diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -3,148 +3,227908 @@
   "best_model_checkpoint": null,
   "epoch": 1.0,
   "eval_steps": 500,
-  "global_step": 20,
+  "global_step": 37980,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.05,
-      "learning_rate": 2e-05,
-      "loss": 5.3746,
+      "epoch": 0.0,
+      "learning_rate": 1.754385964912281e-08,
+      "loss": 2.4553,
       "step": 1
     },
     {
-      "epoch": 0.1,
-      "learning_rate": 1.9863613034027224e-05,
-      "loss": 5.2868,
+      "epoch": 0.0,
+      "learning_rate": 3.508771929824562e-08,
+      "loss": 2.7407,
       "step": 2
     },
     {
-      "epoch": 0.15,
-      "learning_rate": 1.9458172417006347e-05,
-      "loss": 4.0309,
+      "epoch": 0.0,
+      "learning_rate": 5.263157894736842e-08,
+      "loss": 3.1742,
       "step": 3
     },
     {
-      "epoch": 0.2,
-      "learning_rate": 1.879473751206489e-05,
-      "loss": 3.4518,
+      "epoch": 0.0,
+      "learning_rate": 7.017543859649123e-08,
+      "loss": 3.0235,
       "step": 4
     },
     {
-      "epoch": 0.25,
-      "learning_rate": 1.789140509396394e-05,
-      "loss": 2.8369,
+      "epoch": 0.0,
+      "learning_rate": 8.771929824561404e-08,
+      "loss": 3.6353,
       "step": 5
     },
     {
-      "epoch": 0.3,
-      "learning_rate": 1.6772815716257414e-05,
-      "loss": 2.2651,
+      "epoch": 0.0,
+      "learning_rate": 1.0526315789473685e-07,
+      "loss": 2.4897,
       "step": 6
     },
     {
-      "epoch": 0.35,
-      "learning_rate": 1.5469481581224274e-05,
-      "loss": 2.1772,
+      "epoch": 0.0,
+      "learning_rate": 1.2280701754385966e-07,
+      "loss": 2.6602,
       "step": 7
     },
     {
-      "epoch": 0.4,
-      "learning_rate": 1.4016954246529697e-05,
-      "loss": 1.9791,
+      "epoch": 0.0,
+      "learning_rate": 1.4035087719298247e-07,
+      "loss": 2.6685,
       "step": 8
     },
     {
-      "epoch": 0.45,
-      "learning_rate": 1.2454854871407993e-05,
-      "loss": 1.8638,
+      "epoch": 0.0,
+      "learning_rate": 1.5789473684210527e-07,
+      "loss": 2.7283,
       "step": 9
     },
     {
-      "epoch": 0.5,
-      "learning_rate": 1.0825793454723325e-05,
-      "loss": 1.8646,
+      "epoch": 0.0,
+      "learning_rate": 1.7543859649122808e-07,
+      "loss": 2.8069,
       "step": 10
     },
     {
-      "epoch": 0.55,
-      "learning_rate": 9.174206545276678e-06,
-      "loss": 1.6432,
+      "epoch": 0.0,
+      "learning_rate": 1.929824561403509e-07,
+      "loss": 1.0193,
       "step": 11
     },
     {
-      "epoch": 0.6,
-      "learning_rate": 7.545145128592009e-06,
-      "loss": 1.6805,
+      "epoch": 0.0,
+      "learning_rate": 2.105263157894737e-07,
+      "loss": 2.6766,
       "step": 12
     },
     {
-      "epoch": 0.65,
-      "learning_rate": 5.983045753470308e-06,
-      "loss": 1.6166,
+      "epoch": 0.0,
+      "learning_rate": 2.2807017543859652e-07,
+      "loss": 2.7424,
       "step": 13
     },
     {
-      "epoch": 0.7,
-      "learning_rate": 4.530518418775734e-06,
-      "loss": 1.6272,
+      "epoch": 0.0,
+      "learning_rate": 2.4561403508771933e-07,
+      "loss": 2.7001,
       "step": 14
     },
     {
-      "epoch": 0.75,
-      "learning_rate": 3.2271842837425917e-06,
-      "loss": 1.5397,
+      "epoch": 0.0,
+      "learning_rate": 2.6315789473684213e-07,
+      "loss": 2.9494,
       "step": 15
     },
     {
-      "epoch": 0.8,
-      "learning_rate": 2.1085949060360654e-06,
-      "loss": 1.6125,
+      "epoch": 0.0,
+      "learning_rate": 2.8070175438596494e-07,
+      "loss": 3.2771,
       "step": 16
     },
     {
-      "epoch": 0.85,
-      "learning_rate": 1.2052624879351105e-06,
-      "loss": 1.4847,
+      "epoch": 0.0,
+      "learning_rate": 2.9824561403508774e-07,
+      "loss": 2.8604,
       "step": 17
     },
     {
-      "epoch": 0.9,
-      "learning_rate": 5.418275829936537e-07,
-      "loss": 1.5863,
+      "epoch": 0.0,
+      "learning_rate": 3.1578947368421055e-07,
+      "loss": 2.345,
       "step": 18
     },
     {
-      "epoch": 0.95,
-      "learning_rate": 1.3638696597277678e-07,
-      "loss": 1.5869,
+      "epoch": 0.0,
+      "learning_rate": 3.3333333333333335e-07,
+      "loss": 2.3108,
       "step": 19
     },
     {
-      "epoch": 1.0,
-      "learning_rate": 0.0,
-      "loss": 1.5296,
+      "epoch": 0.0,
+      "learning_rate": 3.5087719298245616e-07,
+      "loss": 2.6492,
       "step": 20
     },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.6842105263157896e-07,
+      "loss": 2.3272,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.859649122807018e-07,
+      "loss": 2.228,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.035087719298246e-07,
+      "loss": 2.8794,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.210526315789474e-07,
+      "loss": 2.0903,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.385964912280702e-07,
+      "loss": 2.3888,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5614035087719304e-07,
+      "loss": 1.9659,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.7368421052631585e-07,
+      "loss": 2.116,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.912280701754387e-07,
+      "loss": 2.1326,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.087719298245614e-07,
+      "loss": 1.9259,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.263157894736843e-07,
+      "loss": 1.8884,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.438596491228071e-07,
+      "loss": 1.8779,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.614035087719299e-07,
+      "loss": 1.7505,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.789473684210526e-07,
+      "loss": 1.6648,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.964912280701755e-07,
+      "loss": 2.085,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.140350877192982e-07,
+      "loss": 1.5822,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.315789473684211e-07,
+      "loss": 1.5436,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.491228070175438e-07,
+      "loss": 1.3527,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.666666666666667e-07,
+      "loss": 1.5822,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.842105263157896e-07,
+      "loss": 1.5135,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.017543859649123e-07,
+      "loss": 1.4301,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.192982456140352e-07,
+      "loss": 1.5832,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.368421052631579e-07,
+      "loss": 1.3545,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.543859649122808e-07,
+      "loss": 1.4589,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.719298245614036e-07,
+      "loss": 1.6706,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.894736842105263e-07,
+      "loss": 1.3748,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.070175438596491e-07,
+      "loss": 1.3513,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.245614035087721e-07,
+      "loss": 1.5239,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.421052631578948e-07,
+      "loss": 1.1275,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.596491228070176e-07,
+      "loss": 1.1408,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.771929824561404e-07,
+      "loss": 1.2613,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.947368421052632e-07,
+      "loss": 1.0288,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.122807017543861e-07,
+      "loss": 0.9299,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.298245614035088e-07,
+      "loss": 1.2694,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.473684210526317e-07,
+      "loss": 1.2684,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.649122807017545e-07,
+      "loss": 1.0341,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.824561403508773e-07,
+      "loss": 1.2754,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 1.3553,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0175438596491228e-06,
+      "loss": 1.1791,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0350877192982457e-06,
+      "loss": 1.3598,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0526315789473685e-06,
+      "loss": 1.3786,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0701754385964914e-06,
+      "loss": 1.1471,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0877192982456142e-06,
+      "loss": 1.1717,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1052631578947369e-06,
+      "loss": 1.2091,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1228070175438598e-06,
+      "loss": 1.141,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1403508771929824e-06,
+      "loss": 0.8788,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1578947368421053e-06,
+      "loss": 1.0374,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1754385964912281e-06,
+      "loss": 1.138,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.192982456140351e-06,
+      "loss": 0.9808,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2105263157894738e-06,
+      "loss": 1.2457,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2280701754385965e-06,
+      "loss": 1.159,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2456140350877193e-06,
+      "loss": 1.2903,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2631578947368422e-06,
+      "loss": 1.0387,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.280701754385965e-06,
+      "loss": 1.3623,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2982456140350877e-06,
+      "loss": 0.8376,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3157894736842106e-06,
+      "loss": 1.1285,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3333333333333334e-06,
+      "loss": 1.2805,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3508771929824563e-06,
+      "loss": 0.9176,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3684210526315791e-06,
+      "loss": 0.9612,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3859649122807018e-06,
+      "loss": 0.96,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4035087719298246e-06,
+      "loss": 1.0998,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4210526315789475e-06,
+      "loss": 1.2459,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4385964912280704e-06,
+      "loss": 1.0116,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4561403508771932e-06,
+      "loss": 1.2506,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4736842105263159e-06,
+      "loss": 0.9631,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4912280701754387e-06,
+      "loss": 1.2038,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5087719298245616e-06,
+      "loss": 1.2054,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5263157894736844e-06,
+      "loss": 0.8947,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5438596491228073e-06,
+      "loss": 1.3398,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5614035087719301e-06,
+      "loss": 1.0866,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5789473684210526e-06,
+      "loss": 0.9898,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5964912280701754e-06,
+      "loss": 1.1466,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6140350877192983e-06,
+      "loss": 0.9029,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6315789473684212e-06,
+      "loss": 1.0103,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6491228070175442e-06,
+      "loss": 1.0653,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 0.9758,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6842105263157895e-06,
+      "loss": 1.0311,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7017543859649124e-06,
+      "loss": 1.1393,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7192982456140352e-06,
+      "loss": 1.1125,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.736842105263158e-06,
+      "loss": 1.2485,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7543859649122807e-06,
+      "loss": 1.1559,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7719298245614036e-06,
+      "loss": 0.9081,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7894736842105265e-06,
+      "loss": 1.0836,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8070175438596493e-06,
+      "loss": 1.3207,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8245614035087722e-06,
+      "loss": 1.0928,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8421052631578948e-06,
+      "loss": 1.2081,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8596491228070177e-06,
+      "loss": 1.2455,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8771929824561405e-06,
+      "loss": 0.9298,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8947368421052634e-06,
+      "loss": 1.2974,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9122807017543863e-06,
+      "loss": 0.9804,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.929824561403509e-06,
+      "loss": 1.0366,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9473684210526315e-06,
+      "loss": 0.9612,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9649122807017546e-06,
+      "loss": 1.2176,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9824561403508773e-06,
+      "loss": 1.2051,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.0567,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.017543859649123e-06,
+      "loss": 0.8282,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0350877192982456e-06,
+      "loss": 1.1505,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0526315789473687e-06,
+      "loss": 0.9004,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0701754385964913e-06,
+      "loss": 1.4175,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0877192982456144e-06,
+      "loss": 0.8346,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.105263157894737e-06,
+      "loss": 0.8629,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1228070175438597e-06,
+      "loss": 1.1002,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1403508771929828e-06,
+      "loss": 1.4228,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1578947368421054e-06,
+      "loss": 0.9035,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1754385964912285e-06,
+      "loss": 0.9597,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.192982456140351e-06,
+      "loss": 1.2813,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2105263157894738e-06,
+      "loss": 1.1198,
+      "step": 126
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.228070175438597e-06,
+      "loss": 0.944,
+      "step": 127
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2456140350877195e-06,
+      "loss": 1.09,
+      "step": 128
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2631578947368426e-06,
+      "loss": 1.2098,
+      "step": 129
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.280701754385965e-06,
+      "loss": 1.0782,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.298245614035088e-06,
+      "loss": 0.8734,
+      "step": 131
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3157894736842105e-06,
+      "loss": 0.943,
+      "step": 132
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3333333333333336e-06,
+      "loss": 1.0307,
+      "step": 133
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3508771929824562e-06,
+      "loss": 1.4726,
+      "step": 134
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.368421052631579e-06,
+      "loss": 0.8457,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.385964912280702e-06,
+      "loss": 1.0601,
+      "step": 136
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4035087719298246e-06,
+      "loss": 0.9174,
+      "step": 137
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4210526315789477e-06,
+      "loss": 1.1202,
+      "step": 138
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4385964912280703e-06,
+      "loss": 1.3195,
+      "step": 139
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.456140350877193e-06,
+      "loss": 1.0128,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.473684210526316e-06,
+      "loss": 1.2643,
+      "step": 141
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4912280701754387e-06,
+      "loss": 1.1037,
+      "step": 142
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5087719298245617e-06,
+      "loss": 1.1866,
+      "step": 143
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5263157894736844e-06,
+      "loss": 1.0132,
+      "step": 144
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5438596491228075e-06,
+      "loss": 0.9409,
+      "step": 145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.56140350877193e-06,
+      "loss": 1.0844,
+      "step": 146
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.578947368421053e-06,
+      "loss": 1.1117,
+      "step": 147
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5964912280701754e-06,
+      "loss": 1.1306,
+      "step": 148
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6140350877192985e-06,
+      "loss": 1.0374,
+      "step": 149
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.631578947368421e-06,
+      "loss": 1.0534,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.649122807017544e-06,
+      "loss": 0.9418,
+      "step": 151
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.666666666666667e-06,
+      "loss": 1.2872,
+      "step": 152
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.68421052631579e-06,
+      "loss": 1.0378,
+      "step": 153
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7017543859649125e-06,
+      "loss": 0.8894,
+      "step": 154
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7192982456140356e-06,
+      "loss": 1.0686,
+      "step": 155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7368421052631583e-06,
+      "loss": 1.1848,
+      "step": 156
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7543859649122813e-06,
+      "loss": 1.1022,
+      "step": 157
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7719298245614036e-06,
+      "loss": 1.0688,
+      "step": 158
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.789473684210526e-06,
+      "loss": 0.7569,
+      "step": 159
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8070175438596493e-06,
+      "loss": 0.9698,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.824561403508772e-06,
+      "loss": 0.9315,
+      "step": 161
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.842105263157895e-06,
+      "loss": 1.1632,
+      "step": 162
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8596491228070176e-06,
+      "loss": 0.8597,
+      "step": 163
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8771929824561407e-06,
+      "loss": 1.0302,
+      "step": 164
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8947368421052634e-06,
+      "loss": 1.2456,
+      "step": 165
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9122807017543864e-06,
+      "loss": 1.2164,
+      "step": 166
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.929824561403509e-06,
+      "loss": 0.9414,
+      "step": 167
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9473684210526317e-06,
+      "loss": 1.0536,
+      "step": 168
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9649122807017544e-06,
+      "loss": 1.1492,
+      "step": 169
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9824561403508774e-06,
+      "loss": 1.1076,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-06,
+      "loss": 1.1929,
+      "step": 171
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.017543859649123e-06,
+      "loss": 0.8958,
+      "step": 172
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.035087719298246e-06,
+      "loss": 1.0271,
+      "step": 173
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.052631578947369e-06,
+      "loss": 0.9348,
+      "step": 174
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0701754385964915e-06,
+      "loss": 0.9908,
+      "step": 175
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0877192982456146e-06,
+      "loss": 0.8442,
+      "step": 176
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1052631578947372e-06,
+      "loss": 1.1296,
+      "step": 177
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1228070175438603e-06,
+      "loss": 0.7438,
+      "step": 178
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1403508771929825e-06,
+      "loss": 1.0857,
+      "step": 179
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.157894736842105e-06,
+      "loss": 1.1387,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1754385964912282e-06,
+      "loss": 0.9243,
+      "step": 181
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.192982456140351e-06,
+      "loss": 1.0735,
+      "step": 182
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.210526315789474e-06,
+      "loss": 1.1872,
+      "step": 183
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2280701754385966e-06,
+      "loss": 0.7687,
+      "step": 184
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2456140350877197e-06,
+      "loss": 0.9979,
+      "step": 185
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2631578947368423e-06,
+      "loss": 0.8387,
+      "step": 186
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2807017543859654e-06,
+      "loss": 1.0804,
+      "step": 187
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2982456140350885e-06,
+      "loss": 1.1657,
+      "step": 188
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3157894736842107e-06,
+      "loss": 1.0541,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 0.7683,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3508771929824564e-06,
+      "loss": 1.0799,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.368421052631579e-06,
+      "loss": 0.814,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.385964912280702e-06,
+      "loss": 0.7937,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4035087719298248e-06,
+      "loss": 1.0126,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.421052631578948e-06,
+      "loss": 1.2818,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4385964912280705e-06,
+      "loss": 0.8654,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4561403508771935e-06,
+      "loss": 0.9535,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.473684210526316e-06,
+      "loss": 1.0684,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.491228070175439e-06,
+      "loss": 1.0133,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5087719298245615e-06,
+      "loss": 1.1335,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5263157894736846e-06,
+      "loss": 0.8871,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.543859649122807e-06,
+      "loss": 0.8797,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5614035087719303e-06,
+      "loss": 1.2031,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.578947368421053e-06,
+      "loss": 1.3156,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.596491228070176e-06,
+      "loss": 1.1056,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6140350877192986e-06,
+      "loss": 1.1376,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6315789473684217e-06,
+      "loss": 0.9263,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6491228070175443e-06,
+      "loss": 0.9317,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6666666666666666e-06,
+      "loss": 0.9455,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6842105263157896e-06,
+      "loss": 0.9299,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7017543859649123e-06,
+      "loss": 1.2,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7192982456140354e-06,
+      "loss": 0.7953,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.736842105263158e-06,
+      "loss": 0.9019,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.754385964912281e-06,
+      "loss": 1.0189,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7719298245614037e-06,
+      "loss": 1.1987,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.789473684210527e-06,
+      "loss": 0.9483,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8070175438596494e-06,
+      "loss": 0.7918,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8245614035087725e-06,
+      "loss": 1.084,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.842105263157895e-06,
+      "loss": 0.8475,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.859649122807018e-06,
+      "loss": 0.9166,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.877192982456141e-06,
+      "loss": 0.8969,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.894736842105263e-06,
+      "loss": 0.8632,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.912280701754386e-06,
+      "loss": 0.8542,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.929824561403509e-06,
+      "loss": 1.0861,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.947368421052632e-06,
+      "loss": 0.9943,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9649122807017545e-06,
+      "loss": 1.1564,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.982456140350878e-06,
+      "loss": 1.0274,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.1416,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.017543859649124e-06,
+      "loss": 1.1089,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.035087719298246e-06,
+      "loss": 0.9789,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.052631578947368e-06,
+      "loss": 1.0341,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.070175438596491e-06,
+      "loss": 1.1497,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.087719298245614e-06,
+      "loss": 0.7577,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.105263157894737e-06,
+      "loss": 0.8809,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.12280701754386e-06,
+      "loss": 1.0341,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.140350877192983e-06,
+      "loss": 0.9863,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.157894736842106e-06,
+      "loss": 1.115,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.175438596491229e-06,
+      "loss": 0.9441,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.192982456140352e-06,
+      "loss": 1.0188,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.210526315789474e-06,
+      "loss": 0.9384,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.228070175438596e-06,
+      "loss": 1.0152,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.245614035087719e-06,
+      "loss": 1.214,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2631578947368425e-06,
+      "loss": 0.7929,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2807017543859656e-06,
+      "loss": 1.0282,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.298245614035088e-06,
+      "loss": 1.1079,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.315789473684211e-06,
+      "loss": 0.8803,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.333333333333334e-06,
+      "loss": 0.9598,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.350877192982457e-06,
+      "loss": 1.0308,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.368421052631579e-06,
+      "loss": 1.0645,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.385964912280702e-06,
+      "loss": 1.1526,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4035087719298245e-06,
+      "loss": 1.0251,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4210526315789476e-06,
+      "loss": 1.1339,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.438596491228071e-06,
+      "loss": 1.0763,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.456140350877194e-06,
+      "loss": 1.418,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.473684210526316e-06,
+      "loss": 1.0536,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.491228070175439e-06,
+      "loss": 1.0246,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.508771929824562e-06,
+      "loss": 1.0035,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.526315789473685e-06,
+      "loss": 0.8301,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.543859649122807e-06,
+      "loss": 1.1962,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.56140350877193e-06,
+      "loss": 0.9595,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.578947368421053e-06,
+      "loss": 0.8101,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.596491228070176e-06,
+      "loss": 0.9741,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.614035087719299e-06,
+      "loss": 0.8653,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.631578947368421e-06,
+      "loss": 0.7454,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.649122807017544e-06,
+      "loss": 1.1826,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.666666666666667e-06,
+      "loss": 1.1312,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.68421052631579e-06,
+      "loss": 1.0761,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7017543859649125e-06,
+      "loss": 1.0245,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7192982456140355e-06,
+      "loss": 1.0773,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.736842105263158e-06,
+      "loss": 1.124,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.754385964912281e-06,
+      "loss": 1.0003,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.771929824561404e-06,
+      "loss": 1.0588,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.789473684210527e-06,
+      "loss": 0.832,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.807017543859649e-06,
+      "loss": 1.187,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.824561403508772e-06,
+      "loss": 1.0829,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.842105263157895e-06,
+      "loss": 0.6175,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.859649122807018e-06,
+      "loss": 1.0684,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.877192982456141e-06,
+      "loss": 1.0636,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.894736842105264e-06,
+      "loss": 0.9408,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.912280701754386e-06,
+      "loss": 0.856,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.929824561403509e-06,
+      "loss": 1.0438,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.947368421052632e-06,
+      "loss": 1.0124,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.964912280701754e-06,
+      "loss": 0.9888,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.982456140350877e-06,
+      "loss": 1.1487,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5e-06,
+      "loss": 1.0929,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0175438596491235e-06,
+      "loss": 0.794,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0350877192982465e-06,
+      "loss": 1.0421,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.052631578947369e-06,
+      "loss": 0.9812,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.070175438596492e-06,
+      "loss": 0.7759,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.087719298245615e-06,
+      "loss": 0.9864,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.105263157894738e-06,
+      "loss": 0.9828,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.12280701754386e-06,
+      "loss": 0.9838,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.140350877192983e-06,
+      "loss": 0.9338,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.157894736842106e-06,
+      "loss": 1.1047,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.175438596491229e-06,
+      "loss": 1.0237,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.192982456140351e-06,
+      "loss": 0.9353,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.210526315789474e-06,
+      "loss": 1.2348,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.228070175438597e-06,
+      "loss": 1.0197,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.245614035087719e-06,
+      "loss": 0.8202,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.263157894736842e-06,
+      "loss": 0.7498,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.280701754385965e-06,
+      "loss": 0.9979,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.298245614035088e-06,
+      "loss": 0.9351,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.315789473684211e-06,
+      "loss": 1.0506,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.333333333333334e-06,
+      "loss": 0.9962,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.350877192982457e-06,
+      "loss": 0.8879,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.36842105263158e-06,
+      "loss": 1.0515,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.385964912280702e-06,
+      "loss": 0.9529,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.403508771929825e-06,
+      "loss": 0.9557,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.421052631578948e-06,
+      "loss": 1.0944,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.438596491228071e-06,
+      "loss": 0.9321,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4561403508771935e-06,
+      "loss": 1.0391,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4736842105263165e-06,
+      "loss": 1.3102,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.49122807017544e-06,
+      "loss": 1.1362,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.508771929824563e-06,
+      "loss": 0.7733,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.526315789473685e-06,
+      "loss": 0.8421,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.543859649122807e-06,
+      "loss": 1.0246,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.56140350877193e-06,
+      "loss": 0.831,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.578947368421052e-06,
+      "loss": 1.1663,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.5964912280701755e-06,
+      "loss": 1.0598,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.6140350877192985e-06,
+      "loss": 1.107,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.631578947368422e-06,
+      "loss": 1.15,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.649122807017544e-06,
+      "loss": 0.9667,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.666666666666667e-06,
+      "loss": 0.9738,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.68421052631579e-06,
+      "loss": 0.9258,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.701754385964913e-06,
+      "loss": 0.7725,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.719298245614035e-06,
+      "loss": 0.8671,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.736842105263158e-06,
+      "loss": 1.2286,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.754385964912281e-06,
+      "loss": 1.2729,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7719298245614045e-06,
+      "loss": 1.2309,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.789473684210527e-06,
+      "loss": 0.7855,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.80701754385965e-06,
+      "loss": 0.8286,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.824561403508773e-06,
+      "loss": 1.0167,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.842105263157896e-06,
+      "loss": 0.891,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.859649122807018e-06,
+      "loss": 1.0125,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.877192982456141e-06,
+      "loss": 0.9263,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.8947368421052634e-06,
+      "loss": 0.9213,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.912280701754386e-06,
+      "loss": 0.7738,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.929824561403509e-06,
+      "loss": 1.005,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.947368421052632e-06,
+      "loss": 1.0645,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.964912280701755e-06,
+      "loss": 0.7726,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.982456140350877e-06,
+      "loss": 0.9564,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6e-06,
+      "loss": 0.941,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.017543859649123e-06,
+      "loss": 1.0222,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.035087719298246e-06,
+      "loss": 0.9603,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.0526315789473685e-06,
+      "loss": 1.0067,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.070175438596492e-06,
+      "loss": 0.7244,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.087719298245615e-06,
+      "loss": 1.1001,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.105263157894738e-06,
+      "loss": 1.1305,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.12280701754386e-06,
+      "loss": 0.993,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.140350877192983e-06,
+      "loss": 1.0082,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.157894736842106e-06,
+      "loss": 0.8265,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.175438596491229e-06,
+      "loss": 0.8173,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.192982456140351e-06,
+      "loss": 1.0473,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2105263157894745e-06,
+      "loss": 1.3216,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.2280701754385975e-06,
+      "loss": 0.7989,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.245614035087721e-06,
+      "loss": 1.1065,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.263157894736842e-06,
+      "loss": 1.1044,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.280701754385965e-06,
+      "loss": 1.0113,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.298245614035088e-06,
+      "loss": 0.5607,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.31578947368421e-06,
+      "loss": 0.7764,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.333333333333333e-06,
+      "loss": 1.1068,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3508771929824565e-06,
+      "loss": 0.8719,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3684210526315795e-06,
+      "loss": 1.0126,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.385964912280702e-06,
+      "loss": 0.9184,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.403508771929825e-06,
+      "loss": 0.9408,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.421052631578948e-06,
+      "loss": 0.9394,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.438596491228071e-06,
+      "loss": 1.1086,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.456140350877193e-06,
+      "loss": 1.1914,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.473684210526316e-06,
+      "loss": 1.1323,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.491228070175439e-06,
+      "loss": 1.0355,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.508771929824562e-06,
+      "loss": 0.8802,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.526315789473685e-06,
+      "loss": 0.7172,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.543859649122808e-06,
+      "loss": 1.0935,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.561403508771931e-06,
+      "loss": 0.9968,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.578947368421054e-06,
+      "loss": 1.0513,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.596491228070177e-06,
+      "loss": 1.0749,
+      "step": 376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.614035087719298e-06,
+      "loss": 0.6964,
+      "step": 377
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.631578947368421e-06,
+      "loss": 0.9842,
+      "step": 378
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.649122807017544e-06,
+      "loss": 1.1165,
+      "step": 379
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.7735,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.68421052631579e-06,
+      "loss": 0.8401,
+      "step": 381
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.701754385964913e-06,
+      "loss": 0.7087,
+      "step": 382
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.719298245614035e-06,
+      "loss": 1.0295,
+      "step": 383
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.736842105263158e-06,
+      "loss": 1.0347,
+      "step": 384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.754385964912281e-06,
+      "loss": 0.8948,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.771929824561404e-06,
+      "loss": 0.8958,
+      "step": 386
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.789473684210527e-06,
+      "loss": 1.2142,
+      "step": 387
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.8070175438596495e-06,
+      "loss": 1.1371,
+      "step": 388
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.824561403508773e-06,
+      "loss": 0.9696,
+      "step": 389
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.842105263157896e-06,
+      "loss": 1.0501,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.859649122807019e-06,
+      "loss": 1.1613,
+      "step": 391
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.877192982456141e-06,
+      "loss": 0.7816,
+      "step": 392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.894736842105264e-06,
+      "loss": 0.9566,
+      "step": 393
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.912280701754387e-06,
+      "loss": 1.0612,
+      "step": 394
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.92982456140351e-06,
+      "loss": 0.816,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.947368421052632e-06,
+      "loss": 1.17,
+      "step": 396
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.964912280701755e-06,
+      "loss": 1.2625,
+      "step": 397
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.982456140350878e-06,
+      "loss": 1.0696,
+      "step": 398
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7e-06,
+      "loss": 0.9628,
+      "step": 399
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.017543859649123e-06,
+      "loss": 0.9179,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.035087719298246e-06,
+      "loss": 1.2696,
+      "step": 401
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.052631578947369e-06,
+      "loss": 0.7797,
+      "step": 402
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.070175438596491e-06,
+      "loss": 0.891,
+      "step": 403
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.087719298245614e-06,
+      "loss": 1.0193,
+      "step": 404
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1052631578947375e-06,
+      "loss": 1.0222,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1228070175438605e-06,
+      "loss": 1.1458,
+      "step": 406
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.140350877192983e-06,
+      "loss": 1.1716,
+      "step": 407
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.157894736842106e-06,
+      "loss": 0.8149,
+      "step": 408
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.175438596491229e-06,
+      "loss": 1.0436,
+      "step": 409
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.192982456140352e-06,
+      "loss": 0.7327,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.210526315789474e-06,
+      "loss": 1.1869,
+      "step": 411
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.228070175438597e-06,
+      "loss": 1.1745,
+      "step": 412
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.24561403508772e-06,
+      "loss": 1.1948,
+      "step": 413
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.263157894736843e-06,
+      "loss": 1.0847,
+      "step": 414
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.280701754385966e-06,
+      "loss": 0.8019,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.298245614035089e-06,
+      "loss": 0.7411,
+      "step": 416
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.315789473684212e-06,
+      "loss": 0.987,
+      "step": 417
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.333333333333333e-06,
+      "loss": 0.7634,
+      "step": 418
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.350877192982456e-06,
+      "loss": 1.1453,
+      "step": 419
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.368421052631579e-06,
+      "loss": 0.9786,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.385964912280702e-06,
+      "loss": 0.837,
+      "step": 421
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.4035087719298246e-06,
+      "loss": 1.1365,
+      "step": 422
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.421052631578948e-06,
+      "loss": 1.0242,
+      "step": 423
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.438596491228071e-06,
+      "loss": 0.8391,
+      "step": 424
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.456140350877194e-06,
+      "loss": 1.1639,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.473684210526316e-06,
+      "loss": 0.804,
+      "step": 426
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.491228070175439e-06,
+      "loss": 1.1374,
+      "step": 427
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.508771929824562e-06,
+      "loss": 1.1561,
+      "step": 428
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.526315789473685e-06,
+      "loss": 1.0948,
+      "step": 429
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.5438596491228074e-06,
+      "loss": 1.3439,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.5614035087719305e-06,
+      "loss": 1.1546,
+      "step": 431
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.578947368421054e-06,
+      "loss": 1.2243,
+      "step": 432
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.596491228070177e-06,
+      "loss": 1.0803,
+      "step": 433
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.614035087719299e-06,
+      "loss": 0.9577,
+      "step": 434
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.631578947368423e-06,
+      "loss": 1.2241,
+      "step": 435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.649122807017545e-06,
+      "loss": 0.989,
+      "step": 436
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.666666666666667e-06,
+      "loss": 0.9005,
+      "step": 437
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.68421052631579e-06,
+      "loss": 1.0309,
+      "step": 438
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.701754385964912e-06,
+      "loss": 1.0667,
+      "step": 439
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.719298245614036e-06,
+      "loss": 1.2136,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.736842105263158e-06,
+      "loss": 0.99,
+      "step": 441
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.754385964912282e-06,
+      "loss": 1.0694,
+      "step": 442
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.771929824561404e-06,
+      "loss": 0.8888,
+      "step": 443
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.789473684210526e-06,
+      "loss": 1.2299,
+      "step": 444
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.80701754385965e-06,
+      "loss": 0.9102,
+      "step": 445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.824561403508772e-06,
+      "loss": 0.7771,
+      "step": 446
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.842105263157895e-06,
+      "loss": 1.0669,
+      "step": 447
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.859649122807018e-06,
+      "loss": 1.0971,
+      "step": 448
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.87719298245614e-06,
+      "loss": 0.9157,
+      "step": 449
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.894736842105265e-06,
+      "loss": 1.1579,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.912280701754387e-06,
+      "loss": 1.0989,
+      "step": 451
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.929824561403509e-06,
+      "loss": 0.8605,
+      "step": 452
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.947368421052633e-06,
+      "loss": 1.1423,
+      "step": 453
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.964912280701755e-06,
+      "loss": 1.1699,
+      "step": 454
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.982456140350877e-06,
+      "loss": 0.9065,
+      "step": 455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.0492,
+      "step": 456
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.017543859649124e-06,
+      "loss": 0.9314,
+      "step": 457
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.035087719298247e-06,
+      "loss": 0.9839,
+      "step": 458
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.052631578947368e-06,
+      "loss": 0.9035,
+      "step": 459
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.070175438596492e-06,
+      "loss": 0.8354,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.087719298245614e-06,
+      "loss": 1.0902,
+      "step": 461
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.105263157894736e-06,
+      "loss": 0.968,
+      "step": 462
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.12280701754386e-06,
+      "loss": 0.9311,
+      "step": 463
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.140350877192983e-06,
+      "loss": 1.0997,
+      "step": 464
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.157894736842106e-06,
+      "loss": 1.1562,
+      "step": 465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.175438596491229e-06,
+      "loss": 1.1325,
+      "step": 466
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.192982456140351e-06,
+      "loss": 1.1539,
+      "step": 467
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.210526315789475e-06,
+      "loss": 1.1757,
+      "step": 468
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.228070175438597e-06,
+      "loss": 1.0262,
+      "step": 469
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.24561403508772e-06,
+      "loss": 1.031,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.263157894736843e-06,
+      "loss": 0.8904,
+      "step": 471
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.280701754385965e-06,
+      "loss": 0.7997,
+      "step": 472
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.29824561403509e-06,
+      "loss": 1.0904,
+      "step": 473
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.315789473684212e-06,
+      "loss": 1.2179,
+      "step": 474
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 0.9553,
+      "step": 475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.350877192982458e-06,
+      "loss": 0.9374,
+      "step": 476
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.36842105263158e-06,
+      "loss": 1.219,
+      "step": 477
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.385964912280704e-06,
+      "loss": 0.9565,
+      "step": 478
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.403508771929824e-06,
+      "loss": 0.8888,
+      "step": 479
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.421052631578948e-06,
+      "loss": 1.0138,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.43859649122807e-06,
+      "loss": 1.174,
+      "step": 481
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.456140350877193e-06,
+      "loss": 1.0369,
+      "step": 482
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.473684210526317e-06,
+      "loss": 1.0004,
+      "step": 483
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.491228070175439e-06,
+      "loss": 0.7685,
+      "step": 484
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.508771929824563e-06,
+      "loss": 0.9493,
+      "step": 485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.526315789473685e-06,
+      "loss": 1.2283,
+      "step": 486
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.543859649122807e-06,
+      "loss": 1.0672,
+      "step": 487
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.561403508771931e-06,
+      "loss": 1.02,
+      "step": 488
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.578947368421053e-06,
+      "loss": 0.8602,
+      "step": 489
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.596491228070176e-06,
+      "loss": 0.9713,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.6140350877193e-06,
+      "loss": 1.0458,
+      "step": 491
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.631578947368422e-06,
+      "loss": 1.0429,
+      "step": 492
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.649122807017546e-06,
+      "loss": 1.1397,
+      "step": 493
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.666666666666668e-06,
+      "loss": 1.156,
+      "step": 494
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.68421052631579e-06,
+      "loss": 1.0525,
+      "step": 495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.701754385964914e-06,
+      "loss": 0.9308,
+      "step": 496
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.719298245614036e-06,
+      "loss": 1.0383,
+      "step": 497
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.736842105263158e-06,
+      "loss": 1.0441,
+      "step": 498
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.75438596491228e-06,
+      "loss": 1.1479,
+      "step": 499
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.771929824561405e-06,
+      "loss": 1.0932,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.789473684210527e-06,
+      "loss": 1.2225,
+      "step": 501
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.807017543859649e-06,
+      "loss": 0.9591,
+      "step": 502
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.824561403508773e-06,
+      "loss": 0.977,
+      "step": 503
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.842105263157895e-06,
+      "loss": 1.1872,
+      "step": 504
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.859649122807017e-06,
+      "loss": 1.0394,
+      "step": 505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.877192982456141e-06,
+      "loss": 1.0837,
+      "step": 506
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.894736842105264e-06,
+      "loss": 1.166,
+      "step": 507
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.912280701754387e-06,
+      "loss": 1.0377,
+      "step": 508
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.92982456140351e-06,
+      "loss": 0.7799,
+      "step": 509
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.947368421052632e-06,
+      "loss": 0.9775,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.964912280701756e-06,
+      "loss": 1.2611,
+      "step": 511
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.982456140350878e-06,
+      "loss": 1.198,
+      "step": 512
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9e-06,
+      "loss": 1.0267,
+      "step": 513
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.017543859649124e-06,
+      "loss": 1.1184,
+      "step": 514
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.035087719298246e-06,
+      "loss": 0.8782,
+      "step": 515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.05263157894737e-06,
+      "loss": 0.9168,
+      "step": 516
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.070175438596493e-06,
+      "loss": 0.7386,
+      "step": 517
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.087719298245615e-06,
+      "loss": 0.8853,
+      "step": 518
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.105263157894739e-06,
+      "loss": 0.8122,
+      "step": 519
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.12280701754386e-06,
+      "loss": 1.0856,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.140350877192983e-06,
+      "loss": 1.0078,
+      "step": 521
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.157894736842105e-06,
+      "loss": 1.1776,
+      "step": 522
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.17543859649123e-06,
+      "loss": 0.9467,
+      "step": 523
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.192982456140351e-06,
+      "loss": 1.1719,
+      "step": 524
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.210526315789474e-06,
+      "loss": 0.9298,
+      "step": 525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.228070175438598e-06,
+      "loss": 1.2475,
+      "step": 526
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.24561403508772e-06,
+      "loss": 1.1462,
+      "step": 527
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.263157894736842e-06,
+      "loss": 1.23,
+      "step": 528
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.280701754385966e-06,
+      "loss": 1.2136,
+      "step": 529
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.298245614035088e-06,
+      "loss": 0.9937,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.315789473684212e-06,
+      "loss": 1.2422,
+      "step": 531
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 0.9352,
+      "step": 532
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.350877192982457e-06,
+      "loss": 1.1011,
+      "step": 533
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.36842105263158e-06,
+      "loss": 1.0965,
+      "step": 534
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.385964912280703e-06,
+      "loss": 1.2557,
+      "step": 535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.403508771929825e-06,
+      "loss": 0.9597,
+      "step": 536
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.421052631578949e-06,
+      "loss": 1.104,
+      "step": 537
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.438596491228071e-06,
+      "loss": 0.924,
+      "step": 538
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.456140350877195e-06,
+      "loss": 0.8517,
+      "step": 539
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.473684210526315e-06,
+      "loss": 1.1323,
+      "step": 540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.49122807017544e-06,
+      "loss": 1.0587,
+      "step": 541
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.508771929824562e-06,
+      "loss": 1.0834,
+      "step": 542
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.526315789473684e-06,
+      "loss": 1.0963,
+      "step": 543
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.543859649122808e-06,
+      "loss": 1.0666,
+      "step": 544
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.56140350877193e-06,
+      "loss": 0.7954,
+      "step": 545
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.578947368421054e-06,
+      "loss": 0.8422,
+      "step": 546
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.596491228070176e-06,
+      "loss": 1.1167,
+      "step": 547
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.614035087719298e-06,
+      "loss": 1.1641,
+      "step": 548
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.631578947368422e-06,
+      "loss": 1.0707,
+      "step": 549
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.649122807017545e-06,
+      "loss": 1.016,
+      "step": 550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 0.8258,
+      "step": 551
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.68421052631579e-06,
+      "loss": 0.8962,
+      "step": 552
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.701754385964913e-06,
+      "loss": 1.1016,
+      "step": 553
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.719298245614037e-06,
+      "loss": 1.111,
+      "step": 554
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.736842105263159e-06,
+      "loss": 1.091,
+      "step": 555
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.754385964912281e-06,
+      "loss": 0.8872,
+      "step": 556
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.771929824561405e-06,
+      "loss": 0.9987,
+      "step": 557
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.789473684210527e-06,
+      "loss": 0.9146,
+      "step": 558
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.80701754385965e-06,
+      "loss": 0.9675,
+      "step": 559
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.824561403508772e-06,
+      "loss": 0.9445,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.842105263157896e-06,
+      "loss": 1.1286,
+      "step": 561
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.859649122807018e-06,
+      "loss": 1.1503,
+      "step": 562
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.87719298245614e-06,
+      "loss": 0.9469,
+      "step": 563
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.894736842105264e-06,
+      "loss": 0.7715,
+      "step": 564
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.912280701754386e-06,
+      "loss": 1.1631,
+      "step": 565
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.929824561403509e-06,
+      "loss": 1.2528,
+      "step": 566
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.947368421052632e-06,
+      "loss": 0.8508,
+      "step": 567
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.964912280701755e-06,
+      "loss": 0.8783,
+      "step": 568
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.982456140350879e-06,
+      "loss": 1.038,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1e-05,
+      "loss": 1.1247,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0017543859649123e-05,
+      "loss": 1.0616,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0035087719298247e-05,
+      "loss": 0.8011,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.005263157894737e-05,
+      "loss": 1.0072,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0070175438596493e-05,
+      "loss": 1.0392,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0087719298245615e-05,
+      "loss": 1.2751,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0105263157894738e-05,
+      "loss": 1.1658,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0122807017543861e-05,
+      "loss": 1.015,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0140350877192984e-05,
+      "loss": 1.0745,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0157894736842106e-05,
+      "loss": 0.993,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.017543859649123e-05,
+      "loss": 0.9979,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0192982456140352e-05,
+      "loss": 0.8032,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0210526315789476e-05,
+      "loss": 0.7043,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0228070175438598e-05,
+      "loss": 0.8328,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.024561403508772e-05,
+      "loss": 1.0964,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0263157894736844e-05,
+      "loss": 1.383,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0280701754385967e-05,
+      "loss": 1.1641,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0298245614035089e-05,
+      "loss": 1.0233,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0315789473684213e-05,
+      "loss": 1.0318,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0333333333333335e-05,
+      "loss": 1.0459,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0350877192982459e-05,
+      "loss": 0.9806,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.036842105263158e-05,
+      "loss": 1.2092,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0385964912280702e-05,
+      "loss": 1.084,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0403508771929824e-05,
+      "loss": 1.0944,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0421052631578948e-05,
+      "loss": 0.8433,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.043859649122807e-05,
+      "loss": 0.9611,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0456140350877194e-05,
+      "loss": 1.2509,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0473684210526316e-05,
+      "loss": 1.0595,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0491228070175438e-05,
+      "loss": 1.1423,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0508771929824562e-05,
+      "loss": 0.9849,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0526315789473684e-05,
+      "loss": 0.9554,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0543859649122807e-05,
+      "loss": 1.0978,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.056140350877193e-05,
+      "loss": 0.835,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0578947368421053e-05,
+      "loss": 0.9721,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0596491228070177e-05,
+      "loss": 1.0298,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0614035087719299e-05,
+      "loss": 0.9437,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0631578947368421e-05,
+      "loss": 1.0033,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0649122807017545e-05,
+      "loss": 1.0855,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0666666666666667e-05,
+      "loss": 1.1594,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.068421052631579e-05,
+      "loss": 0.7989,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0701754385964913e-05,
+      "loss": 1.0557,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0719298245614036e-05,
+      "loss": 0.909,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.073684210526316e-05,
+      "loss": 0.9135,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0754385964912282e-05,
+      "loss": 0.9685,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0771929824561404e-05,
+      "loss": 0.8867,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0789473684210528e-05,
+      "loss": 1.1026,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.080701754385965e-05,
+      "loss": 1.0418,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0824561403508772e-05,
+      "loss": 0.8435,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0842105263157896e-05,
+      "loss": 0.9627,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0859649122807019e-05,
+      "loss": 0.9045,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0877192982456142e-05,
+      "loss": 1.1091,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0894736842105265e-05,
+      "loss": 0.8572,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0912280701754387e-05,
+      "loss": 0.908,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.092982456140351e-05,
+      "loss": 0.8069,
+      "step": 623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0947368421052633e-05,
+      "loss": 1.2528,
+      "step": 624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0964912280701755e-05,
+      "loss": 0.9571,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.098245614035088e-05,
+      "loss": 1.4321,
+      "step": 626
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 1.4063,
+      "step": 627
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1017543859649125e-05,
+      "loss": 1.2424,
+      "step": 628
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1035087719298248e-05,
+      "loss": 1.5496,
+      "step": 629
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.105263157894737e-05,
+      "loss": 1.1772,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1070175438596494e-05,
+      "loss": 1.2284,
+      "step": 631
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1087719298245614e-05,
+      "loss": 1.084,
+      "step": 632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1105263157894736e-05,
+      "loss": 1.1721,
+      "step": 633
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.112280701754386e-05,
+      "loss": 0.88,
+      "step": 634
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1140350877192983e-05,
+      "loss": 1.153,
+      "step": 635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1157894736842105e-05,
+      "loss": 1.1246,
+      "step": 636
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1175438596491229e-05,
+      "loss": 1.0723,
+      "step": 637
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1192982456140351e-05,
+      "loss": 0.734,
+      "step": 638
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1210526315789473e-05,
+      "loss": 0.9821,
+      "step": 639
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1228070175438597e-05,
+      "loss": 0.6314,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.124561403508772e-05,
+      "loss": 1.1943,
+      "step": 641
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1263157894736843e-05,
+      "loss": 0.9115,
+      "step": 642
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1280701754385965e-05,
+      "loss": 1.2875,
+      "step": 643
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1298245614035088e-05,
+      "loss": 1.2914,
+      "step": 644
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1315789473684212e-05,
+      "loss": 0.9991,
+      "step": 645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1333333333333334e-05,
+      "loss": 1.0421,
+      "step": 646
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1350877192982456e-05,
+      "loss": 1.1517,
+      "step": 647
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.136842105263158e-05,
+      "loss": 1.0619,
+      "step": 648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1385964912280702e-05,
+      "loss": 1.1333,
+      "step": 649
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1403508771929826e-05,
+      "loss": 0.9568,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1421052631578948e-05,
+      "loss": 0.7587,
+      "step": 651
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.143859649122807e-05,
+      "loss": 1.0956,
+      "step": 652
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1456140350877194e-05,
+      "loss": 1.0195,
+      "step": 653
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1473684210526317e-05,
+      "loss": 0.9802,
+      "step": 654
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1491228070175439e-05,
+      "loss": 1.087,
+      "step": 655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1508771929824563e-05,
+      "loss": 1.1547,
+      "step": 656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1526315789473685e-05,
+      "loss": 1.159,
+      "step": 657
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1543859649122809e-05,
+      "loss": 1.2114,
+      "step": 658
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1561403508771931e-05,
+      "loss": 1.0619,
+      "step": 659
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1578947368421053e-05,
+      "loss": 1.111,
+      "step": 660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1596491228070177e-05,
+      "loss": 0.9684,
+      "step": 661
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.16140350877193e-05,
+      "loss": 1.0002,
+      "step": 662
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1631578947368423e-05,
+      "loss": 1.0705,
+      "step": 663
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1649122807017546e-05,
+      "loss": 0.7251,
+      "step": 664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 1.1814,
+      "step": 665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1684210526315792e-05,
+      "loss": 0.8497,
+      "step": 666
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1701754385964914e-05,
+      "loss": 0.958,
+      "step": 667
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1719298245614036e-05,
+      "loss": 1.1915,
+      "step": 668
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.173684210526316e-05,
+      "loss": 0.9939,
+      "step": 669
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1754385964912282e-05,
+      "loss": 0.9879,
+      "step": 670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1771929824561406e-05,
+      "loss": 0.953,
+      "step": 671
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1789473684210527e-05,
+      "loss": 0.7972,
+      "step": 672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1807017543859649e-05,
+      "loss": 0.8307,
+      "step": 673
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1824561403508771e-05,
+      "loss": 1.0201,
+      "step": 674
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1842105263157895e-05,
+      "loss": 1.0268,
+      "step": 675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1859649122807017e-05,
+      "loss": 1.2065,
+      "step": 676
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.187719298245614e-05,
+      "loss": 1.0361,
+      "step": 677
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1894736842105264e-05,
+      "loss": 1.0281,
+      "step": 678
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1912280701754386e-05,
+      "loss": 0.9515,
+      "step": 679
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.192982456140351e-05,
+      "loss": 1.2576,
+      "step": 680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1947368421052632e-05,
+      "loss": 1.2282,
+      "step": 681
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1964912280701754e-05,
+      "loss": 1.0686,
+      "step": 682
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1982456140350878e-05,
+      "loss": 0.8251,
+      "step": 683
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2e-05,
+      "loss": 1.2014,
+      "step": 684
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2017543859649124e-05,
+      "loss": 1.3139,
+      "step": 685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2035087719298246e-05,
+      "loss": 0.8812,
+      "step": 686
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2052631578947369e-05,
+      "loss": 1.1695,
+      "step": 687
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2070175438596493e-05,
+      "loss": 0.8093,
+      "step": 688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2087719298245615e-05,
+      "loss": 1.0055,
+      "step": 689
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2105263157894737e-05,
+      "loss": 0.9394,
+      "step": 690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2122807017543861e-05,
+      "loss": 0.9175,
+      "step": 691
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2140350877192983e-05,
+      "loss": 1.0844,
+      "step": 692
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2157894736842107e-05,
+      "loss": 0.9257,
+      "step": 693
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.217543859649123e-05,
+      "loss": 1.1569,
+      "step": 694
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2192982456140352e-05,
+      "loss": 0.9716,
+      "step": 695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2210526315789475e-05,
+      "loss": 1.0204,
+      "step": 696
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2228070175438598e-05,
+      "loss": 0.8785,
+      "step": 697
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.224561403508772e-05,
+      "loss": 0.8835,
+      "step": 698
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2263157894736844e-05,
+      "loss": 1.1503,
+      "step": 699
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2280701754385966e-05,
+      "loss": 1.0167,
+      "step": 700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.229824561403509e-05,
+      "loss": 0.969,
+      "step": 701
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2315789473684212e-05,
+      "loss": 0.8668,
+      "step": 702
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2333333333333334e-05,
+      "loss": 1.0675,
+      "step": 703
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2350877192982458e-05,
+      "loss": 0.9727,
+      "step": 704
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.236842105263158e-05,
+      "loss": 1.1161,
+      "step": 705
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2385964912280703e-05,
+      "loss": 0.6736,
+      "step": 706
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2403508771929827e-05,
+      "loss": 1.3186,
+      "step": 707
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2421052631578949e-05,
+      "loss": 1.1132,
+      "step": 708
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2438596491228073e-05,
+      "loss": 1.1712,
+      "step": 709
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2456140350877195e-05,
+      "loss": 1.4023,
+      "step": 710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2473684210526317e-05,
+      "loss": 1.0243,
+      "step": 711
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2491228070175441e-05,
+      "loss": 1.0337,
+      "step": 712
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2508771929824562e-05,
+      "loss": 0.9578,
+      "step": 713
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2526315789473684e-05,
+      "loss": 1.1225,
+      "step": 714
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2543859649122808e-05,
+      "loss": 1.2113,
+      "step": 715
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.256140350877193e-05,
+      "loss": 0.8641,
+      "step": 716
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2578947368421052e-05,
+      "loss": 0.9843,
+      "step": 717
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2596491228070176e-05,
+      "loss": 1.0982,
+      "step": 718
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2614035087719298e-05,
+      "loss": 0.8212,
+      "step": 719
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.263157894736842e-05,
+      "loss": 1.2104,
+      "step": 720
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2649122807017545e-05,
+      "loss": 1.0843,
+      "step": 721
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2666666666666667e-05,
+      "loss": 1.1496,
+      "step": 722
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.268421052631579e-05,
+      "loss": 1.1005,
+      "step": 723
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2701754385964913e-05,
+      "loss": 0.7467,
+      "step": 724
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2719298245614035e-05,
+      "loss": 1.2418,
+      "step": 725
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2736842105263159e-05,
+      "loss": 0.8846,
+      "step": 726
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2754385964912281e-05,
+      "loss": 0.7284,
+      "step": 727
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2771929824561404e-05,
+      "loss": 0.9305,
+      "step": 728
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2789473684210527e-05,
+      "loss": 1.204,
+      "step": 729
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.280701754385965e-05,
+      "loss": 1.4428,
+      "step": 730
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2824561403508774e-05,
+      "loss": 1.1387,
+      "step": 731
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2842105263157896e-05,
+      "loss": 1.304,
+      "step": 732
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2859649122807018e-05,
+      "loss": 1.2665,
+      "step": 733
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2877192982456142e-05,
+      "loss": 1.1927,
+      "step": 734
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2894736842105264e-05,
+      "loss": 1.0822,
+      "step": 735
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2912280701754386e-05,
+      "loss": 1.0656,
+      "step": 736
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.292982456140351e-05,
+      "loss": 0.9589,
+      "step": 737
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2947368421052633e-05,
+      "loss": 1.0357,
+      "step": 738
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2964912280701756e-05,
+      "loss": 1.2652,
+      "step": 739
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2982456140350879e-05,
+      "loss": 1.2202,
+      "step": 740
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 1.2575,
+      "step": 741
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3017543859649125e-05,
+      "loss": 1.2252,
+      "step": 742
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3035087719298247e-05,
+      "loss": 1.042,
+      "step": 743
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.305263157894737e-05,
+      "loss": 0.7601,
+      "step": 744
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3070175438596493e-05,
+      "loss": 0.956,
+      "step": 745
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3087719298245615e-05,
+      "loss": 0.8752,
+      "step": 746
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.310526315789474e-05,
+      "loss": 1.0977,
+      "step": 747
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3122807017543862e-05,
+      "loss": 0.7604,
+      "step": 748
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3140350877192984e-05,
+      "loss": 0.8887,
+      "step": 749
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3157894736842108e-05,
+      "loss": 0.8428,
+      "step": 750
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.317543859649123e-05,
+      "loss": 0.8539,
+      "step": 751
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3192982456140354e-05,
+      "loss": 1.2229,
+      "step": 752
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3210526315789476e-05,
+      "loss": 0.8914,
+      "step": 753
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3228070175438597e-05,
+      "loss": 1.0255,
+      "step": 754
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3245614035087719e-05,
+      "loss": 1.1502,
+      "step": 755
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3263157894736843e-05,
+      "loss": 1.1004,
+      "step": 756
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3280701754385965e-05,
+      "loss": 0.8128,
+      "step": 757
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3298245614035087e-05,
+      "loss": 0.983,
+      "step": 758
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3315789473684211e-05,
+      "loss": 1.2527,
+      "step": 759
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.1375,
+      "step": 760
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3350877192982457e-05,
+      "loss": 0.9598,
+      "step": 761
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.336842105263158e-05,
+      "loss": 0.8718,
+      "step": 762
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3385964912280702e-05,
+      "loss": 1.0775,
+      "step": 763
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3403508771929826e-05,
+      "loss": 0.7762,
+      "step": 764
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3421052631578948e-05,
+      "loss": 0.9881,
+      "step": 765
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.343859649122807e-05,
+      "loss": 0.8535,
+      "step": 766
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3456140350877194e-05,
+      "loss": 1.0815,
+      "step": 767
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3473684210526316e-05,
+      "loss": 0.9137,
+      "step": 768
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.349122807017544e-05,
+      "loss": 0.9329,
+      "step": 769
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3508771929824562e-05,
+      "loss": 0.9995,
+      "step": 770
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3526315789473685e-05,
+      "loss": 1.1985,
+      "step": 771
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3543859649122808e-05,
+      "loss": 1.0167,
+      "step": 772
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.356140350877193e-05,
+      "loss": 0.847,
+      "step": 773
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3578947368421055e-05,
+      "loss": 0.8961,
+      "step": 774
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3596491228070177e-05,
+      "loss": 0.9943,
+      "step": 775
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3614035087719299e-05,
+      "loss": 1.0994,
+      "step": 776
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3631578947368423e-05,
+      "loss": 0.8768,
+      "step": 777
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3649122807017545e-05,
+      "loss": 1.0157,
+      "step": 778
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3666666666666667e-05,
+      "loss": 1.1313,
+      "step": 779
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3684210526315791e-05,
+      "loss": 0.8224,
+      "step": 780
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3701754385964914e-05,
+      "loss": 1.1852,
+      "step": 781
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3719298245614037e-05,
+      "loss": 1.3787,
+      "step": 782
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.373684210526316e-05,
+      "loss": 1.2113,
+      "step": 783
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3754385964912282e-05,
+      "loss": 0.9688,
+      "step": 784
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3771929824561406e-05,
+      "loss": 1.002,
+      "step": 785
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3789473684210528e-05,
+      "loss": 1.1651,
+      "step": 786
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.380701754385965e-05,
+      "loss": 1.2225,
+      "step": 787
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3824561403508774e-05,
+      "loss": 0.9985,
+      "step": 788
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3842105263157896e-05,
+      "loss": 1.006,
+      "step": 789
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.385964912280702e-05,
+      "loss": 1.0136,
+      "step": 790
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3877192982456143e-05,
+      "loss": 1.0879,
+      "step": 791
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3894736842105265e-05,
+      "loss": 1.1054,
+      "step": 792
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3912280701754389e-05,
+      "loss": 1.1917,
+      "step": 793
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.392982456140351e-05,
+      "loss": 1.1935,
+      "step": 794
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3947368421052631e-05,
+      "loss": 1.0553,
+      "step": 795
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3964912280701755e-05,
+      "loss": 1.1189,
+      "step": 796
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3982456140350878e-05,
+      "loss": 0.823,
+      "step": 797
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4e-05,
+      "loss": 0.9704,
+      "step": 798
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4017543859649124e-05,
+      "loss": 1.061,
+      "step": 799
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4035087719298246e-05,
+      "loss": 1.2565,
+      "step": 800
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4052631578947368e-05,
+      "loss": 1.1606,
+      "step": 801
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4070175438596492e-05,
+      "loss": 1.3551,
+      "step": 802
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4087719298245614e-05,
+      "loss": 1.0671,
+      "step": 803
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4105263157894738e-05,
+      "loss": 1.0097,
+      "step": 804
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.412280701754386e-05,
+      "loss": 0.9903,
+      "step": 805
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4140350877192983e-05,
+      "loss": 0.8971,
+      "step": 806
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4157894736842107e-05,
+      "loss": 1.1604,
+      "step": 807
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4175438596491229e-05,
+      "loss": 1.1243,
+      "step": 808
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4192982456140351e-05,
+      "loss": 1.2596,
+      "step": 809
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4210526315789475e-05,
+      "loss": 1.174,
+      "step": 810
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4228070175438597e-05,
+      "loss": 0.9938,
+      "step": 811
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4245614035087721e-05,
+      "loss": 1.1276,
+      "step": 812
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4263157894736843e-05,
+      "loss": 1.0996,
+      "step": 813
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4280701754385966e-05,
+      "loss": 1.3807,
+      "step": 814
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.429824561403509e-05,
+      "loss": 1.405,
+      "step": 815
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4315789473684212e-05,
+      "loss": 0.8384,
+      "step": 816
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4333333333333334e-05,
+      "loss": 0.8444,
+      "step": 817
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4350877192982458e-05,
+      "loss": 1.2117,
+      "step": 818
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.436842105263158e-05,
+      "loss": 1.2205,
+      "step": 819
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4385964912280704e-05,
+      "loss": 0.9611,
+      "step": 820
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4403508771929826e-05,
+      "loss": 0.7012,
+      "step": 821
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4421052631578948e-05,
+      "loss": 1.0668,
+      "step": 822
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4438596491228072e-05,
+      "loss": 1.0825,
+      "step": 823
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4456140350877195e-05,
+      "loss": 1.192,
+      "step": 824
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4473684210526317e-05,
+      "loss": 1.256,
+      "step": 825
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.449122807017544e-05,
+      "loss": 1.1059,
+      "step": 826
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4508771929824563e-05,
+      "loss": 1.4136,
+      "step": 827
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4526315789473687e-05,
+      "loss": 1.0849,
+      "step": 828
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4543859649122809e-05,
+      "loss": 1.1107,
+      "step": 829
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4561403508771931e-05,
+      "loss": 0.7909,
+      "step": 830
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4578947368421055e-05,
+      "loss": 1.2108,
+      "step": 831
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4596491228070177e-05,
+      "loss": 1.0972,
+      "step": 832
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.46140350877193e-05,
+      "loss": 0.8296,
+      "step": 833
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4631578947368424e-05,
+      "loss": 1.0321,
+      "step": 834
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4649122807017544e-05,
+      "loss": 1.0029,
+      "step": 835
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4666666666666666e-05,
+      "loss": 1.0309,
+      "step": 836
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.468421052631579e-05,
+      "loss": 1.2494,
+      "step": 837
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4701754385964912e-05,
+      "loss": 1.1208,
+      "step": 838
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4719298245614035e-05,
+      "loss": 0.8229,
+      "step": 839
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4736842105263159e-05,
+      "loss": 0.7945,
+      "step": 840
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.475438596491228e-05,
+      "loss": 1.1749,
+      "step": 841
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4771929824561405e-05,
+      "loss": 0.9241,
+      "step": 842
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4789473684210527e-05,
+      "loss": 1.1322,
+      "step": 843
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4807017543859649e-05,
+      "loss": 1.0446,
+      "step": 844
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4824561403508773e-05,
+      "loss": 0.8806,
+      "step": 845
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4842105263157895e-05,
+      "loss": 1.2939,
+      "step": 846
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4859649122807018e-05,
+      "loss": 1.1352,
+      "step": 847
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4877192982456141e-05,
+      "loss": 1.3506,
+      "step": 848
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4894736842105264e-05,
+      "loss": 0.9783,
+      "step": 849
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4912280701754388e-05,
+      "loss": 0.8205,
+      "step": 850
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.492982456140351e-05,
+      "loss": 1.3085,
+      "step": 851
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4947368421052632e-05,
+      "loss": 1.0982,
+      "step": 852
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4964912280701756e-05,
+      "loss": 1.0424,
+      "step": 853
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4982456140350878e-05,
+      "loss": 0.8731,
+      "step": 854
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 1.1772,
+      "step": 855
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5017543859649124e-05,
+      "loss": 0.7087,
+      "step": 856
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5035087719298247e-05,
+      "loss": 1.0695,
+      "step": 857
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.505263157894737e-05,
+      "loss": 1.0315,
+      "step": 858
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5070175438596493e-05,
+      "loss": 0.8859,
+      "step": 859
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5087719298245615e-05,
+      "loss": 0.8984,
+      "step": 860
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5105263157894739e-05,
+      "loss": 0.9509,
+      "step": 861
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5122807017543861e-05,
+      "loss": 1.0496,
+      "step": 862
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5140350877192985e-05,
+      "loss": 0.975,
+      "step": 863
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5157894736842107e-05,
+      "loss": 1.0933,
+      "step": 864
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.517543859649123e-05,
+      "loss": 1.0101,
+      "step": 865
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5192982456140353e-05,
+      "loss": 1.0221,
+      "step": 866
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5210526315789476e-05,
+      "loss": 1.0975,
+      "step": 867
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5228070175438598e-05,
+      "loss": 0.7969,
+      "step": 868
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5245614035087722e-05,
+      "loss": 1.0853,
+      "step": 869
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5263157894736846e-05,
+      "loss": 0.9898,
+      "step": 870
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5280701754385968e-05,
+      "loss": 1.2489,
+      "step": 871
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.529824561403509e-05,
+      "loss": 1.4088,
+      "step": 872
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5315789473684212e-05,
+      "loss": 1.0436,
+      "step": 873
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5333333333333334e-05,
+      "loss": 1.1316,
+      "step": 874
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.535087719298246e-05,
+      "loss": 1.2688,
+      "step": 875
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.536842105263158e-05,
+      "loss": 0.6944,
+      "step": 876
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.53859649122807e-05,
+      "loss": 0.9406,
+      "step": 877
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5403508771929823e-05,
+      "loss": 1.1895,
+      "step": 878
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.542105263157895e-05,
+      "loss": 1.1016,
+      "step": 879
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.543859649122807e-05,
+      "loss": 1.2026,
+      "step": 880
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5456140350877193e-05,
+      "loss": 0.989,
+      "step": 881
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5473684210526316e-05,
+      "loss": 1.08,
+      "step": 882
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5491228070175438e-05,
+      "loss": 1.1652,
+      "step": 883
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5508771929824563e-05,
+      "loss": 0.896,
+      "step": 884
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5526315789473686e-05,
+      "loss": 0.8213,
+      "step": 885
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5543859649122808e-05,
+      "loss": 1.082,
+      "step": 886
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.556140350877193e-05,
+      "loss": 0.9546,
+      "step": 887
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5578947368421052e-05,
+      "loss": 0.9932,
+      "step": 888
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5596491228070178e-05,
+      "loss": 1.225,
+      "step": 889
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.56140350877193e-05,
+      "loss": 1.1361,
+      "step": 890
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5631578947368422e-05,
+      "loss": 1.2066,
+      "step": 891
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5649122807017545e-05,
+      "loss": 1.3188,
+      "step": 892
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5666666666666667e-05,
+      "loss": 1.0613,
+      "step": 893
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.568421052631579e-05,
+      "loss": 1.2203,
+      "step": 894
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5701754385964915e-05,
+      "loss": 1.1208,
+      "step": 895
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5719298245614037e-05,
+      "loss": 1.2852,
+      "step": 896
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.573684210526316e-05,
+      "loss": 1.1937,
+      "step": 897
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.575438596491228e-05,
+      "loss": 1.1958,
+      "step": 898
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5771929824561404e-05,
+      "loss": 1.229,
+      "step": 899
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.578947368421053e-05,
+      "loss": 0.9207,
+      "step": 900
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.580701754385965e-05,
+      "loss": 0.975,
+      "step": 901
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5824561403508774e-05,
+      "loss": 0.9833,
+      "step": 902
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5842105263157896e-05,
+      "loss": 1.2593,
+      "step": 903
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5859649122807018e-05,
+      "loss": 0.9639,
+      "step": 904
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5877192982456144e-05,
+      "loss": 1.2403,
+      "step": 905
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5894736842105266e-05,
+      "loss": 1.2749,
+      "step": 906
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5912280701754388e-05,
+      "loss": 1.0324,
+      "step": 907
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.592982456140351e-05,
+      "loss": 1.3232,
+      "step": 908
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5947368421052633e-05,
+      "loss": 0.8223,
+      "step": 909
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5964912280701755e-05,
+      "loss": 1.2234,
+      "step": 910
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.598245614035088e-05,
+      "loss": 1.1673,
+      "step": 911
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 1.0598,
+      "step": 912
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6017543859649125e-05,
+      "loss": 0.8896,
+      "step": 913
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6035087719298247e-05,
+      "loss": 1.0088,
+      "step": 914
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.605263157894737e-05,
+      "loss": 1.1186,
+      "step": 915
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6070175438596495e-05,
+      "loss": 1.038,
+      "step": 916
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6087719298245614e-05,
+      "loss": 0.9963,
+      "step": 917
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6105263157894736e-05,
+      "loss": 1.1643,
+      "step": 918
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.612280701754386e-05,
+      "loss": 1.1683,
+      "step": 919
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6140350877192984e-05,
+      "loss": 1.1209,
+      "step": 920
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6157894736842106e-05,
+      "loss": 0.8687,
+      "step": 921
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6175438596491228e-05,
+      "loss": 0.8784,
+      "step": 922
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.619298245614035e-05,
+      "loss": 1.1551,
+      "step": 923
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6210526315789473e-05,
+      "loss": 1.3632,
+      "step": 924
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.62280701754386e-05,
+      "loss": 1.2745,
+      "step": 925
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.624561403508772e-05,
+      "loss": 1.0472,
+      "step": 926
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6263157894736843e-05,
+      "loss": 0.9762,
+      "step": 927
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6280701754385965e-05,
+      "loss": 1.2953,
+      "step": 928
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6298245614035087e-05,
+      "loss": 0.8733,
+      "step": 929
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6315789473684213e-05,
+      "loss": 1.028,
+      "step": 930
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6333333333333335e-05,
+      "loss": 1.1621,
+      "step": 931
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6350877192982457e-05,
+      "loss": 1.3903,
+      "step": 932
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.636842105263158e-05,
+      "loss": 1.1225,
+      "step": 933
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6385964912280702e-05,
+      "loss": 0.8964,
+      "step": 934
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6403508771929827e-05,
+      "loss": 1.1622,
+      "step": 935
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.642105263157895e-05,
+      "loss": 1.1664,
+      "step": 936
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6438596491228072e-05,
+      "loss": 0.9967,
+      "step": 937
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6456140350877194e-05,
+      "loss": 0.9317,
+      "step": 938
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6473684210526316e-05,
+      "loss": 0.932,
+      "step": 939
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.649122807017544e-05,
+      "loss": 1.135,
+      "step": 940
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6508771929824564e-05,
+      "loss": 0.9793,
+      "step": 941
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6526315789473686e-05,
+      "loss": 0.9836,
+      "step": 942
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.654385964912281e-05,
+      "loss": 0.8499,
+      "step": 943
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.656140350877193e-05,
+      "loss": 1.0533,
+      "step": 944
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6578947368421053e-05,
+      "loss": 1.2257,
+      "step": 945
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.659649122807018e-05,
+      "loss": 0.9578,
+      "step": 946
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.66140350877193e-05,
+      "loss": 1.0036,
+      "step": 947
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6631578947368423e-05,
+      "loss": 1.0823,
+      "step": 948
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6649122807017545e-05,
+      "loss": 1.0689,
+      "step": 949
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 1.0037,
+      "step": 950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6684210526315793e-05,
+      "loss": 0.9189,
+      "step": 951
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6701754385964915e-05,
+      "loss": 1.0684,
+      "step": 952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6719298245614038e-05,
+      "loss": 1.39,
+      "step": 953
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.673684210526316e-05,
+      "loss": 1.0716,
+      "step": 954
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6754385964912282e-05,
+      "loss": 1.0211,
+      "step": 955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6771929824561408e-05,
+      "loss": 1.1318,
+      "step": 956
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6789473684210526e-05,
+      "loss": 1.111,
+      "step": 957
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.680701754385965e-05,
+      "loss": 1.3812,
+      "step": 958
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.682456140350877e-05,
+      "loss": 1.3288,
+      "step": 959
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6842105263157896e-05,
+      "loss": 1.2646,
+      "step": 960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.685964912280702e-05,
+      "loss": 0.7787,
+      "step": 961
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.687719298245614e-05,
+      "loss": 0.8659,
+      "step": 962
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6894736842105263e-05,
+      "loss": 1.0055,
+      "step": 963
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6912280701754385e-05,
+      "loss": 1.059,
+      "step": 964
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.692982456140351e-05,
+      "loss": 1.0,
+      "step": 965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6947368421052633e-05,
+      "loss": 1.0276,
+      "step": 966
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6964912280701755e-05,
+      "loss": 0.971,
+      "step": 967
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6982456140350878e-05,
+      "loss": 1.1175,
+      "step": 968
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7e-05,
+      "loss": 1.2067,
+      "step": 969
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7017543859649125e-05,
+      "loss": 1.041,
+      "step": 970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7035087719298248e-05,
+      "loss": 1.1753,
+      "step": 971
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.705263157894737e-05,
+      "loss": 1.2357,
+      "step": 972
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7070175438596492e-05,
+      "loss": 1.1373,
+      "step": 973
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7087719298245614e-05,
+      "loss": 0.9563,
+      "step": 974
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7105263157894737e-05,
+      "loss": 1.1647,
+      "step": 975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7122807017543862e-05,
+      "loss": 0.8773,
+      "step": 976
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7140350877192984e-05,
+      "loss": 1.3904,
+      "step": 977
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7157894736842107e-05,
+      "loss": 0.9985,
+      "step": 978
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.717543859649123e-05,
+      "loss": 0.9097,
+      "step": 979
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.719298245614035e-05,
+      "loss": 1.3807,
+      "step": 980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7210526315789477e-05,
+      "loss": 1.1942,
+      "step": 981
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.72280701754386e-05,
+      "loss": 1.0032,
+      "step": 982
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.724561403508772e-05,
+      "loss": 1.1538,
+      "step": 983
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7263157894736843e-05,
+      "loss": 1.151,
+      "step": 984
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7280701754385966e-05,
+      "loss": 1.2212,
+      "step": 985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.729824561403509e-05,
+      "loss": 0.8538,
+      "step": 986
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7315789473684213e-05,
+      "loss": 1.2312,
+      "step": 987
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7333333333333336e-05,
+      "loss": 0.892,
+      "step": 988
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7350877192982458e-05,
+      "loss": 1.0197,
+      "step": 989
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.736842105263158e-05,
+      "loss": 1.1101,
+      "step": 990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7385964912280702e-05,
+      "loss": 1.1988,
+      "step": 991
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7403508771929828e-05,
+      "loss": 1.0719,
+      "step": 992
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.742105263157895e-05,
+      "loss": 1.1,
+      "step": 993
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7438596491228072e-05,
+      "loss": 1.0638,
+      "step": 994
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7456140350877195e-05,
+      "loss": 1.172,
+      "step": 995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7473684210526317e-05,
+      "loss": 1.195,
+      "step": 996
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7491228070175442e-05,
+      "loss": 1.0515,
+      "step": 997
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.750877192982456e-05,
+      "loss": 1.1854,
+      "step": 998
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7526315789473683e-05,
+      "loss": 0.9488,
+      "step": 999
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.754385964912281e-05,
+      "loss": 1.0717,
+      "step": 1000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.756140350877193e-05,
+      "loss": 1.2441,
+      "step": 1001
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7578947368421054e-05,
+      "loss": 1.3368,
+      "step": 1002
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7596491228070176e-05,
+      "loss": 1.3496,
+      "step": 1003
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7614035087719298e-05,
+      "loss": 1.262,
+      "step": 1004
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.763157894736842e-05,
+      "loss": 1.1156,
+      "step": 1005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7649122807017546e-05,
+      "loss": 1.1583,
+      "step": 1006
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7666666666666668e-05,
+      "loss": 1.0214,
+      "step": 1007
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.768421052631579e-05,
+      "loss": 0.8083,
+      "step": 1008
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7701754385964913e-05,
+      "loss": 1.0007,
+      "step": 1009
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7719298245614035e-05,
+      "loss": 1.1473,
+      "step": 1010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.773684210526316e-05,
+      "loss": 1.0297,
+      "step": 1011
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7754385964912283e-05,
+      "loss": 1.0521,
+      "step": 1012
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7771929824561405e-05,
+      "loss": 1.3241,
+      "step": 1013
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7789473684210527e-05,
+      "loss": 0.7723,
+      "step": 1014
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.780701754385965e-05,
+      "loss": 0.8342,
+      "step": 1015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7824561403508775e-05,
+      "loss": 0.9314,
+      "step": 1016
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7842105263157897e-05,
+      "loss": 0.9717,
+      "step": 1017
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.785964912280702e-05,
+      "loss": 1.2165,
+      "step": 1018
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.787719298245614e-05,
+      "loss": 1.219,
+      "step": 1019
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7894736842105264e-05,
+      "loss": 0.9036,
+      "step": 1020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7912280701754386e-05,
+      "loss": 1.5559,
+      "step": 1021
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.792982456140351e-05,
+      "loss": 0.5845,
+      "step": 1022
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7947368421052634e-05,
+      "loss": 0.9358,
+      "step": 1023
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7964912280701756e-05,
+      "loss": 1.0556,
+      "step": 1024
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7982456140350878e-05,
+      "loss": 1.3088,
+      "step": 1025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8e-05,
+      "loss": 0.9056,
+      "step": 1026
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8017543859649126e-05,
+      "loss": 0.9921,
+      "step": 1027
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8035087719298248e-05,
+      "loss": 1.3691,
+      "step": 1028
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.805263157894737e-05,
+      "loss": 1.1627,
+      "step": 1029
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8070175438596493e-05,
+      "loss": 1.265,
+      "step": 1030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8087719298245615e-05,
+      "loss": 1.2504,
+      "step": 1031
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.810526315789474e-05,
+      "loss": 0.9915,
+      "step": 1032
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8122807017543863e-05,
+      "loss": 1.044,
+      "step": 1033
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8140350877192985e-05,
+      "loss": 1.1971,
+      "step": 1034
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8157894736842107e-05,
+      "loss": 1.1392,
+      "step": 1035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.817543859649123e-05,
+      "loss": 1.0654,
+      "step": 1036
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8192982456140355e-05,
+      "loss": 1.1992,
+      "step": 1037
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8210526315789477e-05,
+      "loss": 1.0087,
+      "step": 1038
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8228070175438596e-05,
+      "loss": 1.2377,
+      "step": 1039
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.824561403508772e-05,
+      "loss": 1.1327,
+      "step": 1040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8263157894736844e-05,
+      "loss": 0.9853,
+      "step": 1041
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8280701754385966e-05,
+      "loss": 1.2115,
+      "step": 1042
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.829824561403509e-05,
+      "loss": 0.8513,
+      "step": 1043
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.831578947368421e-05,
+      "loss": 0.9157,
+      "step": 1044
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 0.9578,
+      "step": 1045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.835087719298246e-05,
+      "loss": 1.14,
+      "step": 1046
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.836842105263158e-05,
+      "loss": 0.8514,
+      "step": 1047
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8385964912280703e-05,
+      "loss": 1.029,
+      "step": 1048
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8403508771929825e-05,
+      "loss": 1.0572,
+      "step": 1049
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8421052631578947e-05,
+      "loss": 1.2166,
+      "step": 1050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.843859649122807e-05,
+      "loss": 1.1296,
+      "step": 1051
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8456140350877195e-05,
+      "loss": 1.1601,
+      "step": 1052
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8473684210526317e-05,
+      "loss": 1.3714,
+      "step": 1053
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.849122807017544e-05,
+      "loss": 0.9589,
+      "step": 1054
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8508771929824562e-05,
+      "loss": 1.0292,
+      "step": 1055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8526315789473684e-05,
+      "loss": 1.1694,
+      "step": 1056
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.854385964912281e-05,
+      "loss": 1.2419,
+      "step": 1057
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8561403508771932e-05,
+      "loss": 1.3277,
+      "step": 1058
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8578947368421054e-05,
+      "loss": 1.0689,
+      "step": 1059
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8596491228070176e-05,
+      "loss": 0.5274,
+      "step": 1060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.86140350877193e-05,
+      "loss": 1.1412,
+      "step": 1061
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8631578947368424e-05,
+      "loss": 1.2094,
+      "step": 1062
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8649122807017546e-05,
+      "loss": 1.1094,
+      "step": 1063
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.866666666666667e-05,
+      "loss": 0.8459,
+      "step": 1064
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.868421052631579e-05,
+      "loss": 1.0134,
+      "step": 1065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8701754385964913e-05,
+      "loss": 1.243,
+      "step": 1066
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.871929824561404e-05,
+      "loss": 1.2985,
+      "step": 1067
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.873684210526316e-05,
+      "loss": 1.0238,
+      "step": 1068
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8754385964912283e-05,
+      "loss": 0.9844,
+      "step": 1069
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8771929824561405e-05,
+      "loss": 1.252,
+      "step": 1070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8789473684210528e-05,
+      "loss": 1.0381,
+      "step": 1071
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.880701754385965e-05,
+      "loss": 1.1357,
+      "step": 1072
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8824561403508775e-05,
+      "loss": 0.9746,
+      "step": 1073
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8842105263157898e-05,
+      "loss": 0.9555,
+      "step": 1074
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.885964912280702e-05,
+      "loss": 1.3354,
+      "step": 1075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8877192982456142e-05,
+      "loss": 1.1028,
+      "step": 1076
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8894736842105264e-05,
+      "loss": 0.9737,
+      "step": 1077
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.891228070175439e-05,
+      "loss": 1.3185,
+      "step": 1078
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.892982456140351e-05,
+      "loss": 1.0967,
+      "step": 1079
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.894736842105263e-05,
+      "loss": 1.0982,
+      "step": 1080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8964912280701757e-05,
+      "loss": 0.8617,
+      "step": 1081
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.898245614035088e-05,
+      "loss": 0.8143,
+      "step": 1082
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9e-05,
+      "loss": 1.0417,
+      "step": 1083
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9017543859649123e-05,
+      "loss": 1.2006,
+      "step": 1084
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9035087719298245e-05,
+      "loss": 1.0916,
+      "step": 1085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9052631578947368e-05,
+      "loss": 0.9642,
+      "step": 1086
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9070175438596493e-05,
+      "loss": 1.0063,
+      "step": 1087
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9087719298245616e-05,
+      "loss": 1.2341,
+      "step": 1088
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9105263157894738e-05,
+      "loss": 1.0834,
+      "step": 1089
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.912280701754386e-05,
+      "loss": 1.3461,
+      "step": 1090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9140350877192982e-05,
+      "loss": 1.1346,
+      "step": 1091
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9157894736842108e-05,
+      "loss": 1.0225,
+      "step": 1092
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.917543859649123e-05,
+      "loss": 1.2253,
+      "step": 1093
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9192982456140352e-05,
+      "loss": 0.885,
+      "step": 1094
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9210526315789474e-05,
+      "loss": 1.029,
+      "step": 1095
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9228070175438597e-05,
+      "loss": 1.3776,
+      "step": 1096
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9245614035087722e-05,
+      "loss": 0.9629,
+      "step": 1097
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9263157894736845e-05,
+      "loss": 0.9204,
+      "step": 1098
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9280701754385967e-05,
+      "loss": 1.0074,
+      "step": 1099
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.929824561403509e-05,
+      "loss": 1.1948,
+      "step": 1100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.931578947368421e-05,
+      "loss": 0.7903,
+      "step": 1101
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9333333333333333e-05,
+      "loss": 0.8462,
+      "step": 1102
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.935087719298246e-05,
+      "loss": 1.1156,
+      "step": 1103
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.936842105263158e-05,
+      "loss": 1.0898,
+      "step": 1104
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9385964912280704e-05,
+      "loss": 1.1591,
+      "step": 1105
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9403508771929826e-05,
+      "loss": 1.2186,
+      "step": 1106
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9421052631578948e-05,
+      "loss": 0.9612,
+      "step": 1107
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9438596491228074e-05,
+      "loss": 1.12,
+      "step": 1108
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9456140350877196e-05,
+      "loss": 1.1685,
+      "step": 1109
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9473684210526318e-05,
+      "loss": 1.2241,
+      "step": 1110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.949122807017544e-05,
+      "loss": 0.9524,
+      "step": 1111
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9508771929824562e-05,
+      "loss": 0.9038,
+      "step": 1112
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9526315789473688e-05,
+      "loss": 0.9145,
+      "step": 1113
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.954385964912281e-05,
+      "loss": 1.0646,
+      "step": 1114
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9561403508771933e-05,
+      "loss": 0.9822,
+      "step": 1115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9578947368421055e-05,
+      "loss": 0.8804,
+      "step": 1116
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9596491228070177e-05,
+      "loss": 0.9315,
+      "step": 1117
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.96140350877193e-05,
+      "loss": 0.8492,
+      "step": 1118
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9631578947368425e-05,
+      "loss": 1.1719,
+      "step": 1119
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9649122807017544e-05,
+      "loss": 0.754,
+      "step": 1120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9666666666666666e-05,
+      "loss": 1.3373,
+      "step": 1121
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.968421052631579e-05,
+      "loss": 1.1798,
+      "step": 1122
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9701754385964914e-05,
+      "loss": 1.3161,
+      "step": 1123
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9719298245614036e-05,
+      "loss": 0.8323,
+      "step": 1124
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9736842105263158e-05,
+      "loss": 1.2111,
+      "step": 1125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.975438596491228e-05,
+      "loss": 1.1471,
+      "step": 1126
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9771929824561406e-05,
+      "loss": 1.2269,
+      "step": 1127
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9789473684210528e-05,
+      "loss": 1.0589,
+      "step": 1128
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980701754385965e-05,
+      "loss": 1.2055,
+      "step": 1129
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9824561403508773e-05,
+      "loss": 1.3279,
+      "step": 1130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9842105263157895e-05,
+      "loss": 0.7549,
+      "step": 1131
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9859649122807017e-05,
+      "loss": 1.0925,
+      "step": 1132
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9877192982456143e-05,
+      "loss": 1.2299,
+      "step": 1133
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9894736842105265e-05,
+      "loss": 0.9983,
+      "step": 1134
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9912280701754387e-05,
+      "loss": 1.0903,
+      "step": 1135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.992982456140351e-05,
+      "loss": 1.0666,
+      "step": 1136
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.994736842105263e-05,
+      "loss": 0.9941,
+      "step": 1137
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9964912280701757e-05,
+      "loss": 1.2467,
+      "step": 1138
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.998245614035088e-05,
+      "loss": 1.4287,
+      "step": 1139
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 1.2683,
+      "step": 1140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999963639445e-05,
+      "loss": 1.0055,
+      "step": 1141
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999985455778e-05,
+      "loss": 1.1835,
+      "step": 1142
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999672755006e-05,
+      "loss": 1.0415,
+      "step": 1143
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999418231126e-05,
+      "loss": 1.0454,
+      "step": 1144
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999090986133e-05,
+      "loss": 1.0834,
+      "step": 1145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998691020042e-05,
+      "loss": 1.0714,
+      "step": 1146
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998218332845e-05,
+      "loss": 1.383,
+      "step": 1147
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999997672924553e-05,
+      "loss": 1.1417,
+      "step": 1148
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999705479517e-05,
+      "loss": 1.0533,
+      "step": 1149
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999996363944695e-05,
+      "loss": 1.2503,
+      "step": 1150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999560037314e-05,
+      "loss": 1.0174,
+      "step": 1151
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999994764080502e-05,
+      "loss": 1.2363,
+      "step": 1152
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999993855066793e-05,
+      "loss": 1.1296,
+      "step": 1153
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999992873332016e-05,
+      "loss": 1.2797,
+      "step": 1154
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999991818876183e-05,
+      "loss": 1.2832,
+      "step": 1155
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99999906916993e-05,
+      "loss": 1.3047,
+      "step": 1156
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998949180137e-05,
+      "loss": 1.0196,
+      "step": 1157
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999998821918241e-05,
+      "loss": 1.1626,
+      "step": 1158
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999986873842424e-05,
+      "loss": 1.1241,
+      "step": 1159
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999985455781423e-05,
+      "loss": 1.0679,
+      "step": 1160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999983964999417e-05,
+      "loss": 1.3478,
+      "step": 1161
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999982401496418e-05,
+      "loss": 0.8488,
+      "step": 1162
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999980765272435e-05,
+      "loss": 1.2771,
+      "step": 1163
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999979056327482e-05,
+      "loss": 0.9716,
+      "step": 1164
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999977274661572e-05,
+      "loss": 1.0718,
+      "step": 1165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999975420274716e-05,
+      "loss": 1.1358,
+      "step": 1166
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999973493166927e-05,
+      "loss": 0.8779,
+      "step": 1167
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999971493338225e-05,
+      "loss": 1.2505,
+      "step": 1168
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999969420788614e-05,
+      "loss": 0.9535,
+      "step": 1169
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999967275518118e-05,
+      "loss": 0.8939,
+      "step": 1170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996505752675e-05,
+      "loss": 0.8391,
+      "step": 1171
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996276681452e-05,
+      "loss": 1.0448,
+      "step": 1172
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996040338146e-05,
+      "loss": 1.3107,
+      "step": 1173
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999957967227567e-05,
+      "loss": 1.2447,
+      "step": 1174
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999955458352874e-05,
+      "loss": 0.8962,
+      "step": 1175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999952876757397e-05,
+      "loss": 1.1596,
+      "step": 1176
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999995022244115e-05,
+      "loss": 0.8473,
+      "step": 1177
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999947495404153e-05,
+      "loss": 1.4114,
+      "step": 1178
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999994469564643e-05,
+      "loss": 1.0946,
+      "step": 1179
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999941823167997e-05,
+      "loss": 1.0328,
+      "step": 1180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999938877968876e-05,
+      "loss": 0.9958,
+      "step": 1181
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999993586004909e-05,
+      "loss": 1.0808,
+      "step": 1182
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999993276940866e-05,
+      "loss": 0.8656,
+      "step": 1183
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999992960604761e-05,
+      "loss": 1.3684,
+      "step": 1184
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999992636996596e-05,
+      "loss": 1.0556,
+      "step": 1185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999923061163735e-05,
+      "loss": 0.9367,
+      "step": 1186
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999919679640958e-05,
+      "loss": 1.1604,
+      "step": 1187
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999916225397654e-05,
+      "loss": 1.296,
+      "step": 1188
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999912698433852e-05,
+      "loss": 1.2365,
+      "step": 1189
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999909098749574e-05,
+      "loss": 0.9639,
+      "step": 1190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999905426344844e-05,
+      "loss": 0.9282,
+      "step": 1191
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999901681219693e-05,
+      "loss": 1.2474,
+      "step": 1192
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999897863374144e-05,
+      "loss": 1.0792,
+      "step": 1193
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999893972808228e-05,
+      "loss": 0.9826,
+      "step": 1194
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999890009521976e-05,
+      "loss": 1.1482,
+      "step": 1195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999988597351541e-05,
+      "loss": 1.0908,
+      "step": 1196
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999881864788567e-05,
+      "loss": 1.1777,
+      "step": 1197
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999877683341467e-05,
+      "loss": 1.4993,
+      "step": 1198
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999987342917415e-05,
+      "loss": 1.3103,
+      "step": 1199
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999869102286638e-05,
+      "loss": 1.0128,
+      "step": 1200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999864702678976e-05,
+      "loss": 1.0843,
+      "step": 1201
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999986023035118e-05,
+      "loss": 1.0224,
+      "step": 1202
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999855685303294e-05,
+      "loss": 1.2283,
+      "step": 1203
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999851067535344e-05,
+      "loss": 1.3109,
+      "step": 1204
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999846377047366e-05,
+      "loss": 1.2431,
+      "step": 1205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999841613839397e-05,
+      "loss": 1.2294,
+      "step": 1206
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999983677791147e-05,
+      "loss": 0.9487,
+      "step": 1207
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999983186926362e-05,
+      "loss": 0.9101,
+      "step": 1208
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999982688789588e-05,
+      "loss": 1.136,
+      "step": 1209
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999821833808288e-05,
+      "loss": 0.9063,
+      "step": 1210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999816707000883e-05,
+      "loss": 0.912,
+      "step": 1211
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99998115074737e-05,
+      "loss": 1.0574,
+      "step": 1212
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999806235226776e-05,
+      "loss": 1.1634,
+      "step": 1213
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999980089026015e-05,
+      "loss": 1.1873,
+      "step": 1214
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999795472573865e-05,
+      "loss": 1.3301,
+      "step": 1215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999789982167958e-05,
+      "loss": 1.0992,
+      "step": 1216
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999784419042463e-05,
+      "loss": 0.9019,
+      "step": 1217
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999778783197425e-05,
+      "loss": 1.2261,
+      "step": 1218
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999773074632887e-05,
+      "loss": 1.1671,
+      "step": 1219
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999976729334889e-05,
+      "loss": 0.8246,
+      "step": 1220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999761439345474e-05,
+      "loss": 1.0134,
+      "step": 1221
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999975551262268e-05,
+      "loss": 1.0026,
+      "step": 1222
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999749513180557e-05,
+      "loss": 1.1062,
+      "step": 1223
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999743441019144e-05,
+      "loss": 1.0156,
+      "step": 1224
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999737296138486e-05,
+      "loss": 1.1023,
+      "step": 1225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999731078538625e-05,
+      "loss": 1.146,
+      "step": 1226
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999724788219614e-05,
+      "loss": 0.9968,
+      "step": 1227
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999971842518149e-05,
+      "loss": 1.2705,
+      "step": 1228
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999711989424306e-05,
+      "loss": 1.0904,
+      "step": 1229
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999705480948107e-05,
+      "loss": 1.2308,
+      "step": 1230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999698899752937e-05,
+      "loss": 1.123,
+      "step": 1231
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999692245838847e-05,
+      "loss": 1.2317,
+      "step": 1232
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999685519205884e-05,
+      "loss": 0.9715,
+      "step": 1233
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999678719854095e-05,
+      "loss": 1.0455,
+      "step": 1234
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999671847783535e-05,
+      "loss": 1.1611,
+      "step": 1235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999664902994255e-05,
+      "loss": 1.1453,
+      "step": 1236
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999965788548629e-05,
+      "loss": 1.33,
+      "step": 1237
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999650795259713e-05,
+      "loss": 1.3018,
+      "step": 1238
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999964363231456e-05,
+      "loss": 1.221,
+      "step": 1239
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999636396650888e-05,
+      "loss": 1.0408,
+      "step": 1240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999962908826875e-05,
+      "loss": 1.3194,
+      "step": 1241
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999621707168205e-05,
+      "loss": 1.0089,
+      "step": 1242
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999961425334929e-05,
+      "loss": 1.0471,
+      "step": 1243
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999606726812078e-05,
+      "loss": 1.0058,
+      "step": 1244
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999599127556614e-05,
+      "loss": 1.0187,
+      "step": 1245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999591455582952e-05,
+      "loss": 1.1965,
+      "step": 1246
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999583710891152e-05,
+      "loss": 1.011,
+      "step": 1247
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999575893481265e-05,
+      "loss": 1.1448,
+      "step": 1248
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999956800335336e-05,
+      "loss": 1.2595,
+      "step": 1249
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999956004050748e-05,
+      "loss": 1.1369,
+      "step": 1250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999552004943687e-05,
+      "loss": 1.0403,
+      "step": 1251
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999543896662046e-05,
+      "loss": 1.1396,
+      "step": 1252
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999953571566261e-05,
+      "loss": 1.1457,
+      "step": 1253
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999952746194544e-05,
+      "loss": 1.0991,
+      "step": 1254
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999519135510597e-05,
+      "loss": 0.972,
+      "step": 1255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999951073635814e-05,
+      "loss": 1.1184,
+      "step": 1256
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999502264488128e-05,
+      "loss": 1.2063,
+      "step": 1257
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999493719900625e-05,
+      "loss": 1.3882,
+      "step": 1258
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999485102595696e-05,
+      "loss": 1.0689,
+      "step": 1259
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99994764125734e-05,
+      "loss": 1.3353,
+      "step": 1260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999467649833798e-05,
+      "loss": 1.1645,
+      "step": 1261
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999458814376963e-05,
+      "loss": 1.3335,
+      "step": 1262
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999944990620295e-05,
+      "loss": 1.0755,
+      "step": 1263
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999440925311825e-05,
+      "loss": 1.237,
+      "step": 1264
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999431871703658e-05,
+      "loss": 1.2492,
+      "step": 1265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999942274537851e-05,
+      "loss": 1.2176,
+      "step": 1266
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999941354633645e-05,
+      "loss": 1.3376,
+      "step": 1267
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999940427457755e-05,
+      "loss": 1.1707,
+      "step": 1268
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999394930101862e-05,
+      "loss": 1.2223,
+      "step": 1269
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999385512909467e-05,
+      "loss": 1.0458,
+      "step": 1270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999376023000432e-05,
+      "loss": 1.2121,
+      "step": 1271
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999936646037482e-05,
+      "loss": 1.332,
+      "step": 1272
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999356825032713e-05,
+      "loss": 0.9713,
+      "step": 1273
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999347116974164e-05,
+      "loss": 1.341,
+      "step": 1274
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999337336199257e-05,
+      "loss": 0.9023,
+      "step": 1275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999327482708058e-05,
+      "loss": 1.0834,
+      "step": 1276
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999317556500643e-05,
+      "loss": 1.2142,
+      "step": 1277
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999307557577072e-05,
+      "loss": 1.2975,
+      "step": 1278
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999929748593743e-05,
+      "loss": 1.0373,
+      "step": 1279
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999928734158179e-05,
+      "loss": 0.8161,
+      "step": 1280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999277124510215e-05,
+      "loss": 1.3758,
+      "step": 1281
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999926683472279e-05,
+      "loss": 1.2474,
+      "step": 1282
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999256472219587e-05,
+      "loss": 0.9025,
+      "step": 1283
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999924603700068e-05,
+      "loss": 1.0554,
+      "step": 1284
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999235529066143e-05,
+      "loss": 1.0005,
+      "step": 1285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999224948416058e-05,
+      "loss": 1.1398,
+      "step": 1286
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999214295050498e-05,
+      "loss": 1.0466,
+      "step": 1287
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999203568969542e-05,
+      "loss": 0.8539,
+      "step": 1288
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999192770173264e-05,
+      "loss": 1.1029,
+      "step": 1289
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999918189866175e-05,
+      "loss": 1.1151,
+      "step": 1290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999917095443507e-05,
+      "loss": 0.9231,
+      "step": 1291
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999159937493313e-05,
+      "loss": 0.7677,
+      "step": 1292
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999148847836552e-05,
+      "loss": 1.2756,
+      "step": 1293
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999137685464873e-05,
+      "loss": 0.7705,
+      "step": 1294
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999912645037835e-05,
+      "loss": 1.287,
+      "step": 1295
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999115142577072e-05,
+      "loss": 1.1005,
+      "step": 1296
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999103762061117e-05,
+      "loss": 1.0752,
+      "step": 1297
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999909230883057e-05,
+      "loss": 1.0501,
+      "step": 1298
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999080782885514e-05,
+      "loss": 1.1778,
+      "step": 1299
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999069184226032e-05,
+      "loss": 1.4039,
+      "step": 1300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999057512852207e-05,
+      "loss": 0.8852,
+      "step": 1301
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999045768764124e-05,
+      "loss": 1.1554,
+      "step": 1302
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999033951961874e-05,
+      "loss": 1.133,
+      "step": 1303
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999022062445535e-05,
+      "loss": 0.9518,
+      "step": 1304
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999010100215202e-05,
+      "loss": 1.2884,
+      "step": 1305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998998065270953e-05,
+      "loss": 1.2971,
+      "step": 1306
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999898595761288e-05,
+      "loss": 1.1456,
+      "step": 1307
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998973777241074e-05,
+      "loss": 0.7988,
+      "step": 1308
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999896152415562e-05,
+      "loss": 0.9706,
+      "step": 1309
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99989491983566e-05,
+      "loss": 1.0249,
+      "step": 1310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998936799844116e-05,
+      "loss": 1.1688,
+      "step": 1311
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998924328618258e-05,
+      "loss": 1.0747,
+      "step": 1312
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998911784679107e-05,
+      "loss": 1.1316,
+      "step": 1313
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999889916802676e-05,
+      "loss": 1.1987,
+      "step": 1314
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998886478661304e-05,
+      "loss": 0.9934,
+      "step": 1315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998873716582842e-05,
+      "loss": 1.0798,
+      "step": 1316
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998860881791456e-05,
+      "loss": 1.0819,
+      "step": 1317
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998847974287245e-05,
+      "loss": 1.139,
+      "step": 1318
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99988349940703e-05,
+      "loss": 0.9025,
+      "step": 1319
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998821941140716e-05,
+      "loss": 1.2083,
+      "step": 1320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999880881549859e-05,
+      "loss": 0.6639,
+      "step": 1321
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998795617144016e-05,
+      "loss": 0.8696,
+      "step": 1322
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999878234607709e-05,
+      "loss": 0.8798,
+      "step": 1323
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998769002297906e-05,
+      "loss": 1.2156,
+      "step": 1324
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998755585806563e-05,
+      "loss": 1.4233,
+      "step": 1325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998742096603166e-05,
+      "loss": 1.1021,
+      "step": 1326
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99987285346878e-05,
+      "loss": 1.0061,
+      "step": 1327
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999871490006057e-05,
+      "loss": 1.0218,
+      "step": 1328
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998701192721576e-05,
+      "loss": 1.2328,
+      "step": 1329
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998687412670915e-05,
+      "loss": 1.0408,
+      "step": 1330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998673559908692e-05,
+      "loss": 1.1007,
+      "step": 1331
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998659634435006e-05,
+      "loss": 1.3031,
+      "step": 1332
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998645636249952e-05,
+      "loss": 0.9648,
+      "step": 1333
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998631565353638e-05,
+      "loss": 1.116,
+      "step": 1334
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998617421746166e-05,
+      "loss": 1.0871,
+      "step": 1335
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998603205427635e-05,
+      "loss": 1.0091,
+      "step": 1336
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998588916398155e-05,
+      "loss": 0.9397,
+      "step": 1337
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998574554657825e-05,
+      "loss": 1.0302,
+      "step": 1338
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998560120206748e-05,
+      "loss": 1.1061,
+      "step": 1339
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998545613045034e-05,
+      "loss": 1.0393,
+      "step": 1340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998531033172788e-05,
+      "loss": 0.9089,
+      "step": 1341
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999851638059011e-05,
+      "loss": 1.113,
+      "step": 1342
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999850165529711e-05,
+      "loss": 0.8273,
+      "step": 1343
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99984868572939e-05,
+      "loss": 1.2776,
+      "step": 1344
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999847198658058e-05,
+      "loss": 1.0919,
+      "step": 1345
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999845704315726e-05,
+      "loss": 0.8794,
+      "step": 1346
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998442027024057e-05,
+      "loss": 1.2238,
+      "step": 1347
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998426938181068e-05,
+      "loss": 0.9472,
+      "step": 1348
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998411776628407e-05,
+      "loss": 1.0045,
+      "step": 1349
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998396542366188e-05,
+      "loss": 1.1596,
+      "step": 1350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998381235394518e-05,
+      "loss": 1.2015,
+      "step": 1351
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999836585571351e-05,
+      "loss": 0.9454,
+      "step": 1352
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998350403323274e-05,
+      "loss": 1.253,
+      "step": 1353
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998334878223923e-05,
+      "loss": 1.0895,
+      "step": 1354
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998319280415573e-05,
+      "loss": 0.8967,
+      "step": 1355
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998303609898335e-05,
+      "loss": 1.3196,
+      "step": 1356
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999828786667232e-05,
+      "loss": 0.8669,
+      "step": 1357
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998272050737648e-05,
+      "loss": 0.8656,
+      "step": 1358
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998256162094426e-05,
+      "loss": 1.1724,
+      "step": 1359
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998240200742784e-05,
+      "loss": 1.223,
+      "step": 1360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998224166682823e-05,
+      "loss": 1.409,
+      "step": 1361
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998208059914665e-05,
+      "loss": 1.3452,
+      "step": 1362
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999819188043843e-05,
+      "loss": 1.0814,
+      "step": 1363
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999817562825423e-05,
+      "loss": 1.1223,
+      "step": 1364
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998159303362193e-05,
+      "loss": 1.0759,
+      "step": 1365
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998142905762426e-05,
+      "loss": 1.1853,
+      "step": 1366
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998126435455053e-05,
+      "loss": 1.0405,
+      "step": 1367
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998109892440197e-05,
+      "loss": 1.1407,
+      "step": 1368
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998093276717975e-05,
+      "loss": 1.047,
+      "step": 1369
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998076588288507e-05,
+      "loss": 1.039,
+      "step": 1370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999805982715192e-05,
+      "loss": 0.9584,
+      "step": 1371
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999804299330833e-05,
+      "loss": 0.9814,
+      "step": 1372
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999802608675786e-05,
+      "loss": 1.1483,
+      "step": 1373
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999800910750063e-05,
+      "loss": 1.377,
+      "step": 1374
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997992055536774e-05,
+      "loss": 1.1111,
+      "step": 1375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997974930866406e-05,
+      "loss": 1.028,
+      "step": 1376
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997957733489658e-05,
+      "loss": 1.0186,
+      "step": 1377
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997940463406646e-05,
+      "loss": 0.9654,
+      "step": 1378
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997923120617502e-05,
+      "loss": 1.0291,
+      "step": 1379
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997905705122352e-05,
+      "loss": 1.1169,
+      "step": 1380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999788821692132e-05,
+      "loss": 0.9294,
+      "step": 1381
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997870656014537e-05,
+      "loss": 1.0312,
+      "step": 1382
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997853022402126e-05,
+      "loss": 1.0381,
+      "step": 1383
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999783531608422e-05,
+      "loss": 1.1849,
+      "step": 1384
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997817537060945e-05,
+      "loss": 1.1931,
+      "step": 1385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997799685332427e-05,
+      "loss": 1.0392,
+      "step": 1386
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997781760898805e-05,
+      "loss": 1.1691,
+      "step": 1387
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997763763760202e-05,
+      "loss": 0.8152,
+      "step": 1388
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999774569391675e-05,
+      "loss": 1.0905,
+      "step": 1389
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999772755136858e-05,
+      "loss": 1.3333,
+      "step": 1390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997709336115827e-05,
+      "loss": 1.2176,
+      "step": 1391
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999769104815862e-05,
+      "loss": 1.1283,
+      "step": 1392
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999767268749709e-05,
+      "loss": 1.1896,
+      "step": 1393
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999765425413138e-05,
+      "loss": 1.056,
+      "step": 1394
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997635748061615e-05,
+      "loss": 1.0795,
+      "step": 1395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997617169287937e-05,
+      "loss": 1.0576,
+      "step": 1396
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997598517810473e-05,
+      "loss": 1.3364,
+      "step": 1397
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999757979362936e-05,
+      "loss": 1.0765,
+      "step": 1398
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997560996744737e-05,
+      "loss": 1.0046,
+      "step": 1399
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997542127156743e-05,
+      "loss": 1.2089,
+      "step": 1400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997523184865516e-05,
+      "loss": 1.3838,
+      "step": 1401
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997504169871184e-05,
+      "loss": 1.3847,
+      "step": 1402
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997485082173894e-05,
+      "loss": 1.089,
+      "step": 1403
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999746592177378e-05,
+      "loss": 1.289,
+      "step": 1404
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997446688670985e-05,
+      "loss": 0.7916,
+      "step": 1405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997427382865648e-05,
+      "loss": 1.0995,
+      "step": 1406
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999740800435791e-05,
+      "loss": 0.9187,
+      "step": 1407
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999738855314791e-05,
+      "loss": 1.1083,
+      "step": 1408
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999736902923579e-05,
+      "loss": 0.7855,
+      "step": 1409
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999734943262169e-05,
+      "loss": 0.9234,
+      "step": 1410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997329763305758e-05,
+      "loss": 1.1771,
+      "step": 1411
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997310021288133e-05,
+      "loss": 1.078,
+      "step": 1412
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999729020656896e-05,
+      "loss": 1.1717,
+      "step": 1413
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999727031914838e-05,
+      "loss": 1.2241,
+      "step": 1414
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997250359026542e-05,
+      "loss": 1.1932,
+      "step": 1415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999723032620359e-05,
+      "loss": 1.0542,
+      "step": 1416
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997210220679665e-05,
+      "loss": 1.2717,
+      "step": 1417
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997190042454923e-05,
+      "loss": 1.0668,
+      "step": 1418
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997169791529498e-05,
+      "loss": 1.2482,
+      "step": 1419
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999714946790355e-05,
+      "loss": 1.1491,
+      "step": 1420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997129071577216e-05,
+      "loss": 1.0173,
+      "step": 1421
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997108602550654e-05,
+      "loss": 1.1269,
+      "step": 1422
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997088060824005e-05,
+      "loss": 1.1297,
+      "step": 1423
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997067446397424e-05,
+      "loss": 1.3324,
+      "step": 1424
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997046759271055e-05,
+      "loss": 0.9554,
+      "step": 1425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997025999445054e-05,
+      "loss": 1.138,
+      "step": 1426
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999700516691957e-05,
+      "loss": 1.1493,
+      "step": 1427
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996984261694754e-05,
+      "loss": 1.1095,
+      "step": 1428
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996963283770756e-05,
+      "loss": 0.9003,
+      "step": 1429
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996942233147733e-05,
+      "loss": 1.2666,
+      "step": 1430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996921109825837e-05,
+      "loss": 0.8113,
+      "step": 1431
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996899913805217e-05,
+      "loss": 1.0943,
+      "step": 1432
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996878645086037e-05,
+      "loss": 1.0773,
+      "step": 1433
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999685730366844e-05,
+      "loss": 0.8857,
+      "step": 1434
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996835889552585e-05,
+      "loss": 1.0916,
+      "step": 1435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996814402738637e-05,
+      "loss": 0.9718,
+      "step": 1436
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996792843226736e-05,
+      "loss": 1.1417,
+      "step": 1437
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996771211017055e-05,
+      "loss": 1.1382,
+      "step": 1438
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996749506109737e-05,
+      "loss": 1.1973,
+      "step": 1439
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999672772850495e-05,
+      "loss": 1.11,
+      "step": 1440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996705878202846e-05,
+      "loss": 1.0825,
+      "step": 1441
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999668395520359e-05,
+      "loss": 0.7518,
+      "step": 1442
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996661959507336e-05,
+      "loss": 1.2609,
+      "step": 1443
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999663989111425e-05,
+      "loss": 0.9814,
+      "step": 1444
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996617750024484e-05,
+      "loss": 1.2997,
+      "step": 1445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996595536238206e-05,
+      "loss": 1.037,
+      "step": 1446
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996573249755573e-05,
+      "loss": 1.1774,
+      "step": 1447
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996550890576754e-05,
+      "loss": 1.1471,
+      "step": 1448
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99965284587019e-05,
+      "loss": 1.0606,
+      "step": 1449
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996505954131188e-05,
+      "loss": 1.1485,
+      "step": 1450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999648337686477e-05,
+      "loss": 1.0823,
+      "step": 1451
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996460726902816e-05,
+      "loss": 1.2956,
+      "step": 1452
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999643800424549e-05,
+      "loss": 0.9943,
+      "step": 1453
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996415208892953e-05,
+      "loss": 0.995,
+      "step": 1454
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999639234084538e-05,
+      "loss": 0.9919,
+      "step": 1455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999636940010293e-05,
+      "loss": 1.1777,
+      "step": 1456
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999634638666577e-05,
+      "loss": 1.2279,
+      "step": 1457
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999632330053407e-05,
+      "loss": 1.1803,
+      "step": 1458
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996300141708e-05,
+      "loss": 1.3461,
+      "step": 1459
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999627691018772e-05,
+      "loss": 0.8977,
+      "step": 1460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996253605973406e-05,
+      "loss": 0.8255,
+      "step": 1461
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996230229065226e-05,
+      "loss": 1.0349,
+      "step": 1462
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996206779463353e-05,
+      "loss": 1.1667,
+      "step": 1463
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996183257167948e-05,
+      "loss": 1.0969,
+      "step": 1464
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996159662179195e-05,
+      "loss": 1.4209,
+      "step": 1465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996135994497256e-05,
+      "loss": 1.1652,
+      "step": 1466
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996112254122304e-05,
+      "loss": 1.1973,
+      "step": 1467
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996088441054515e-05,
+      "loss": 1.0492,
+      "step": 1468
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996064555294063e-05,
+      "loss": 1.214,
+      "step": 1469
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996040596841118e-05,
+      "loss": 1.2639,
+      "step": 1470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996016565695853e-05,
+      "loss": 0.9908,
+      "step": 1471
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995992461858448e-05,
+      "loss": 1.1392,
+      "step": 1472
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995968285329078e-05,
+      "loss": 1.1717,
+      "step": 1473
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995944036107913e-05,
+      "loss": 1.2582,
+      "step": 1474
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995919714195133e-05,
+      "loss": 1.1129,
+      "step": 1475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995895319590914e-05,
+      "loss": 1.1903,
+      "step": 1476
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999587085229544e-05,
+      "loss": 1.1799,
+      "step": 1477
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995846312308876e-05,
+      "loss": 1.2283,
+      "step": 1478
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995821699631413e-05,
+      "loss": 1.1076,
+      "step": 1479
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995797014263223e-05,
+      "loss": 1.4272,
+      "step": 1480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995772256204488e-05,
+      "loss": 1.05,
+      "step": 1481
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995747425455384e-05,
+      "loss": 1.1268,
+      "step": 1482
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995722522016098e-05,
+      "loss": 1.4081,
+      "step": 1483
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995697545886803e-05,
+      "loss": 0.9227,
+      "step": 1484
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999567249706769e-05,
+      "loss": 0.9619,
+      "step": 1485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995647375558936e-05,
+      "loss": 0.8427,
+      "step": 1486
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995622181360724e-05,
+      "loss": 0.9519,
+      "step": 1487
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995596914473237e-05,
+      "loss": 1.1148,
+      "step": 1488
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995571574896656e-05,
+      "loss": 1.0353,
+      "step": 1489
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995546162631174e-05,
+      "loss": 1.2727,
+      "step": 1490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999552067767697e-05,
+      "loss": 1.19,
+      "step": 1491
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995495120034226e-05,
+      "loss": 1.2126,
+      "step": 1492
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995469489703134e-05,
+      "loss": 0.9671,
+      "step": 1493
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999544378668388e-05,
+      "loss": 0.9676,
+      "step": 1494
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995418010976643e-05,
+      "loss": 1.0228,
+      "step": 1495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999539216258162e-05,
+      "loss": 0.973,
+      "step": 1496
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995366241498995e-05,
+      "loss": 1.232,
+      "step": 1497
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995340247728955e-05,
+      "loss": 1.0657,
+      "step": 1498
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999531418127169e-05,
+      "loss": 1.1275,
+      "step": 1499
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995288042127396e-05,
+      "loss": 0.9188,
+      "step": 1500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995261830296252e-05,
+      "loss": 1.4431,
+      "step": 1501
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995235545778456e-05,
+      "loss": 1.174,
+      "step": 1502
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99952091885742e-05,
+      "loss": 1.7424,
+      "step": 1503
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999518275868367e-05,
+      "loss": 1.1415,
+      "step": 1504
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995156256107057e-05,
+      "loss": 1.2054,
+      "step": 1505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995129680844565e-05,
+      "loss": 1.0688,
+      "step": 1506
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995103032896374e-05,
+      "loss": 0.9315,
+      "step": 1507
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999507631226269e-05,
+      "loss": 1.1853,
+      "step": 1508
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995049518943698e-05,
+      "loss": 1.2664,
+      "step": 1509
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995022652939594e-05,
+      "loss": 1.0364,
+      "step": 1510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999499571425058e-05,
+      "loss": 1.3188,
+      "step": 1511
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999496870287684e-05,
+      "loss": 0.7992,
+      "step": 1512
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999494161881858e-05,
+      "loss": 0.96,
+      "step": 1513
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994914462075997e-05,
+      "loss": 1.292,
+      "step": 1514
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999488723264929e-05,
+      "loss": 1.0348,
+      "step": 1515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994859930538648e-05,
+      "loss": 1.2891,
+      "step": 1516
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994832555744274e-05,
+      "loss": 1.1904,
+      "step": 1517
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999480510826637e-05,
+      "loss": 1.4191,
+      "step": 1518
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994777588105134e-05,
+      "loss": 1.3574,
+      "step": 1519
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994749995260763e-05,
+      "loss": 1.0319,
+      "step": 1520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999472232973346e-05,
+      "loss": 0.9876,
+      "step": 1521
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999469459152343e-05,
+      "loss": 1.1674,
+      "step": 1522
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994666780630867e-05,
+      "loss": 0.8545,
+      "step": 1523
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999463889705598e-05,
+      "loss": 1.212,
+      "step": 1524
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994610940798967e-05,
+      "loss": 0.968,
+      "step": 1525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994582911860035e-05,
+      "loss": 1.0869,
+      "step": 1526
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994554810239385e-05,
+      "loss": 1.1101,
+      "step": 1527
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999452663593722e-05,
+      "loss": 0.9114,
+      "step": 1528
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994498388953753e-05,
+      "loss": 1.0902,
+      "step": 1529
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999447006928918e-05,
+      "loss": 1.3939,
+      "step": 1530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994441676943715e-05,
+      "loss": 1.1024,
+      "step": 1531
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994413211917555e-05,
+      "loss": 0.9184,
+      "step": 1532
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994384674210917e-05,
+      "loss": 1.1315,
+      "step": 1533
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994356063824e-05,
+      "loss": 1.1665,
+      "step": 1534
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994327380757017e-05,
+      "loss": 1.2697,
+      "step": 1535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994298625010176e-05,
+      "loss": 1.0013,
+      "step": 1536
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994269796583686e-05,
+      "loss": 1.1478,
+      "step": 1537
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994240895477755e-05,
+      "loss": 0.9132,
+      "step": 1538
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994211921692595e-05,
+      "loss": 1.2188,
+      "step": 1539
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994182875228417e-05,
+      "loss": 0.9362,
+      "step": 1540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999415375608543e-05,
+      "loss": 1.2116,
+      "step": 1541
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994124564263847e-05,
+      "loss": 0.7864,
+      "step": 1542
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999409529976388e-05,
+      "loss": 0.9323,
+      "step": 1543
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994065962585744e-05,
+      "loss": 0.9002,
+      "step": 1544
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999403655272965e-05,
+      "loss": 1.0923,
+      "step": 1545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999400707019581e-05,
+      "loss": 0.936,
+      "step": 1546
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993977514984444e-05,
+      "loss": 1.1055,
+      "step": 1547
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993947887095763e-05,
+      "loss": 0.9508,
+      "step": 1548
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999391818652998e-05,
+      "loss": 1.1053,
+      "step": 1549
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999388841328732e-05,
+      "loss": 0.782,
+      "step": 1550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993858567367988e-05,
+      "loss": 0.7153,
+      "step": 1551
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999382864877221e-05,
+      "loss": 1.1427,
+      "step": 1552
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993798657500197e-05,
+      "loss": 1.0163,
+      "step": 1553
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993768593552175e-05,
+      "loss": 1.0055,
+      "step": 1554
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993738456928354e-05,
+      "loss": 1.0626,
+      "step": 1555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993708247628957e-05,
+      "loss": 1.0073,
+      "step": 1556
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993677965654208e-05,
+      "loss": 1.1666,
+      "step": 1557
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999364761100432e-05,
+      "loss": 0.9275,
+      "step": 1558
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999361718367952e-05,
+      "loss": 1.3338,
+      "step": 1559
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999358668368002e-05,
+      "loss": 1.0597,
+      "step": 1560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993556111006054e-05,
+      "loss": 1.4465,
+      "step": 1561
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993525465657836e-05,
+      "loss": 0.9418,
+      "step": 1562
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999349474763559e-05,
+      "loss": 1.048,
+      "step": 1563
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999346395693954e-05,
+      "loss": 0.7536,
+      "step": 1564
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999343309356991e-05,
+      "loss": 1.1767,
+      "step": 1565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993402157526928e-05,
+      "loss": 0.8506,
+      "step": 1566
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993371148810815e-05,
+      "loss": 1.1862,
+      "step": 1567
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993340067421796e-05,
+      "loss": 0.7096,
+      "step": 1568
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99933089133601e-05,
+      "loss": 1.2287,
+      "step": 1569
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993277686625948e-05,
+      "loss": 0.9763,
+      "step": 1570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993246387219573e-05,
+      "loss": 1.2408,
+      "step": 1571
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993215015141203e-05,
+      "loss": 1.2376,
+      "step": 1572
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993183570391062e-05,
+      "loss": 0.8417,
+      "step": 1573
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999315205296938e-05,
+      "loss": 0.9464,
+      "step": 1574
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993120462876385e-05,
+      "loss": 1.0813,
+      "step": 1575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999308880011231e-05,
+      "loss": 1.362,
+      "step": 1576
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993057064677385e-05,
+      "loss": 1.1013,
+      "step": 1577
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993025256571838e-05,
+      "loss": 0.7969,
+      "step": 1578
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99929933757959e-05,
+      "loss": 0.863,
+      "step": 1579
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992961422349808e-05,
+      "loss": 1.1561,
+      "step": 1580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992929396233785e-05,
+      "loss": 1.4757,
+      "step": 1581
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992897297448076e-05,
+      "loss": 0.9793,
+      "step": 1582
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992865125992906e-05,
+      "loss": 0.9574,
+      "step": 1583
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999283288186851e-05,
+      "loss": 0.9989,
+      "step": 1584
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992800565075123e-05,
+      "loss": 1.1583,
+      "step": 1585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999276817561298e-05,
+      "loss": 1.4204,
+      "step": 1586
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999273571348232e-05,
+      "loss": 1.1746,
+      "step": 1587
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992703178683376e-05,
+      "loss": 1.1429,
+      "step": 1588
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992670571216384e-05,
+      "loss": 1.2516,
+      "step": 1589
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992637891081585e-05,
+      "loss": 1.133,
+      "step": 1590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999260513827921e-05,
+      "loss": 0.9257,
+      "step": 1591
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99925723128095e-05,
+      "loss": 1.0445,
+      "step": 1592
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992539414672696e-05,
+      "loss": 1.1989,
+      "step": 1593
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992506443869034e-05,
+      "loss": 0.8889,
+      "step": 1594
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999247340039876e-05,
+      "loss": 1.0054,
+      "step": 1595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992440284262104e-05,
+      "loss": 1.0832,
+      "step": 1596
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999240709545932e-05,
+      "loss": 1.1206,
+      "step": 1597
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992373833990633e-05,
+      "loss": 0.9658,
+      "step": 1598
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992340499856298e-05,
+      "loss": 1.0058,
+      "step": 1599
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992307093056555e-05,
+      "loss": 1.0248,
+      "step": 1600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992273613591644e-05,
+      "loss": 1.1169,
+      "step": 1601
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999224006146181e-05,
+      "loss": 1.1035,
+      "step": 1602
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992206436667295e-05,
+      "loss": 1.1379,
+      "step": 1603
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992172739208348e-05,
+      "loss": 1.1886,
+      "step": 1604
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999213896908521e-05,
+      "loss": 0.9253,
+      "step": 1605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992105126298127e-05,
+      "loss": 1.0619,
+      "step": 1606
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992071210847346e-05,
+      "loss": 1.1855,
+      "step": 1607
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992037222733113e-05,
+      "loss": 1.1562,
+      "step": 1608
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992003161955677e-05,
+      "loss": 0.8318,
+      "step": 1609
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991969028515282e-05,
+      "loss": 1.3783,
+      "step": 1610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991934822412183e-05,
+      "loss": 1.0023,
+      "step": 1611
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991900543646624e-05,
+      "loss": 1.0694,
+      "step": 1612
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999186619221885e-05,
+      "loss": 1.0327,
+      "step": 1613
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999183176812912e-05,
+      "loss": 1.0453,
+      "step": 1614
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991797271377676e-05,
+      "loss": 1.3319,
+      "step": 1615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991762701964775e-05,
+      "loss": 0.8679,
+      "step": 1616
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991728059890668e-05,
+      "loss": 1.0058,
+      "step": 1617
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99916933451556e-05,
+      "loss": 1.1046,
+      "step": 1618
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999165855775983e-05,
+      "loss": 1.0201,
+      "step": 1619
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991623697703613e-05,
+      "loss": 1.016,
+      "step": 1620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99915887649872e-05,
+      "loss": 0.8036,
+      "step": 1621
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991553759610838e-05,
+      "loss": 1.2449,
+      "step": 1622
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999151868157479e-05,
+      "loss": 1.1021,
+      "step": 1623
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999148353087931e-05,
+      "loss": 1.122,
+      "step": 1624
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999144830752465e-05,
+      "loss": 1.2944,
+      "step": 1625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991413011511068e-05,
+      "loss": 1.0677,
+      "step": 1626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991377642838823e-05,
+      "loss": 1.0025,
+      "step": 1627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999134220150817e-05,
+      "loss": 1.15,
+      "step": 1628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991306687519366e-05,
+      "loss": 0.805,
+      "step": 1629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991271100872673e-05,
+      "loss": 1.3236,
+      "step": 1630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991235441568343e-05,
+      "loss": 1.2936,
+      "step": 1631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991199709606642e-05,
+      "loss": 1.2737,
+      "step": 1632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991163904987827e-05,
+      "loss": 1.1123,
+      "step": 1633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991128027712156e-05,
+      "loss": 1.0469,
+      "step": 1634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991092077779895e-05,
+      "loss": 0.9782,
+      "step": 1635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991056055191304e-05,
+      "loss": 1.2503,
+      "step": 1636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999101995994664e-05,
+      "loss": 0.8285,
+      "step": 1637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990983792046176e-05,
+      "loss": 1.0134,
+      "step": 1638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990947551490162e-05,
+      "loss": 1.388,
+      "step": 1639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999091123827887e-05,
+      "loss": 1.2947,
+      "step": 1640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990874852412563e-05,
+      "loss": 1.2509,
+      "step": 1641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990838393891505e-05,
+      "loss": 1.1865,
+      "step": 1642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999080186271596e-05,
+      "loss": 1.0412,
+      "step": 1643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999076525888619e-05,
+      "loss": 1.2833,
+      "step": 1644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999072858240247e-05,
+      "loss": 1.2274,
+      "step": 1645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990691833265064e-05,
+      "loss": 1.3615,
+      "step": 1646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990655011474235e-05,
+      "loss": 0.9981,
+      "step": 1647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990618117030254e-05,
+      "loss": 0.8769,
+      "step": 1648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999058114993339e-05,
+      "loss": 1.0496,
+      "step": 1649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990544110183907e-05,
+      "loss": 1.3178,
+      "step": 1650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999050699778208e-05,
+      "loss": 1.038,
+      "step": 1651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990469812728174e-05,
+      "loss": 1.1808,
+      "step": 1652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990432555022464e-05,
+      "loss": 1.3547,
+      "step": 1653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999039522466522e-05,
+      "loss": 0.9559,
+      "step": 1654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999035782165671e-05,
+      "loss": 1.3506,
+      "step": 1655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990320345997213e-05,
+      "loss": 1.2245,
+      "step": 1656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990282797686995e-05,
+      "loss": 0.8915,
+      "step": 1657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999024517672633e-05,
+      "loss": 1.1189,
+      "step": 1658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990207483115494e-05,
+      "loss": 1.2399,
+      "step": 1659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999016971685476e-05,
+      "loss": 0.9359,
+      "step": 1660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99901318779444e-05,
+      "loss": 1.002,
+      "step": 1661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990093966384695e-05,
+      "loss": 1.2597,
+      "step": 1662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990055982175913e-05,
+      "loss": 1.1865,
+      "step": 1663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990017925318338e-05,
+      "loss": 1.0799,
+      "step": 1664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998997979581224e-05,
+      "loss": 1.0494,
+      "step": 1665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989941593657903e-05,
+      "loss": 0.8845,
+      "step": 1666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99899033188556e-05,
+      "loss": 1.0104,
+      "step": 1667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989864971405612e-05,
+      "loss": 1.1026,
+      "step": 1668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989826551308217e-05,
+      "loss": 1.0971,
+      "step": 1669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989788058563692e-05,
+      "loss": 1.2488,
+      "step": 1670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989749493172316e-05,
+      "loss": 0.9818,
+      "step": 1671
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998971085513438e-05,
+      "loss": 1.2969,
+      "step": 1672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989672144450152e-05,
+      "loss": 0.9308,
+      "step": 1673
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989633361119916e-05,
+      "loss": 0.9353,
+      "step": 1674
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989594505143964e-05,
+      "loss": 1.0325,
+      "step": 1675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989555576522567e-05,
+      "loss": 1.0761,
+      "step": 1676
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989516575256015e-05,
+      "loss": 1.0111,
+      "step": 1677
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989477501344584e-05,
+      "loss": 1.0791,
+      "step": 1678
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989438354788572e-05,
+      "loss": 0.959,
+      "step": 1679
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989399135588246e-05,
+      "loss": 0.9498,
+      "step": 1680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989359843743905e-05,
+      "loss": 0.9789,
+      "step": 1681
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989320479255832e-05,
+      "loss": 1.3286,
+      "step": 1682
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989281042124307e-05,
+      "loss": 1.2085,
+      "step": 1683
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989241532349623e-05,
+      "loss": 0.9246,
+      "step": 1684
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989201949932066e-05,
+      "loss": 1.2253,
+      "step": 1685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998916229487192e-05,
+      "loss": 1.3397,
+      "step": 1686
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989122567169475e-05,
+      "loss": 1.1694,
+      "step": 1687
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989082766825028e-05,
+      "loss": 1.1646,
+      "step": 1688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998904289383886e-05,
+      "loss": 0.9615,
+      "step": 1689
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998900294821126e-05,
+      "loss": 1.4539,
+      "step": 1690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988962929942525e-05,
+      "loss": 1.354,
+      "step": 1691
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998892283903294e-05,
+      "loss": 1.2681,
+      "step": 1692
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988882675482797e-05,
+      "loss": 1.2293,
+      "step": 1693
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998884243929239e-05,
+      "loss": 1.058,
+      "step": 1694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988802130462017e-05,
+      "loss": 1.122,
+      "step": 1695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998876174899196e-05,
+      "loss": 1.1182,
+      "step": 1696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998872129488252e-05,
+      "loss": 1.2132,
+      "step": 1697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988680768133992e-05,
+      "loss": 1.2092,
+      "step": 1698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988640168746666e-05,
+      "loss": 1.0607,
+      "step": 1699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988599496720836e-05,
+      "loss": 0.9808,
+      "step": 1700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998855875205681e-05,
+      "loss": 1.0794,
+      "step": 1701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998851793475487e-05,
+      "loss": 1.1726,
+      "step": 1702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988477044815318e-05,
+      "loss": 0.8629,
+      "step": 1703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998843608223845e-05,
+      "loss": 1.2011,
+      "step": 1704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988395047024572e-05,
+      "loss": 1.0673,
+      "step": 1705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998835393917397e-05,
+      "loss": 1.3007,
+      "step": 1706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988312758686953e-05,
+      "loss": 1.0527,
+      "step": 1707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988271505563812e-05,
+      "loss": 1.2461,
+      "step": 1708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988230179804856e-05,
+      "loss": 1.4143,
+      "step": 1709
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988188781410377e-05,
+      "loss": 1.0833,
+      "step": 1710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988147310380683e-05,
+      "loss": 1.3033,
+      "step": 1711
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988105766716073e-05,
+      "loss": 1.4594,
+      "step": 1712
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988064150416847e-05,
+      "loss": 1.2968,
+      "step": 1713
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998802246148331e-05,
+      "loss": 0.9565,
+      "step": 1714
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987980699915767e-05,
+      "loss": 1.1562,
+      "step": 1715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987938865714512e-05,
+      "loss": 0.9104,
+      "step": 1716
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998789695887986e-05,
+      "loss": 1.4653,
+      "step": 1717
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987854979412115e-05,
+      "loss": 0.8281,
+      "step": 1718
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987812927311577e-05,
+      "loss": 1.3751,
+      "step": 1719
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987770802578557e-05,
+      "loss": 0.9755,
+      "step": 1720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987728605213354e-05,
+      "loss": 1.2131,
+      "step": 1721
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987686335216284e-05,
+      "loss": 1.3846,
+      "step": 1722
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987643992587647e-05,
+      "loss": 0.9685,
+      "step": 1723
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987601577327754e-05,
+      "loss": 1.1225,
+      "step": 1724
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987559089436917e-05,
+      "loss": 0.9758,
+      "step": 1725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987516528915435e-05,
+      "loss": 1.1199,
+      "step": 1726
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987473895763628e-05,
+      "loss": 1.2408,
+      "step": 1727
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99874311899818e-05,
+      "loss": 1.0347,
+      "step": 1728
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987388411570265e-05,
+      "loss": 1.1856,
+      "step": 1729
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998734556052933e-05,
+      "loss": 0.8289,
+      "step": 1730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987302636859312e-05,
+      "loss": 1.1132,
+      "step": 1731
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987259640560517e-05,
+      "loss": 1.0114,
+      "step": 1732
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987216571633264e-05,
+      "loss": 1.1646,
+      "step": 1733
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998717343007786e-05,
+      "loss": 1.0668,
+      "step": 1734
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987130215894623e-05,
+      "loss": 1.4439,
+      "step": 1735
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987086929083866e-05,
+      "loss": 0.8977,
+      "step": 1736
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987043569645904e-05,
+      "loss": 1.135,
+      "step": 1737
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998700013758105e-05,
+      "loss": 0.9504,
+      "step": 1738
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986956632889628e-05,
+      "loss": 1.3318,
+      "step": 1739
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998691305557194e-05,
+      "loss": 1.1045,
+      "step": 1740
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998686940562832e-05,
+      "loss": 1.3403,
+      "step": 1741
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998682568305907e-05,
+      "loss": 0.9612,
+      "step": 1742
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986781887864514e-05,
+      "loss": 1.0336,
+      "step": 1743
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986738020044977e-05,
+      "loss": 1.3329,
+      "step": 1744
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986694079600768e-05,
+      "loss": 0.8967,
+      "step": 1745
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998665006653221e-05,
+      "loss": 1.007,
+      "step": 1746
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986605980839626e-05,
+      "loss": 0.9454,
+      "step": 1747
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986561822523333e-05,
+      "loss": 0.8641,
+      "step": 1748
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986517591583656e-05,
+      "loss": 0.9683,
+      "step": 1749
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998647328802091e-05,
+      "loss": 0.9627,
+      "step": 1750
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986428911835425e-05,
+      "loss": 1.497,
+      "step": 1751
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998638446302752e-05,
+      "loss": 1.2471,
+      "step": 1752
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986339941597517e-05,
+      "loss": 1.0652,
+      "step": 1753
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986295347545738e-05,
+      "loss": 1.2105,
+      "step": 1754
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986250680872515e-05,
+      "loss": 0.9757,
+      "step": 1755
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986205941578167e-05,
+      "loss": 0.976,
+      "step": 1756
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998616112966302e-05,
+      "loss": 1.0347,
+      "step": 1757
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986116245127397e-05,
+      "loss": 1.1628,
+      "step": 1758
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998607128797163e-05,
+      "loss": 1.1424,
+      "step": 1759
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986026258196048e-05,
+      "loss": 0.8287,
+      "step": 1760
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998598115580097e-05,
+      "loss": 1.1194,
+      "step": 1761
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985935980786728e-05,
+      "loss": 1.0815,
+      "step": 1762
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998589073315365e-05,
+      "loss": 1.0141,
+      "step": 1763
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985845412902067e-05,
+      "loss": 0.779,
+      "step": 1764
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998580002003231e-05,
+      "loss": 0.9735,
+      "step": 1765
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99857545545447e-05,
+      "loss": 0.9289,
+      "step": 1766
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998570901643958e-05,
+      "loss": 1.1869,
+      "step": 1767
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985663405717272e-05,
+      "loss": 1.2933,
+      "step": 1768
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998561772237811e-05,
+      "loss": 1.035,
+      "step": 1769
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998557196642243e-05,
+      "loss": 1.0687,
+      "step": 1770
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985526137850558e-05,
+      "loss": 1.0916,
+      "step": 1771
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985480236662834e-05,
+      "loss": 1.1263,
+      "step": 1772
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998543426285959e-05,
+      "loss": 0.9741,
+      "step": 1773
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985388216441157e-05,
+      "loss": 1.3301,
+      "step": 1774
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985342097407872e-05,
+      "loss": 1.1715,
+      "step": 1775
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985295905760072e-05,
+      "loss": 1.0918,
+      "step": 1776
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998524964149809e-05,
+      "loss": 0.992,
+      "step": 1777
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998520330462226e-05,
+      "loss": 1.0446,
+      "step": 1778
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985156895132932e-05,
+      "loss": 0.8639,
+      "step": 1779
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985110413030426e-05,
+      "loss": 0.912,
+      "step": 1780
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985063858315092e-05,
+      "loss": 1.0423,
+      "step": 1781
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985017230987266e-05,
+      "loss": 1.0643,
+      "step": 1782
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984970531047283e-05,
+      "loss": 1.1231,
+      "step": 1783
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984923758495485e-05,
+      "loss": 1.1204,
+      "step": 1784
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984876913332215e-05,
+      "loss": 1.1866,
+      "step": 1785
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998482999555781e-05,
+      "loss": 1.3895,
+      "step": 1786
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984783005172616e-05,
+      "loss": 1.2336,
+      "step": 1787
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984735942176964e-05,
+      "loss": 0.8746,
+      "step": 1788
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998468880657121e-05,
+      "loss": 1.0611,
+      "step": 1789
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998464159835569e-05,
+      "loss": 0.8999,
+      "step": 1790
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998459431753074e-05,
+      "loss": 1.2681,
+      "step": 1791
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998454696409672e-05,
+      "loss": 0.8602,
+      "step": 1792
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998449953805396e-05,
+      "loss": 0.8457,
+      "step": 1793
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984452039402814e-05,
+      "loss": 1.2257,
+      "step": 1794
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998440446814362e-05,
+      "loss": 1.0431,
+      "step": 1795
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984356824276734e-05,
+      "loss": 1.1937,
+      "step": 1796
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984309107802492e-05,
+      "loss": 1.0856,
+      "step": 1797
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984261318721247e-05,
+      "loss": 1.2106,
+      "step": 1798
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984213457033347e-05,
+      "loss": 1.434,
+      "step": 1799
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984165522739135e-05,
+      "loss": 1.0884,
+      "step": 1800
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984117515838962e-05,
+      "loss": 0.9444,
+      "step": 1801
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998406943633318e-05,
+      "loss": 1.0978,
+      "step": 1802
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984021284222134e-05,
+      "loss": 1.0958,
+      "step": 1803
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998397305950618e-05,
+      "loss": 1.3674,
+      "step": 1804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998392476218566e-05,
+      "loss": 0.9597,
+      "step": 1805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983876392260935e-05,
+      "loss": 1.2274,
+      "step": 1806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983827949732352e-05,
+      "loss": 1.0383,
+      "step": 1807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998377943460026e-05,
+      "loss": 1.3432,
+      "step": 1808
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998373084686502e-05,
+      "loss": 1.0412,
+      "step": 1809
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983682186526976e-05,
+      "loss": 0.8891,
+      "step": 1810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998363345358649e-05,
+      "loss": 0.9802,
+      "step": 1811
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998358464804391e-05,
+      "loss": 0.981,
+      "step": 1812
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983535769899595e-05,
+      "loss": 1.6183,
+      "step": 1813
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983486819153902e-05,
+      "loss": 1.0241,
+      "step": 1814
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998343779580718e-05,
+      "loss": 1.1042,
+      "step": 1815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998338869985979e-05,
+      "loss": 1.2418,
+      "step": 1816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998333953131209e-05,
+      "loss": 1.0414,
+      "step": 1817
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998329029016444e-05,
+      "loss": 1.098,
+      "step": 1818
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998324097641719e-05,
+      "loss": 1.2767,
+      "step": 1819
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983191590070704e-05,
+      "loss": 1.0792,
+      "step": 1820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998314213112534e-05,
+      "loss": 1.2967,
+      "step": 1821
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983092599581457e-05,
+      "loss": 1.2078,
+      "step": 1822
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983042995439418e-05,
+      "loss": 1.3285,
+      "step": 1823
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998299331869958e-05,
+      "loss": 1.3527,
+      "step": 1824
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982943569362305e-05,
+      "loss": 1.2173,
+      "step": 1825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982893747427957e-05,
+      "loss": 1.0571,
+      "step": 1826
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99828438528969e-05,
+      "loss": 1.1577,
+      "step": 1827
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998279388576949e-05,
+      "loss": 0.8728,
+      "step": 1828
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982743846046095e-05,
+      "loss": 1.0621,
+      "step": 1829
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998269373372708e-05,
+      "loss": 0.9038,
+      "step": 1830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982643548812808e-05,
+      "loss": 1.0698,
+      "step": 1831
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998259329130364e-05,
+      "loss": 0.8518,
+      "step": 1832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982542961199946e-05,
+      "loss": 0.8711,
+      "step": 1833
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982492558502096e-05,
+      "loss": 1.1311,
+      "step": 1834
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998244208321045e-05,
+      "loss": 1.1266,
+      "step": 1835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998239153532537e-05,
+      "loss": 1.0137,
+      "step": 1836
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982340914847235e-05,
+      "loss": 0.9829,
+      "step": 1837
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982290221776407e-05,
+      "loss": 0.8996,
+      "step": 1838
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982239456113256e-05,
+      "loss": 1.1209,
+      "step": 1839
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982188617858155e-05,
+      "loss": 1.2487,
+      "step": 1840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982137707011466e-05,
+      "loss": 1.2894,
+      "step": 1841
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982086723573567e-05,
+      "loss": 1.0039,
+      "step": 1842
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982035667544818e-05,
+      "loss": 1.1578,
+      "step": 1843
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981984538925603e-05,
+      "loss": 1.112,
+      "step": 1844
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981933337716288e-05,
+      "loss": 0.9937,
+      "step": 1845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981882063917242e-05,
+      "loss": 1.0449,
+      "step": 1846
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981830717528844e-05,
+      "loss": 1.058,
+      "step": 1847
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998177929855146e-05,
+      "loss": 0.9652,
+      "step": 1848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981727806985473e-05,
+      "loss": 0.8951,
+      "step": 1849
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998167624283125e-05,
+      "loss": 1.0628,
+      "step": 1850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998162460608917e-05,
+      "loss": 1.1042,
+      "step": 1851
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998157289675961e-05,
+      "loss": 0.9839,
+      "step": 1852
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981521114842942e-05,
+      "loss": 1.1041,
+      "step": 1853
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981469260339544e-05,
+      "loss": 1.2857,
+      "step": 1854
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998141733324979e-05,
+      "loss": 1.139,
+      "step": 1855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981365333574066e-05,
+      "loss": 0.8382,
+      "step": 1856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998131326131274e-05,
+      "loss": 1.1966,
+      "step": 1857
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981261116466198e-05,
+      "loss": 0.944,
+      "step": 1858
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981208899034817e-05,
+      "loss": 1.0404,
+      "step": 1859
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981156609018977e-05,
+      "loss": 1.072,
+      "step": 1860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981104246419057e-05,
+      "loss": 1.0817,
+      "step": 1861
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981051811235438e-05,
+      "loss": 1.0395,
+      "step": 1862
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99809993034685e-05,
+      "loss": 1.2325,
+      "step": 1863
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998094672311863e-05,
+      "loss": 1.1833,
+      "step": 1864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980894070186204e-05,
+      "loss": 1.2314,
+      "step": 1865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980841344671614e-05,
+      "loss": 0.8501,
+      "step": 1866
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998078854657523e-05,
+      "loss": 1.3013,
+      "step": 1867
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998073567589745e-05,
+      "loss": 1.0738,
+      "step": 1868
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980682732638648e-05,
+      "loss": 0.8985,
+      "step": 1869
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998062971679921e-05,
+      "loss": 1.2084,
+      "step": 1870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998057662837953e-05,
+      "loss": 1.0929,
+      "step": 1871
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980523467379986e-05,
+      "loss": 0.9993,
+      "step": 1872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998047023380097e-05,
+      "loss": 0.8268,
+      "step": 1873
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980416927642864e-05,
+      "loss": 0.6244,
+      "step": 1874
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980363548906056e-05,
+      "loss": 0.7906,
+      "step": 1875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980310097590935e-05,
+      "loss": 0.9631,
+      "step": 1876
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980256573697894e-05,
+      "loss": 1.1226,
+      "step": 1877
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998020297722732e-05,
+      "loss": 1.0256,
+      "step": 1878
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980149308179598e-05,
+      "loss": 1.3132,
+      "step": 1879
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980095566555122e-05,
+      "loss": 1.0766,
+      "step": 1880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980041752354285e-05,
+      "loss": 1.1727,
+      "step": 1881
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979987865577476e-05,
+      "loss": 0.9051,
+      "step": 1882
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979933906225085e-05,
+      "loss": 1.0593,
+      "step": 1883
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997987987429751e-05,
+      "loss": 0.9747,
+      "step": 1884
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979825769795136e-05,
+      "loss": 0.9151,
+      "step": 1885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997977159271836e-05,
+      "loss": 0.8965,
+      "step": 1886
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979717343067582e-05,
+      "loss": 0.9703,
+      "step": 1887
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997966302084319e-05,
+      "loss": 1.2903,
+      "step": 1888
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979608626045578e-05,
+      "loss": 1.5569,
+      "step": 1889
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979554158675145e-05,
+      "loss": 1.1699,
+      "step": 1890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979499618732287e-05,
+      "loss": 0.9024,
+      "step": 1891
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979445006217398e-05,
+      "loss": 1.2143,
+      "step": 1892
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979390321130875e-05,
+      "loss": 0.9742,
+      "step": 1893
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979335563473123e-05,
+      "loss": 1.2646,
+      "step": 1894
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997928073324453e-05,
+      "loss": 1.2846,
+      "step": 1895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99792258304455e-05,
+      "loss": 1.081,
+      "step": 1896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997917085507643e-05,
+      "loss": 1.3479,
+      "step": 1897
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979115807137727e-05,
+      "loss": 0.8582,
+      "step": 1898
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997906068662978e-05,
+      "loss": 1.3015,
+      "step": 1899
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979005493552996e-05,
+      "loss": 1.1465,
+      "step": 1900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997895022790778e-05,
+      "loss": 1.1327,
+      "step": 1901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978894889694527e-05,
+      "loss": 1.1434,
+      "step": 1902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978839478913644e-05,
+      "loss": 0.866,
+      "step": 1903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997878399556553e-05,
+      "loss": 0.9927,
+      "step": 1904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997872843965059e-05,
+      "loss": 1.3275,
+      "step": 1905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997867281116923e-05,
+      "loss": 0.9746,
+      "step": 1906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978617110121858e-05,
+      "loss": 1.0005,
+      "step": 1907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978561336508865e-05,
+      "loss": 0.9679,
+      "step": 1908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978505490330672e-05,
+      "loss": 1.3542,
+      "step": 1909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978449571587678e-05,
+      "loss": 1.1858,
+      "step": 1910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978393580280293e-05,
+      "loss": 1.3948,
+      "step": 1911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978337516408913e-05,
+      "loss": 0.9505,
+      "step": 1912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978281379973962e-05,
+      "loss": 1.1211,
+      "step": 1913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997822517097584e-05,
+      "loss": 1.2211,
+      "step": 1914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978168889414956e-05,
+      "loss": 0.9361,
+      "step": 1915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997811253529172e-05,
+      "loss": 0.9579,
+      "step": 1916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997805610860654e-05,
+      "loss": 1.0768,
+      "step": 1917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997799960935983e-05,
+      "loss": 1.2754,
+      "step": 1918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977943037551998e-05,
+      "loss": 1.1672,
+      "step": 1919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977886393183454e-05,
+      "loss": 1.3137,
+      "step": 1920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977829676254615e-05,
+      "loss": 1.1547,
+      "step": 1921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997777288676589e-05,
+      "loss": 1.0181,
+      "step": 1922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977716024717693e-05,
+      "loss": 1.0304,
+      "step": 1923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977659090110436e-05,
+      "loss": 0.9665,
+      "step": 1924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977602082944535e-05,
+      "loss": 1.339,
+      "step": 1925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977545003220405e-05,
+      "loss": 0.8859,
+      "step": 1926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977487850938458e-05,
+      "loss": 1.2161,
+      "step": 1927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997743062609911e-05,
+      "loss": 0.8694,
+      "step": 1928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997737332870278e-05,
+      "loss": 0.9201,
+      "step": 1929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997731595874988e-05,
+      "loss": 1.0004,
+      "step": 1930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977258516240836e-05,
+      "loss": 0.8708,
+      "step": 1931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977201001176056e-05,
+      "loss": 1.257,
+      "step": 1932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977143413555966e-05,
+      "loss": 1.0968,
+      "step": 1933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977085753380978e-05,
+      "loss": 0.9924,
+      "step": 1934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977028020651516e-05,
+      "loss": 0.9704,
+      "step": 1935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976970215367996e-05,
+      "loss": 1.3622,
+      "step": 1936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976912337530843e-05,
+      "loss": 1.1551,
+      "step": 1937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976854387140477e-05,
+      "loss": 1.1281,
+      "step": 1938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976796364197314e-05,
+      "loss": 0.7375,
+      "step": 1939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976738268701784e-05,
+      "loss": 1.0617,
+      "step": 1940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976680100654304e-05,
+      "loss": 0.9838,
+      "step": 1941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976621860055298e-05,
+      "loss": 0.904,
+      "step": 1942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997656354690519e-05,
+      "loss": 1.2209,
+      "step": 1943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976505161204405e-05,
+      "loss": 0.7966,
+      "step": 1944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976446702953364e-05,
+      "loss": 0.8732,
+      "step": 1945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976388172152495e-05,
+      "loss": 1.2579,
+      "step": 1946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976329568802228e-05,
+      "loss": 0.8495,
+      "step": 1947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976270892902978e-05,
+      "loss": 0.9928,
+      "step": 1948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997621214445518e-05,
+      "loss": 1.421,
+      "step": 1949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976153323459262e-05,
+      "loss": 1.137,
+      "step": 1950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997609442991565e-05,
+      "loss": 0.9948,
+      "step": 1951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976035463824768e-05,
+      "loss": 1.171,
+      "step": 1952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997597642518705e-05,
+      "loss": 0.7784,
+      "step": 1953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975917314002923e-05,
+      "loss": 0.9621,
+      "step": 1954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975858130272818e-05,
+      "loss": 1.0722,
+      "step": 1955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975798873997164e-05,
+      "loss": 1.1919,
+      "step": 1956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975739545176394e-05,
+      "loss": 1.034,
+      "step": 1957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997568014381094e-05,
+      "loss": 0.9861,
+      "step": 1958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975620669901228e-05,
+      "loss": 0.849,
+      "step": 1959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975561123447695e-05,
+      "loss": 1.2914,
+      "step": 1960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975501504450778e-05,
+      "loss": 0.9019,
+      "step": 1961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975441812910903e-05,
+      "loss": 0.9717,
+      "step": 1962
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975382048828508e-05,
+      "loss": 1.1392,
+      "step": 1963
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975322212204025e-05,
+      "loss": 1.0778,
+      "step": 1964
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975262303037896e-05,
+      "loss": 1.0416,
+      "step": 1965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975202321330548e-05,
+      "loss": 1.2467,
+      "step": 1966
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975142267082424e-05,
+      "loss": 1.0713,
+      "step": 1967
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975082140293954e-05,
+      "loss": 0.7529,
+      "step": 1968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975021940965583e-05,
+      "loss": 0.9466,
+      "step": 1969
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997496166909774e-05,
+      "loss": 1.3995,
+      "step": 1970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974901324690873e-05,
+      "loss": 1.1858,
+      "step": 1971
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974840907745413e-05,
+      "loss": 1.3521,
+      "step": 1972
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974780418261802e-05,
+      "loss": 1.1956,
+      "step": 1973
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974719856240482e-05,
+      "loss": 1.0208,
+      "step": 1974
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997465922168189e-05,
+      "loss": 0.9555,
+      "step": 1975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997459851458647e-05,
+      "loss": 1.2421,
+      "step": 1976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997453773495466e-05,
+      "loss": 1.2211,
+      "step": 1977
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974476882786903e-05,
+      "loss": 0.9727,
+      "step": 1978
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974415958083646e-05,
+      "loss": 0.9876,
+      "step": 1979
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974354960845326e-05,
+      "loss": 1.1532,
+      "step": 1980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997429389107239e-05,
+      "loss": 1.2356,
+      "step": 1981
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997423274876528e-05,
+      "loss": 1.2726,
+      "step": 1982
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974171533924443e-05,
+      "loss": 1.137,
+      "step": 1983
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974110246550325e-05,
+      "loss": 0.9809,
+      "step": 1984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974048886643366e-05,
+      "loss": 0.885,
+      "step": 1985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973987454204016e-05,
+      "loss": 1.1059,
+      "step": 1986
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973925949232728e-05,
+      "loss": 1.0587,
+      "step": 1987
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973864371729936e-05,
+      "loss": 1.0841,
+      "step": 1988
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973802721696097e-05,
+      "loss": 1.0835,
+      "step": 1989
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973740999131656e-05,
+      "loss": 1.3183,
+      "step": 1990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973679204037065e-05,
+      "loss": 1.018,
+      "step": 1991
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997361733641277e-05,
+      "loss": 1.2569,
+      "step": 1992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973555396259223e-05,
+      "loss": 1.3514,
+      "step": 1993
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973493383576873e-05,
+      "loss": 1.1969,
+      "step": 1994
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997343129836617e-05,
+      "loss": 1.0256,
+      "step": 1995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973369140627567e-05,
+      "loss": 1.1066,
+      "step": 1996
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997330691036152e-05,
+      "loss": 0.9922,
+      "step": 1997
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973244607568474e-05,
+      "loss": 1.1481,
+      "step": 1998
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973182232248887e-05,
+      "loss": 1.075,
+      "step": 1999
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997311978440321e-05,
+      "loss": 1.1863,
+      "step": 2000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973057264031902e-05,
+      "loss": 1.099,
+      "step": 2001
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997299467113541e-05,
+      "loss": 0.9986,
+      "step": 2002
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972932005714197e-05,
+      "loss": 1.0858,
+      "step": 2003
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972869267768712e-05,
+      "loss": 0.9879,
+      "step": 2004
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972806457299413e-05,
+      "loss": 1.2911,
+      "step": 2005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972743574306763e-05,
+      "loss": 1.3815,
+      "step": 2006
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997268061879121e-05,
+      "loss": 1.0251,
+      "step": 2007
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972617590753218e-05,
+      "loss": 1.0998,
+      "step": 2008
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972554490193244e-05,
+      "loss": 1.2237,
+      "step": 2009
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972491317111745e-05,
+      "loss": 0.9611,
+      "step": 2010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972428071509183e-05,
+      "loss": 1.1601,
+      "step": 2011
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972364753386015e-05,
+      "loss": 1.029,
+      "step": 2012
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972301362742705e-05,
+      "loss": 1.1336,
+      "step": 2013
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997223789957971e-05,
+      "loss": 0.8938,
+      "step": 2014
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972174363897494e-05,
+      "loss": 0.9191,
+      "step": 2015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997211075569652e-05,
+      "loss": 0.9853,
+      "step": 2016
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997204707497725e-05,
+      "loss": 0.8143,
+      "step": 2017
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971983321740144e-05,
+      "loss": 1.2849,
+      "step": 2018
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971919495985665e-05,
+      "loss": 0.851,
+      "step": 2019
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971855597714286e-05,
+      "loss": 1.3486,
+      "step": 2020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997179162692646e-05,
+      "loss": 1.2412,
+      "step": 2021
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971727583622663e-05,
+      "loss": 0.9234,
+      "step": 2022
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971663467803353e-05,
+      "loss": 1.347,
+      "step": 2023
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971599279469e-05,
+      "loss": 1.1963,
+      "step": 2024
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997153501862007e-05,
+      "loss": 1.0818,
+      "step": 2025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971470685257032e-05,
+      "loss": 1.3302,
+      "step": 2026
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997140627938035e-05,
+      "loss": 1.1883,
+      "step": 2027
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971341800990495e-05,
+      "loss": 1.0444,
+      "step": 2028
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997127725008793e-05,
+      "loss": 1.1933,
+      "step": 2029
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971212626673134e-05,
+      "loss": 0.8221,
+      "step": 2030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971147930746577e-05,
+      "loss": 1.1868,
+      "step": 2031
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971083162308717e-05,
+      "loss": 0.9796,
+      "step": 2032
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971018321360036e-05,
+      "loss": 1.4202,
+      "step": 2033
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970953407901e-05,
+      "loss": 1.0156,
+      "step": 2034
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970888421932088e-05,
+      "loss": 0.9811,
+      "step": 2035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997082336345377e-05,
+      "loss": 1.2224,
+      "step": 2036
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970758232466512e-05,
+      "loss": 1.0004,
+      "step": 2037
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970693028970793e-05,
+      "loss": 1.1237,
+      "step": 2038
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970627752967088e-05,
+      "loss": 0.9812,
+      "step": 2039
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970562404455872e-05,
+      "loss": 0.8415,
+      "step": 2040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970496983437616e-05,
+      "loss": 1.1413,
+      "step": 2041
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970431489912804e-05,
+      "loss": 1.1275,
+      "step": 2042
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970365923881904e-05,
+      "loss": 1.0461,
+      "step": 2043
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970300285345394e-05,
+      "loss": 1.0685,
+      "step": 2044
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970234574303755e-05,
+      "loss": 0.9128,
+      "step": 2045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970168790757464e-05,
+      "loss": 0.9306,
+      "step": 2046
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970102934707e-05,
+      "loss": 0.9458,
+      "step": 2047
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970037006152835e-05,
+      "loss": 1.1166,
+      "step": 2048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969971005095456e-05,
+      "loss": 1.3631,
+      "step": 2049
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996990493153534e-05,
+      "loss": 1.1339,
+      "step": 2050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996983878547297e-05,
+      "loss": 1.0065,
+      "step": 2051
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969772566908824e-05,
+      "loss": 1.1647,
+      "step": 2052
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969706275843387e-05,
+      "loss": 1.0039,
+      "step": 2053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969639912277137e-05,
+      "loss": 1.0823,
+      "step": 2054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969573476210558e-05,
+      "loss": 1.3034,
+      "step": 2055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969506967644136e-05,
+      "loss": 1.2342,
+      "step": 2056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969440386578348e-05,
+      "loss": 1.0366,
+      "step": 2057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969373733013685e-05,
+      "loss": 1.0954,
+      "step": 2058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969307006950628e-05,
+      "loss": 1.0819,
+      "step": 2059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969240208389666e-05,
+      "loss": 1.0407,
+      "step": 2060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969173337331283e-05,
+      "loss": 1.1885,
+      "step": 2061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969106393775962e-05,
+      "loss": 1.0665,
+      "step": 2062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996903937772419e-05,
+      "loss": 1.4073,
+      "step": 2063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996897228917646e-05,
+      "loss": 1.1199,
+      "step": 2064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968905128133254e-05,
+      "loss": 1.1115,
+      "step": 2065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968837894595066e-05,
+      "loss": 1.139,
+      "step": 2066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968770588562377e-05,
+      "loss": 0.8682,
+      "step": 2067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968703210035686e-05,
+      "loss": 0.9009,
+      "step": 2068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968635759015476e-05,
+      "loss": 1.1176,
+      "step": 2069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996856823550224e-05,
+      "loss": 1.283,
+      "step": 2070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968500639496472e-05,
+      "loss": 1.0194,
+      "step": 2071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968432970998656e-05,
+      "loss": 1.0148,
+      "step": 2072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996836523000929e-05,
+      "loss": 1.0449,
+      "step": 2073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968297416528867e-05,
+      "loss": 1.2101,
+      "step": 2074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968229530557876e-05,
+      "loss": 1.2321,
+      "step": 2075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968161572096813e-05,
+      "loss": 1.2638,
+      "step": 2076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968093541146172e-05,
+      "loss": 1.2233,
+      "step": 2077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968025437706448e-05,
+      "loss": 1.1655,
+      "step": 2078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967957261778136e-05,
+      "loss": 1.1668,
+      "step": 2079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967889013361734e-05,
+      "loss": 1.1391,
+      "step": 2080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967820692457732e-05,
+      "loss": 1.049,
+      "step": 2081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967752299066633e-05,
+      "loss": 1.2199,
+      "step": 2082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967683833188934e-05,
+      "loss": 1.0521,
+      "step": 2083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967615294825128e-05,
+      "loss": 1.2596,
+      "step": 2084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996754668397572e-05,
+      "loss": 1.0972,
+      "step": 2085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967478000641205e-05,
+      "loss": 1.1654,
+      "step": 2086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967409244822082e-05,
+      "loss": 0.9117,
+      "step": 2087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967340416518854e-05,
+      "loss": 1.1061,
+      "step": 2088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967271515732017e-05,
+      "loss": 1.1421,
+      "step": 2089
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967202542462078e-05,
+      "loss": 1.0838,
+      "step": 2090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967133496709533e-05,
+      "loss": 1.0503,
+      "step": 2091
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967064378474886e-05,
+      "loss": 1.1194,
+      "step": 2092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996699518775864e-05,
+      "loss": 1.0629,
+      "step": 2093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99669259245613e-05,
+      "loss": 0.8932,
+      "step": 2094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966856588883366e-05,
+      "loss": 1.2488,
+      "step": 2095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966787180725343e-05,
+      "loss": 0.9924,
+      "step": 2096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996671770008774e-05,
+      "loss": 1.2957,
+      "step": 2097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996664814697106e-05,
+      "loss": 1.415,
+      "step": 2098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966578521375802e-05,
+      "loss": 1.0796,
+      "step": 2099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966508823302484e-05,
+      "loss": 0.9715,
+      "step": 2100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966439052751605e-05,
+      "loss": 1.0795,
+      "step": 2101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966369209723675e-05,
+      "loss": 1.0714,
+      "step": 2102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99662992942192e-05,
+      "loss": 1.1045,
+      "step": 2103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966229306238692e-05,
+      "loss": 1.1605,
+      "step": 2104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966159245782656e-05,
+      "loss": 1.0385,
+      "step": 2105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99660891128516e-05,
+      "loss": 1.1942,
+      "step": 2106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996601890744604e-05,
+      "loss": 0.9127,
+      "step": 2107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996594862956649e-05,
+      "loss": 0.9642,
+      "step": 2108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965878279213448e-05,
+      "loss": 1.1803,
+      "step": 2109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965807856387434e-05,
+      "loss": 1.2446,
+      "step": 2110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996573736108896e-05,
+      "loss": 1.1545,
+      "step": 2111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965666793318534e-05,
+      "loss": 1.3081,
+      "step": 2112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965596153076675e-05,
+      "loss": 1.2487,
+      "step": 2113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965525440363893e-05,
+      "loss": 0.9642,
+      "step": 2114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965454655180704e-05,
+      "loss": 1.0319,
+      "step": 2115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965383797527623e-05,
+      "loss": 1.2032,
+      "step": 2116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965312867405163e-05,
+      "loss": 0.8193,
+      "step": 2117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996524186481384e-05,
+      "loss": 1.0835,
+      "step": 2118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965170789754175e-05,
+      "loss": 1.2857,
+      "step": 2119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965099642226676e-05,
+      "loss": 1.2737,
+      "step": 2120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965028422231873e-05,
+      "loss": 1.1639,
+      "step": 2121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964957129770274e-05,
+      "loss": 0.902,
+      "step": 2122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964885764842398e-05,
+      "loss": 1.3435,
+      "step": 2123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964814327448766e-05,
+      "loss": 1.0797,
+      "step": 2124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964742817589896e-05,
+      "loss": 1.4594,
+      "step": 2125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964671235266314e-05,
+      "loss": 1.1851,
+      "step": 2126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964599580478535e-05,
+      "loss": 0.9256,
+      "step": 2127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996452785322708e-05,
+      "loss": 0.8864,
+      "step": 2128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964456053512477e-05,
+      "loss": 1.1442,
+      "step": 2129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964384181335237e-05,
+      "loss": 1.0748,
+      "step": 2130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996431223669589e-05,
+      "loss": 1.1768,
+      "step": 2131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996424021959496e-05,
+      "loss": 1.2071,
+      "step": 2132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964168130032968e-05,
+      "loss": 0.9081,
+      "step": 2133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996409596801044e-05,
+      "loss": 1.0679,
+      "step": 2134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964023733527897e-05,
+      "loss": 1.1394,
+      "step": 2135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963951426585867e-05,
+      "loss": 1.0084,
+      "step": 2136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963879047184875e-05,
+      "loss": 0.9835,
+      "step": 2137
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963806595325452e-05,
+      "loss": 1.224,
+      "step": 2138
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963734071008117e-05,
+      "loss": 1.2365,
+      "step": 2139
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99636614742334e-05,
+      "loss": 0.8986,
+      "step": 2140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963588805001834e-05,
+      "loss": 1.2039,
+      "step": 2141
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963516063313944e-05,
+      "loss": 0.9678,
+      "step": 2142
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996344324917026e-05,
+      "loss": 0.9319,
+      "step": 2143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996337036257131e-05,
+      "loss": 1.2557,
+      "step": 2144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996329740351762e-05,
+      "loss": 1.0267,
+      "step": 2145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963224372009726e-05,
+      "loss": 0.9839,
+      "step": 2146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996315126804816e-05,
+      "loss": 1.1504,
+      "step": 2147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963078091633452e-05,
+      "loss": 1.2437,
+      "step": 2148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996300484276613e-05,
+      "loss": 1.402,
+      "step": 2149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962931521446736e-05,
+      "loss": 0.668,
+      "step": 2150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962858127675794e-05,
+      "loss": 0.9263,
+      "step": 2151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996278466145384e-05,
+      "loss": 1.2386,
+      "step": 2152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996271112278141e-05,
+      "loss": 1.3117,
+      "step": 2153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962637511659043e-05,
+      "loss": 1.1528,
+      "step": 2154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962563828087266e-05,
+      "loss": 1.0702,
+      "step": 2155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996249007206662e-05,
+      "loss": 1.1621,
+      "step": 2156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996241624359764e-05,
+      "loss": 1.1169,
+      "step": 2157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962342342680862e-05,
+      "loss": 1.2399,
+      "step": 2158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962268369316824e-05,
+      "loss": 1.1273,
+      "step": 2159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962194323506064e-05,
+      "loss": 1.3943,
+      "step": 2160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962120205249122e-05,
+      "loss": 1.1093,
+      "step": 2161
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962046014546533e-05,
+      "loss": 1.1124,
+      "step": 2162
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961971751398842e-05,
+      "loss": 1.2511,
+      "step": 2163
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996189741580658e-05,
+      "loss": 1.0712,
+      "step": 2164
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961823007770303e-05,
+      "loss": 1.1902,
+      "step": 2165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961748527290536e-05,
+      "loss": 1.1182,
+      "step": 2166
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996167397436783e-05,
+      "loss": 1.1025,
+      "step": 2167
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961599349002724e-05,
+      "loss": 1.1524,
+      "step": 2168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961524651195763e-05,
+      "loss": 1.0006,
+      "step": 2169
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961449880947485e-05,
+      "loss": 1.1783,
+      "step": 2170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961375038258442e-05,
+      "loss": 1.1119,
+      "step": 2171
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996130012312917e-05,
+      "loss": 1.201,
+      "step": 2172
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996122513556022e-05,
+      "loss": 1.3081,
+      "step": 2173
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996115007555213e-05,
+      "loss": 1.0971,
+      "step": 2174
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961074943105457e-05,
+      "loss": 1.238,
+      "step": 2175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960999738220738e-05,
+      "loss": 1.317,
+      "step": 2176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996092446089852e-05,
+      "loss": 1.0002,
+      "step": 2177
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960849111139357e-05,
+      "loss": 1.2612,
+      "step": 2178
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996077368894379e-05,
+      "loss": 0.9765,
+      "step": 2179
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996069819431237e-05,
+      "loss": 1.2268,
+      "step": 2180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996062262724565e-05,
+      "loss": 1.1676,
+      "step": 2181
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960546987744177e-05,
+      "loss": 0.8076,
+      "step": 2182
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960471275808497e-05,
+      "loss": 1.191,
+      "step": 2183
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960395491439164e-05,
+      "loss": 0.8865,
+      "step": 2184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996031963463673e-05,
+      "loss": 0.9206,
+      "step": 2185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960243705401744e-05,
+      "loss": 1.162,
+      "step": 2186
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996016770373476e-05,
+      "loss": 1.206,
+      "step": 2187
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960091629636335e-05,
+      "loss": 1.2747,
+      "step": 2188
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960015483107017e-05,
+      "loss": 0.9217,
+      "step": 2189
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959939264147355e-05,
+      "loss": 0.9884,
+      "step": 2190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995986297275791e-05,
+      "loss": 0.8499,
+      "step": 2191
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959786608939236e-05,
+      "loss": 1.0147,
+      "step": 2192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995971017269189e-05,
+      "loss": 1.1348,
+      "step": 2193
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959633664016424e-05,
+      "loss": 1.2241,
+      "step": 2194
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959557082913396e-05,
+      "loss": 1.135,
+      "step": 2195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959480429383366e-05,
+      "loss": 1.1995,
+      "step": 2196
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959403703426882e-05,
+      "loss": 1.0654,
+      "step": 2197
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995932690504451e-05,
+      "loss": 0.8657,
+      "step": 2198
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995925003423681e-05,
+      "loss": 1.1943,
+      "step": 2199
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959173091004335e-05,
+      "loss": 1.0653,
+      "step": 2200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995909607534765e-05,
+      "loss": 0.9953,
+      "step": 2201
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959018987267308e-05,
+      "loss": 0.9773,
+      "step": 2202
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958941826763878e-05,
+      "loss": 1.1006,
+      "step": 2203
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958864593837917e-05,
+      "loss": 1.4119,
+      "step": 2204
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958787288489983e-05,
+      "loss": 1.2876,
+      "step": 2205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958709910720647e-05,
+      "loss": 1.1414,
+      "step": 2206
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958632460530465e-05,
+      "loss": 1.4198,
+      "step": 2207
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995855493792e-05,
+      "loss": 1.16,
+      "step": 2208
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958477342889822e-05,
+      "loss": 0.825,
+      "step": 2209
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958399675440487e-05,
+      "loss": 1.067,
+      "step": 2210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958321935572565e-05,
+      "loss": 0.9372,
+      "step": 2211
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995824412328662e-05,
+      "loss": 1.2953,
+      "step": 2212
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958166238583216e-05,
+      "loss": 1.2367,
+      "step": 2213
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958088281462924e-05,
+      "loss": 1.2964,
+      "step": 2214
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958010251926306e-05,
+      "loss": 1.2518,
+      "step": 2215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957932149973934e-05,
+      "loss": 1.4668,
+      "step": 2216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995785397560637e-05,
+      "loss": 1.1474,
+      "step": 2217
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995777572882419e-05,
+      "loss": 0.9563,
+      "step": 2218
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957697409627954e-05,
+      "loss": 1.0529,
+      "step": 2219
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957619018018243e-05,
+      "loss": 1.077,
+      "step": 2220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957540553995617e-05,
+      "loss": 1.4693,
+      "step": 2221
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995746201756065e-05,
+      "loss": 1.2695,
+      "step": 2222
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957383408713915e-05,
+      "loss": 0.9619,
+      "step": 2223
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957304727455982e-05,
+      "loss": 1.0539,
+      "step": 2224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957225973787424e-05,
+      "loss": 1.3842,
+      "step": 2225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995714714770881e-05,
+      "loss": 1.1771,
+      "step": 2226
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995706824922072e-05,
+      "loss": 1.0219,
+      "step": 2227
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995698927832372e-05,
+      "loss": 1.1834,
+      "step": 2228
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995691023501839e-05,
+      "loss": 1.1964,
+      "step": 2229
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956831119305305e-05,
+      "loss": 1.5054,
+      "step": 2230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956751931185038e-05,
+      "loss": 0.9796,
+      "step": 2231
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956672670658165e-05,
+      "loss": 1.2147,
+      "step": 2232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956593337725262e-05,
+      "loss": 1.2981,
+      "step": 2233
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956513932386902e-05,
+      "loss": 1.0545,
+      "step": 2234
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956434454643675e-05,
+      "loss": 1.1395,
+      "step": 2235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956354904496145e-05,
+      "loss": 1.386,
+      "step": 2236
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99562752819449e-05,
+      "loss": 1.0395,
+      "step": 2237
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956195586990516e-05,
+      "loss": 1.1615,
+      "step": 2238
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995611581963357e-05,
+      "loss": 0.9791,
+      "step": 2239
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956035979874645e-05,
+      "loss": 1.0892,
+      "step": 2240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995595606771432e-05,
+      "loss": 1.1024,
+      "step": 2241
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955876083153178e-05,
+      "loss": 1.0786,
+      "step": 2242
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99557960261918e-05,
+      "loss": 1.0617,
+      "step": 2243
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955715896830767e-05,
+      "loss": 1.3617,
+      "step": 2244
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955635695070657e-05,
+      "loss": 0.9611,
+      "step": 2245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955555420912067e-05,
+      "loss": 1.081,
+      "step": 2246
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995547507435557e-05,
+      "loss": 1.378,
+      "step": 2247
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995539465540175e-05,
+      "loss": 1.1478,
+      "step": 2248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955314164051197e-05,
+      "loss": 1.1217,
+      "step": 2249
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955233600304496e-05,
+      "loss": 1.101,
+      "step": 2250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955152964162225e-05,
+      "loss": 1.072,
+      "step": 2251
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995507225562498e-05,
+      "loss": 1.1626,
+      "step": 2252
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954991474693344e-05,
+      "loss": 1.0331,
+      "step": 2253
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954910621367905e-05,
+      "loss": 1.0439,
+      "step": 2254
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995482969564925e-05,
+      "loss": 1.0827,
+      "step": 2255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954748697537963e-05,
+      "loss": 0.7448,
+      "step": 2256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954667627034645e-05,
+      "loss": 1.0807,
+      "step": 2257
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954586484139877e-05,
+      "loss": 0.9943,
+      "step": 2258
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954505268854248e-05,
+      "loss": 1.0526,
+      "step": 2259
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954423981178354e-05,
+      "loss": 0.9462,
+      "step": 2260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995434262111278e-05,
+      "loss": 1.0457,
+      "step": 2261
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954261188658127e-05,
+      "loss": 0.9944,
+      "step": 2262
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954179683814976e-05,
+      "loss": 1.16,
+      "step": 2263
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954098106583925e-05,
+      "loss": 1.1761,
+      "step": 2264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995401645696557e-05,
+      "loss": 1.3073,
+      "step": 2265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953934734960502e-05,
+      "loss": 1.0052,
+      "step": 2266
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953852940569312e-05,
+      "loss": 1.2086,
+      "step": 2267
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99537710737926e-05,
+      "loss": 1.2225,
+      "step": 2268
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995368913463096e-05,
+      "loss": 1.2599,
+      "step": 2269
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953607123084987e-05,
+      "loss": 1.0699,
+      "step": 2270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953525039155277e-05,
+      "loss": 1.1175,
+      "step": 2271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995344288284243e-05,
+      "loss": 0.843,
+      "step": 2272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995336065414704e-05,
+      "loss": 1.0299,
+      "step": 2273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953278353069706e-05,
+      "loss": 0.8635,
+      "step": 2274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953195979611026e-05,
+      "loss": 0.8899,
+      "step": 2275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953113533771597e-05,
+      "loss": 0.8842,
+      "step": 2276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953031015552025e-05,
+      "loss": 0.9274,
+      "step": 2277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952948424952904e-05,
+      "loss": 0.9013,
+      "step": 2278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952865761974835e-05,
+      "loss": 1.0124,
+      "step": 2279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952783026618424e-05,
+      "loss": 1.1031,
+      "step": 2280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995270021888427e-05,
+      "loss": 1.0742,
+      "step": 2281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995261733877297e-05,
+      "loss": 1.0563,
+      "step": 2282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995253438628514e-05,
+      "loss": 0.9834,
+      "step": 2283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952451361421368e-05,
+      "loss": 1.1652,
+      "step": 2284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952368264182266e-05,
+      "loss": 1.1614,
+      "step": 2285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952285094568438e-05,
+      "loss": 1.1119,
+      "step": 2286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952201852580486e-05,
+      "loss": 0.9625,
+      "step": 2287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995211853821902e-05,
+      "loss": 1.047,
+      "step": 2288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995203515148464e-05,
+      "loss": 0.6852,
+      "step": 2289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951951692377952e-05,
+      "loss": 1.0404,
+      "step": 2290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951868160899574e-05,
+      "loss": 0.9522,
+      "step": 2291
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99517845570501e-05,
+      "loss": 0.9916,
+      "step": 2292
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951700880830145e-05,
+      "loss": 1.4025,
+      "step": 2293
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951617132240316e-05,
+      "loss": 0.9685,
+      "step": 2294
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995153331128122e-05,
+      "loss": 1.222,
+      "step": 2295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951449417953474e-05,
+      "loss": 0.9751,
+      "step": 2296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995136545225768e-05,
+      "loss": 1.2749,
+      "step": 2297
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995128141419445e-05,
+      "loss": 1.299,
+      "step": 2298
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951197303764397e-05,
+      "loss": 1.0532,
+      "step": 2299
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951113120968135e-05,
+      "loss": 0.8415,
+      "step": 2300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995102886580627e-05,
+      "loss": 0.8772,
+      "step": 2301
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995094453827942e-05,
+      "loss": 1.2047,
+      "step": 2302
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950860138388195e-05,
+      "loss": 1.3104,
+      "step": 2303
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950775666133212e-05,
+      "loss": 1.0394,
+      "step": 2304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950691121515083e-05,
+      "loss": 1.2277,
+      "step": 2305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995060650453442e-05,
+      "loss": 1.0778,
+      "step": 2306
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950521815191843e-05,
+      "loss": 0.599,
+      "step": 2307
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950437053487968e-05,
+      "loss": 1.0221,
+      "step": 2308
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950352219423407e-05,
+      "loss": 1.0459,
+      "step": 2309
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950267312998783e-05,
+      "loss": 1.1392,
+      "step": 2310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950182334214706e-05,
+      "loss": 1.0475,
+      "step": 2311
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950097283071803e-05,
+      "loss": 1.2673,
+      "step": 2312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950012159570687e-05,
+      "loss": 1.2043,
+      "step": 2313
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949926963711974e-05,
+      "loss": 1.0667,
+      "step": 2314
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994984169549629e-05,
+      "loss": 1.3207,
+      "step": 2315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949756354924253e-05,
+      "loss": 0.8189,
+      "step": 2316
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994967094199648e-05,
+      "loss": 0.9687,
+      "step": 2317
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949585456713597e-05,
+      "loss": 0.9307,
+      "step": 2318
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949499899076224e-05,
+      "loss": 1.1052,
+      "step": 2319
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994941426908498e-05,
+      "loss": 0.7889,
+      "step": 2320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994932856674049e-05,
+      "loss": 1.183,
+      "step": 2321
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949242792043382e-05,
+      "loss": 1.2919,
+      "step": 2322
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949156944994274e-05,
+      "loss": 1.1617,
+      "step": 2323
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949071025593794e-05,
+      "loss": 0.9399,
+      "step": 2324
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994898503384256e-05,
+      "loss": 1.0718,
+      "step": 2325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948898969741207e-05,
+      "loss": 1.2944,
+      "step": 2326
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948812833290353e-05,
+      "loss": 1.066,
+      "step": 2327
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948726624490628e-05,
+      "loss": 0.7617,
+      "step": 2328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948640343342657e-05,
+      "loss": 1.204,
+      "step": 2329
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994855398984707e-05,
+      "loss": 0.9387,
+      "step": 2330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994846756400449e-05,
+      "loss": 1.0234,
+      "step": 2331
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948381065815553e-05,
+      "loss": 1.1644,
+      "step": 2332
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948294495280883e-05,
+      "loss": 1.0447,
+      "step": 2333
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948207852401112e-05,
+      "loss": 0.9153,
+      "step": 2334
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994812113717687e-05,
+      "loss": 1.2258,
+      "step": 2335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948034349608782e-05,
+      "loss": 0.9893,
+      "step": 2336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947947489697487e-05,
+      "loss": 1.292,
+      "step": 2337
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947860557443614e-05,
+      "loss": 1.0671,
+      "step": 2338
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994777355284779e-05,
+      "loss": 1.5249,
+      "step": 2339
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947686475910656e-05,
+      "loss": 1.2635,
+      "step": 2340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947599326632837e-05,
+      "loss": 1.2056,
+      "step": 2341
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947512105014973e-05,
+      "loss": 1.1664,
+      "step": 2342
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99474248110577e-05,
+      "loss": 1.0192,
+      "step": 2343
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947337444761645e-05,
+      "loss": 1.2159,
+      "step": 2344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994725000612745e-05,
+      "loss": 0.9244,
+      "step": 2345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947162495155746e-05,
+      "loss": 1.2159,
+      "step": 2346
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994707491184717e-05,
+      "loss": 1.1281,
+      "step": 2347
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946987256202366e-05,
+      "loss": 1.2305,
+      "step": 2348
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994689952822196e-05,
+      "loss": 1.2013,
+      "step": 2349
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99468117279066e-05,
+      "loss": 1.1846,
+      "step": 2350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946723855256922e-05,
+      "loss": 1.1853,
+      "step": 2351
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994663591027356e-05,
+      "loss": 1.0429,
+      "step": 2352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994654789295716e-05,
+      "loss": 1.0805,
+      "step": 2353
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994645980330836e-05,
+      "loss": 1.2127,
+      "step": 2354
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946371641327794e-05,
+      "loss": 1.1015,
+      "step": 2355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946283407016115e-05,
+      "loss": 1.0045,
+      "step": 2356
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994619510037396e-05,
+      "loss": 1.0206,
+      "step": 2357
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946106721401964e-05,
+      "loss": 1.1413,
+      "step": 2358
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946018270100776e-05,
+      "loss": 1.0968,
+      "step": 2359
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994592974647104e-05,
+      "loss": 1.1465,
+      "step": 2360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945841150513398e-05,
+      "loss": 1.1959,
+      "step": 2361
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945752482228495e-05,
+      "loss": 1.1047,
+      "step": 2362
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945663741616976e-05,
+      "loss": 1.2367,
+      "step": 2363
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945574928679483e-05,
+      "loss": 1.3177,
+      "step": 2364
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994548604341667e-05,
+      "loss": 1.1396,
+      "step": 2365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945397085829173e-05,
+      "loss": 1.0152,
+      "step": 2366
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945308055917648e-05,
+      "loss": 1.1431,
+      "step": 2367
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945218953682736e-05,
+      "loss": 0.8569,
+      "step": 2368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994512977912509e-05,
+      "loss": 1.1356,
+      "step": 2369
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945040532245352e-05,
+      "loss": 1.0803,
+      "step": 2370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944951213044174e-05,
+      "loss": 0.884,
+      "step": 2371
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994486182152221e-05,
+      "loss": 1.1752,
+      "step": 2372
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944772357680105e-05,
+      "loss": 1.1204,
+      "step": 2373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944682821518512e-05,
+      "loss": 1.3136,
+      "step": 2374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944593213038083e-05,
+      "loss": 1.0109,
+      "step": 2375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944503532239465e-05,
+      "loss": 1.1322,
+      "step": 2376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994441377912331e-05,
+      "loss": 1.1592,
+      "step": 2377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994432395369028e-05,
+      "loss": 1.1796,
+      "step": 2378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944234055941018e-05,
+      "loss": 0.9489,
+      "step": 2379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944144085876186e-05,
+      "loss": 1.2844,
+      "step": 2380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944054043496428e-05,
+      "loss": 0.995,
+      "step": 2381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994396392880241e-05,
+      "loss": 1.3441,
+      "step": 2382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994387374179478e-05,
+      "loss": 1.2399,
+      "step": 2383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943783482474192e-05,
+      "loss": 0.9138,
+      "step": 2384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943693150841312e-05,
+      "loss": 1.0443,
+      "step": 2385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994360274689679e-05,
+      "loss": 1.3732,
+      "step": 2386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943512270641283e-05,
+      "loss": 1.0724,
+      "step": 2387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943421722075452e-05,
+      "loss": 1.0249,
+      "step": 2388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943331101199953e-05,
+      "loss": 1.1766,
+      "step": 2389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943240408015444e-05,
+      "loss": 1.0459,
+      "step": 2390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994314964252259e-05,
+      "loss": 1.1524,
+      "step": 2391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994305880472205e-05,
+      "loss": 1.157,
+      "step": 2392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942967894614476e-05,
+      "loss": 1.2524,
+      "step": 2393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942876912200538e-05,
+      "loss": 1.257,
+      "step": 2394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942785857480896e-05,
+      "loss": 1.2218,
+      "step": 2395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942694730456208e-05,
+      "loss": 1.3312,
+      "step": 2396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942603531127144e-05,
+      "loss": 0.9545,
+      "step": 2397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994251225949436e-05,
+      "loss": 1.1189,
+      "step": 2398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994242091555852e-05,
+      "loss": 0.9848,
+      "step": 2399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942329499320298e-05,
+      "loss": 1.0813,
+      "step": 2400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942238010780352e-05,
+      "loss": 1.0974,
+      "step": 2401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994214644993934e-05,
+      "loss": 1.4066,
+      "step": 2402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994205481679794e-05,
+      "loss": 1.1245,
+      "step": 2403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994196311135681e-05,
+      "loss": 1.1588,
+      "step": 2404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941871333616625e-05,
+      "loss": 1.0257,
+      "step": 2405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941779483578047e-05,
+      "loss": 1.1799,
+      "step": 2406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941687561241737e-05,
+      "loss": 0.886,
+      "step": 2407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941595566608377e-05,
+      "loss": 0.888,
+      "step": 2408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994150349967863e-05,
+      "loss": 1.3842,
+      "step": 2409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941411360453165e-05,
+      "loss": 0.9542,
+      "step": 2410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941319148932655e-05,
+      "loss": 0.99,
+      "step": 2411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941226865117764e-05,
+      "loss": 1.0552,
+      "step": 2412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941134509009173e-05,
+      "loss": 1.0831,
+      "step": 2413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994104208060754e-05,
+      "loss": 1.2106,
+      "step": 2414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994094957991355e-05,
+      "loss": 0.6975,
+      "step": 2415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940857006927872e-05,
+      "loss": 1.1558,
+      "step": 2416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940764361651175e-05,
+      "loss": 1.2507,
+      "step": 2417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940671644084135e-05,
+      "loss": 1.1536,
+      "step": 2418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994057885422743e-05,
+      "loss": 0.9487,
+      "step": 2419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994048599208173e-05,
+      "loss": 1.0048,
+      "step": 2420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940393057647714e-05,
+      "loss": 0.9533,
+      "step": 2421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940300050926053e-05,
+      "loss": 1.1363,
+      "step": 2422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940206971917427e-05,
+      "loss": 0.89,
+      "step": 2423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940113820622512e-05,
+      "loss": 1.2319,
+      "step": 2424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940020597041988e-05,
+      "loss": 1.0711,
+      "step": 2425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939927301176526e-05,
+      "loss": 1.161,
+      "step": 2426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993983393302681e-05,
+      "loss": 1.2717,
+      "step": 2427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993974049259352e-05,
+      "loss": 1.0457,
+      "step": 2428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939646979877333e-05,
+      "loss": 0.8157,
+      "step": 2429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939553394878926e-05,
+      "loss": 0.9802,
+      "step": 2430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939459737598986e-05,
+      "loss": 1.1657,
+      "step": 2431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993936600803819e-05,
+      "loss": 1.2495,
+      "step": 2432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939272206197226e-05,
+      "loss": 0.9763,
+      "step": 2433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939178332076767e-05,
+      "loss": 1.1543,
+      "step": 2434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939084385677496e-05,
+      "loss": 1.104,
+      "step": 2435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99389903670001e-05,
+      "loss": 1.1318,
+      "step": 2436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938896276045265e-05,
+      "loss": 0.8552,
+      "step": 2437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993880211281367e-05,
+      "loss": 1.3193,
+      "step": 2438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993870787730601e-05,
+      "loss": 1.3478,
+      "step": 2439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938613569522953e-05,
+      "loss": 1.2522,
+      "step": 2440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99385191894652e-05,
+      "loss": 1.0587,
+      "step": 2441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938424737133432e-05,
+      "loss": 1.4498,
+      "step": 2442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938330212528333e-05,
+      "loss": 1.05,
+      "step": 2443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938235615650594e-05,
+      "loss": 0.8806,
+      "step": 2444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99381409465009e-05,
+      "loss": 1.1702,
+      "step": 2445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938046205079945e-05,
+      "loss": 1.0035,
+      "step": 2446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937951391388413e-05,
+      "loss": 1.3707,
+      "step": 2447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937856505426993e-05,
+      "loss": 1.0716,
+      "step": 2448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993776154719638e-05,
+      "loss": 1.2157,
+      "step": 2449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993766651669726e-05,
+      "loss": 1.0763,
+      "step": 2450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937571413930325e-05,
+      "loss": 0.8729,
+      "step": 2451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993747623889627e-05,
+      "loss": 1.1738,
+      "step": 2452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993738099159578e-05,
+      "loss": 1.1698,
+      "step": 2453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937285672029556e-05,
+      "loss": 0.9411,
+      "step": 2454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937190280198284e-05,
+      "loss": 1.0407,
+      "step": 2455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993709481610266e-05,
+      "loss": 1.0267,
+      "step": 2456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936999279743384e-05,
+      "loss": 1.1737,
+      "step": 2457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936903671121142e-05,
+      "loss": 1.3818,
+      "step": 2458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936807990236635e-05,
+      "loss": 1.0708,
+      "step": 2459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936712237090554e-05,
+      "loss": 0.8319,
+      "step": 2460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99366164116836e-05,
+      "loss": 1.1298,
+      "step": 2461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936520514016465e-05,
+      "loss": 0.9793,
+      "step": 2462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936424544089855e-05,
+      "loss": 1.1969,
+      "step": 2463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993632850190446e-05,
+      "loss": 1.122,
+      "step": 2464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993623238746098e-05,
+      "loss": 0.9648,
+      "step": 2465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936136200760116e-05,
+      "loss": 1.0545,
+      "step": 2466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993603994180256e-05,
+      "loss": 1.2875,
+      "step": 2467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9935943610589027e-05,
+      "loss": 0.8343,
+      "step": 2468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935847207120205e-05,
+      "loss": 1.2269,
+      "step": 2469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99357507313968e-05,
+      "loss": 1.1445,
+      "step": 2470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993565418341951e-05,
+      "loss": 0.9594,
+      "step": 2471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935557563189037e-05,
+      "loss": 1.2007,
+      "step": 2472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935460870706095e-05,
+      "loss": 0.6891,
+      "step": 2473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993536410597137e-05,
+      "loss": 1.1079,
+      "step": 2474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935267268985577e-05,
+      "loss": 1.2089,
+      "step": 2475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935170359749417e-05,
+      "loss": 1.2309,
+      "step": 2476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935073378263594e-05,
+      "loss": 1.0874,
+      "step": 2477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934976324528814e-05,
+      "loss": 1.1698,
+      "step": 2478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934879198545784e-05,
+      "loss": 0.976,
+      "step": 2479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993478200031521e-05,
+      "loss": 1.3771,
+      "step": 2480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934684729837798e-05,
+      "loss": 0.7241,
+      "step": 2481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934587387114255e-05,
+      "loss": 1.1936,
+      "step": 2482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934489972145287e-05,
+      "loss": 1.0247,
+      "step": 2483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993439248493161e-05,
+      "loss": 1.1667,
+      "step": 2484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934294925473924e-05,
+      "loss": 0.8607,
+      "step": 2485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993419729377294e-05,
+      "loss": 1.2617,
+      "step": 2486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934099589829374e-05,
+      "loss": 0.9859,
+      "step": 2487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934001813643927e-05,
+      "loss": 0.7663,
+      "step": 2488
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933903965217324e-05,
+      "loss": 0.7174,
+      "step": 2489
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933806044550262e-05,
+      "loss": 1.2762,
+      "step": 2490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993370805164346e-05,
+      "loss": 1.465,
+      "step": 2491
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993360998649763e-05,
+      "loss": 0.755,
+      "step": 2492
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993351184911349e-05,
+      "loss": 1.2061,
+      "step": 2493
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933413639491742e-05,
+      "loss": 1.0398,
+      "step": 2494
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993331535763311e-05,
+      "loss": 1.1109,
+      "step": 2495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933217003538305e-05,
+      "loss": 1.1648,
+      "step": 2496
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933118577208047e-05,
+      "loss": 1.2891,
+      "step": 2497
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933020078643043e-05,
+      "loss": 1.1434,
+      "step": 2498
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932921507844015e-05,
+      "loss": 0.953,
+      "step": 2499
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932822864811678e-05,
+      "loss": 1.0066,
+      "step": 2500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932724149546752e-05,
+      "loss": 1.1744,
+      "step": 2501
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993262536204995e-05,
+      "loss": 1.2738,
+      "step": 2502
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932526502321997e-05,
+      "loss": 0.8741,
+      "step": 2503
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932427570363607e-05,
+      "loss": 0.9923,
+      "step": 2504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99323285661755e-05,
+      "loss": 0.9531,
+      "step": 2505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932229489758395e-05,
+      "loss": 1.2404,
+      "step": 2506
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993213034111302e-05,
+      "loss": 0.9469,
+      "step": 2507
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932031120240084e-05,
+      "loss": 1.1348,
+      "step": 2508
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931931827140317e-05,
+      "loss": 1.0716,
+      "step": 2509
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993183246181444e-05,
+      "loss": 1.1066,
+      "step": 2510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931733024263172e-05,
+      "loss": 0.8764,
+      "step": 2511
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993163351448724e-05,
+      "loss": 0.8837,
+      "step": 2512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931533932487368e-05,
+      "loss": 1.1376,
+      "step": 2513
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931434278264275e-05,
+      "loss": 1.2783,
+      "step": 2514
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931334551818692e-05,
+      "loss": 0.8279,
+      "step": 2515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993123475315134e-05,
+      "loss": 1.1728,
+      "step": 2516
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931134882262944e-05,
+      "loss": 1.0994,
+      "step": 2517
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931034939154237e-05,
+      "loss": 1.2276,
+      "step": 2518
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993093492382594e-05,
+      "loss": 1.1848,
+      "step": 2519
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993083483627878e-05,
+      "loss": 1.2189,
+      "step": 2520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930734676513484e-05,
+      "loss": 1.0972,
+      "step": 2521
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930634444530787e-05,
+      "loss": 0.9157,
+      "step": 2522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993053414033141e-05,
+      "loss": 0.8834,
+      "step": 2523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930433763916086e-05,
+      "loss": 1.1095,
+      "step": 2524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930333315285546e-05,
+      "loss": 1.1787,
+      "step": 2525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993023279444052e-05,
+      "loss": 1.1246,
+      "step": 2526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993013220138174e-05,
+      "loss": 1.1821,
+      "step": 2527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930031536109933e-05,
+      "loss": 1.1231,
+      "step": 2528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929930798625837e-05,
+      "loss": 0.9053,
+      "step": 2529
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992982998893018e-05,
+      "loss": 0.8455,
+      "step": 2530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929729107023693e-05,
+      "loss": 1.1952,
+      "step": 2531
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929628152907116e-05,
+      "loss": 1.2606,
+      "step": 2532
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929527126581183e-05,
+      "loss": 0.9924,
+      "step": 2533
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992942602804662e-05,
+      "loss": 1.1831,
+      "step": 2534
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929324857304175e-05,
+      "loss": 0.9876,
+      "step": 2535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929223614354574e-05,
+      "loss": 1.0755,
+      "step": 2536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929122299198555e-05,
+      "loss": 1.3121,
+      "step": 2537
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992902091183686e-05,
+      "loss": 1.2189,
+      "step": 2538
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992891945227022e-05,
+      "loss": 1.24,
+      "step": 2539
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928817920499375e-05,
+      "loss": 0.8944,
+      "step": 2540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992871631652506e-05,
+      "loss": 1.1177,
+      "step": 2541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928614640348024e-05,
+      "loss": 1.2484,
+      "step": 2542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928512891968994e-05,
+      "loss": 0.9694,
+      "step": 2543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992841107138872e-05,
+      "loss": 1.1125,
+      "step": 2544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928309178607937e-05,
+      "loss": 1.5507,
+      "step": 2545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928207213627387e-05,
+      "loss": 1.07,
+      "step": 2546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992810517644781e-05,
+      "loss": 0.9982,
+      "step": 2547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992800306706995e-05,
+      "loss": 0.9339,
+      "step": 2548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992790088549455e-05,
+      "loss": 0.9388,
+      "step": 2549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927798631722353e-05,
+      "loss": 0.9986,
+      "step": 2550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927696305754104e-05,
+      "loss": 0.4942,
+      "step": 2551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992759390759054e-05,
+      "loss": 1.0146,
+      "step": 2552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927491437232414e-05,
+      "loss": 1.0281,
+      "step": 2553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927388894680467e-05,
+      "loss": 1.1066,
+      "step": 2554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927286279935444e-05,
+      "loss": 0.9642,
+      "step": 2555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927183592998095e-05,
+      "loss": 1.3086,
+      "step": 2556
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927080833869167e-05,
+      "loss": 1.1985,
+      "step": 2557
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99269780025494e-05,
+      "loss": 0.9744,
+      "step": 2558
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992687509903955e-05,
+      "loss": 1.1467,
+      "step": 2559
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926772123340357e-05,
+      "loss": 1.1757,
+      "step": 2560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992666907545258e-05,
+      "loss": 1.1034,
+      "step": 2561
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992656595537696e-05,
+      "loss": 1.0499,
+      "step": 2562
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992646276311425e-05,
+      "loss": 1.1295,
+      "step": 2563
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926359498665204e-05,
+      "loss": 0.9099,
+      "step": 2564
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926256162030564e-05,
+      "loss": 1.0909,
+      "step": 2565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992615275321109e-05,
+      "loss": 1.0877,
+      "step": 2566
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926049272207532e-05,
+      "loss": 0.9292,
+      "step": 2567
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992594571902064e-05,
+      "loss": 0.8997,
+      "step": 2568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925842093651168e-05,
+      "loss": 1.1317,
+      "step": 2569
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925738396099872e-05,
+      "loss": 1.1696,
+      "step": 2570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925634626367502e-05,
+      "loss": 1.1023,
+      "step": 2571
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925530784454817e-05,
+      "loss": 0.8319,
+      "step": 2572
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925426870362567e-05,
+      "loss": 1.1994,
+      "step": 2573
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925322884091513e-05,
+      "loss": 0.9805,
+      "step": 2574
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992521882564241e-05,
+      "loss": 1.3143,
+      "step": 2575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992511469501601e-05,
+      "loss": 1.0474,
+      "step": 2576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925010492213076e-05,
+      "loss": 0.905,
+      "step": 2577
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924906217234362e-05,
+      "loss": 0.7852,
+      "step": 2578
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992480187008063e-05,
+      "loss": 0.9294,
+      "step": 2579
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924697450752636e-05,
+      "loss": 1.0492,
+      "step": 2580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924592959251135e-05,
+      "loss": 1.2682,
+      "step": 2581
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924488395576898e-05,
+      "loss": 1.0526,
+      "step": 2582
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992438375973068e-05,
+      "loss": 0.7244,
+      "step": 2583
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992427905171324e-05,
+      "loss": 1.2636,
+      "step": 2584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924174271525336e-05,
+      "loss": 1.0493,
+      "step": 2585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992406941916774e-05,
+      "loss": 1.2236,
+      "step": 2586
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923964494641206e-05,
+      "loss": 0.9804,
+      "step": 2587
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923859497946504e-05,
+      "loss": 1.0077,
+      "step": 2588
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992375442908439e-05,
+      "loss": 0.9458,
+      "step": 2589
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992364928805563e-05,
+      "loss": 1.0189,
+      "step": 2590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923544074860996e-05,
+      "loss": 1.2377,
+      "step": 2591
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923438789501243e-05,
+      "loss": 1.2866,
+      "step": 2592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923333431977147e-05,
+      "loss": 1.1078,
+      "step": 2593
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992322800228946e-05,
+      "loss": 1.1758,
+      "step": 2594
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923122500438964e-05,
+      "loss": 1.2235,
+      "step": 2595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923016926426415e-05,
+      "loss": 0.9052,
+      "step": 2596
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992291128025259e-05,
+      "loss": 0.99,
+      "step": 2597
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922805561918247e-05,
+      "loss": 1.1077,
+      "step": 2598
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922699771424163e-05,
+      "loss": 1.4899,
+      "step": 2599
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99225939087711e-05,
+      "loss": 1.059,
+      "step": 2600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922487973959837e-05,
+      "loss": 1.1147,
+      "step": 2601
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992238196699114e-05,
+      "loss": 1.1976,
+      "step": 2602
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922275887865775e-05,
+      "loss": 1.2322,
+      "step": 2603
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922169736584522e-05,
+      "loss": 0.7922,
+      "step": 2604
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922063513148147e-05,
+      "loss": 0.8712,
+      "step": 2605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921957217557427e-05,
+      "loss": 1.0858,
+      "step": 2606
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992185084981313e-05,
+      "loss": 1.0786,
+      "step": 2607
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921744409916028e-05,
+      "loss": 1.1287,
+      "step": 2608
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921637897866905e-05,
+      "loss": 1.3175,
+      "step": 2609
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921531313666526e-05,
+      "loss": 1.1341,
+      "step": 2610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992142465731567e-05,
+      "loss": 1.0589,
+      "step": 2611
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921317928815115e-05,
+      "loss": 1.1563,
+      "step": 2612
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992121112816563e-05,
+      "loss": 1.1735,
+      "step": 2613
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921104255368e-05,
+      "loss": 0.9816,
+      "step": 2614
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920997310422992e-05,
+      "loss": 0.8143,
+      "step": 2615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920890293331396e-05,
+      "loss": 1.3345,
+      "step": 2616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920783204093982e-05,
+      "loss": 0.949,
+      "step": 2617
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920676042711533e-05,
+      "loss": 1.0712,
+      "step": 2618
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920568809184824e-05,
+      "loss": 0.9627,
+      "step": 2619
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920461503514636e-05,
+      "loss": 1.0225,
+      "step": 2620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992035412570175e-05,
+      "loss": 0.8813,
+      "step": 2621
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992024667574695e-05,
+      "loss": 0.764,
+      "step": 2622
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920139153651014e-05,
+      "loss": 1.265,
+      "step": 2623
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920031559414725e-05,
+      "loss": 1.1352,
+      "step": 2624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919923893038863e-05,
+      "loss": 1.1628,
+      "step": 2625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919816154524213e-05,
+      "loss": 1.2375,
+      "step": 2626
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919708343871558e-05,
+      "loss": 1.1306,
+      "step": 2627
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991960046108168e-05,
+      "loss": 1.2212,
+      "step": 2628
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919492506155372e-05,
+      "loss": 1.0651,
+      "step": 2629
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991938447909341e-05,
+      "loss": 1.2186,
+      "step": 2630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919276379896583e-05,
+      "loss": 0.9917,
+      "step": 2631
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919168208565675e-05,
+      "loss": 1.0667,
+      "step": 2632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919059965101474e-05,
+      "loss": 1.2172,
+      "step": 2633
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991895164950477e-05,
+      "loss": 1.0144,
+      "step": 2634
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918843261776346e-05,
+      "loss": 0.9197,
+      "step": 2635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918734801916994e-05,
+      "loss": 1.1848,
+      "step": 2636
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99186262699275e-05,
+      "loss": 1.0669,
+      "step": 2637
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918517665808658e-05,
+      "loss": 1.1427,
+      "step": 2638
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918408989561247e-05,
+      "loss": 1.1289,
+      "step": 2639
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991830024118607e-05,
+      "loss": 0.8715,
+      "step": 2640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991819142068391e-05,
+      "loss": 1.1423,
+      "step": 2641
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991808252805556e-05,
+      "loss": 1.2125,
+      "step": 2642
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917973563301817e-05,
+      "loss": 1.0481,
+      "step": 2643
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991786452642346e-05,
+      "loss": 1.2126,
+      "step": 2644
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99177554174213e-05,
+      "loss": 1.0495,
+      "step": 2645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917646236296117e-05,
+      "loss": 1.0809,
+      "step": 2646
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991753698304871e-05,
+      "loss": 1.2022,
+      "step": 2647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917427657679874e-05,
+      "loss": 1.1704,
+      "step": 2648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99173182601904e-05,
+      "loss": 1.1783,
+      "step": 2649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917208790581086e-05,
+      "loss": 1.2392,
+      "step": 2650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991709924885273e-05,
+      "loss": 1.0125,
+      "step": 2651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991698963500613e-05,
+      "loss": 1.0682,
+      "step": 2652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916879949042075e-05,
+      "loss": 0.8714,
+      "step": 2653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991677019096137e-05,
+      "loss": 1.1975,
+      "step": 2654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991666036076481e-05,
+      "loss": 0.9543,
+      "step": 2655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99165504584532e-05,
+      "loss": 1.0514,
+      "step": 2656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916440484027326e-05,
+      "loss": 1.0647,
+      "step": 2657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916330437488e-05,
+      "loss": 1.1579,
+      "step": 2658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916220318836018e-05,
+      "loss": 1.2185,
+      "step": 2659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991611012807218e-05,
+      "loss": 1.2037,
+      "step": 2660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915999865197285e-05,
+      "loss": 0.8726,
+      "step": 2661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991588953021214e-05,
+      "loss": 1.0402,
+      "step": 2662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915779123117546e-05,
+      "loss": 0.9032,
+      "step": 2663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915668643914303e-05,
+      "loss": 1.1815,
+      "step": 2664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991555809260322e-05,
+      "loss": 1.062,
+      "step": 2665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915447469185096e-05,
+      "loss": 0.8735,
+      "step": 2666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915336773660735e-05,
+      "loss": 1.0158,
+      "step": 2667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991522600603095e-05,
+      "loss": 1.019,
+      "step": 2668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991511516629653e-05,
+      "loss": 0.9447,
+      "step": 2669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99150042544583e-05,
+      "loss": 1.1206,
+      "step": 2670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914893270517054e-05,
+      "loss": 1.1788,
+      "step": 2671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914782214473606e-05,
+      "loss": 0.8827,
+      "step": 2672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991467108632876e-05,
+      "loss": 1.3338,
+      "step": 2673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914559886083323e-05,
+      "loss": 1.1499,
+      "step": 2674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914448613738107e-05,
+      "loss": 0.8531,
+      "step": 2675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914337269293917e-05,
+      "loss": 1.4087,
+      "step": 2676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991422585275157e-05,
+      "loss": 0.9231,
+      "step": 2677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914114364111866e-05,
+      "loss": 1.1939,
+      "step": 2678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914002803375627e-05,
+      "loss": 1.4278,
+      "step": 2679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913891170543658e-05,
+      "loss": 1.395,
+      "step": 2680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913779465616767e-05,
+      "loss": 1.2758,
+      "step": 2681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913667688595775e-05,
+      "loss": 1.2071,
+      "step": 2682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913555839481488e-05,
+      "loss": 0.955,
+      "step": 2683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913443918274723e-05,
+      "loss": 0.8912,
+      "step": 2684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913331924976295e-05,
+      "loss": 0.8519,
+      "step": 2685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913219859587015e-05,
+      "loss": 1.0756,
+      "step": 2686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99131077221077e-05,
+      "loss": 1.252,
+      "step": 2687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912995512539162e-05,
+      "loss": 1.249,
+      "step": 2688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912883230882222e-05,
+      "loss": 0.9447,
+      "step": 2689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912770877137696e-05,
+      "loss": 0.9618,
+      "step": 2690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912658451306396e-05,
+      "loss": 0.9519,
+      "step": 2691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912545953389144e-05,
+      "loss": 0.9909,
+      "step": 2692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991243338338676e-05,
+      "loss": 1.1127,
+      "step": 2693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912320741300054e-05,
+      "loss": 1.2989,
+      "step": 2694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912208027129856e-05,
+      "loss": 0.9538,
+      "step": 2695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912095240876978e-05,
+      "loss": 1.0042,
+      "step": 2696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911982382542243e-05,
+      "loss": 1.0627,
+      "step": 2697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911869452126472e-05,
+      "loss": 0.8869,
+      "step": 2698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911756449630487e-05,
+      "loss": 1.1697,
+      "step": 2699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991164337505511e-05,
+      "loss": 0.9475,
+      "step": 2700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911530228401156e-05,
+      "loss": 0.9393,
+      "step": 2701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911417009669454e-05,
+      "loss": 1.1721,
+      "step": 2702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991130371886083e-05,
+      "loss": 0.946,
+      "step": 2703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911190355976102e-05,
+      "loss": 0.9955,
+      "step": 2704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99110769210161e-05,
+      "loss": 0.8737,
+      "step": 2705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910963413981642e-05,
+      "loss": 1.7607,
+      "step": 2706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910849834873562e-05,
+      "loss": 1.1382,
+      "step": 2707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910736183692682e-05,
+      "loss": 1.1109,
+      "step": 2708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910622460439825e-05,
+      "loss": 0.9277,
+      "step": 2709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991050866511582e-05,
+      "loss": 1.3015,
+      "step": 2710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910394797721496e-05,
+      "loss": 0.7767,
+      "step": 2711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910280858257686e-05,
+      "loss": 1.6245,
+      "step": 2712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910166846725205e-05,
+      "loss": 1.0151,
+      "step": 2713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910052763124894e-05,
+      "loss": 1.3772,
+      "step": 2714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990993860745758e-05,
+      "loss": 1.2097,
+      "step": 2715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990982437972409e-05,
+      "loss": 1.1305,
+      "step": 2716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909710079925254e-05,
+      "loss": 1.0119,
+      "step": 2717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990959570806191e-05,
+      "loss": 1.0972,
+      "step": 2718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909481264134886e-05,
+      "loss": 1.0723,
+      "step": 2719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909366748145015e-05,
+      "loss": 1.1871,
+      "step": 2720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909252160093125e-05,
+      "loss": 1.102,
+      "step": 2721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909137499980052e-05,
+      "loss": 1.2951,
+      "step": 2722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909022767806634e-05,
+      "loss": 0.9322,
+      "step": 2723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99089079635737e-05,
+      "loss": 1.1441,
+      "step": 2724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990879308728209e-05,
+      "loss": 0.9423,
+      "step": 2725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908678138932632e-05,
+      "loss": 1.1401,
+      "step": 2726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908563118526167e-05,
+      "loss": 1.2051,
+      "step": 2727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990844802606353e-05,
+      "loss": 1.2402,
+      "step": 2728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908332861545564e-05,
+      "loss": 0.8656,
+      "step": 2729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99082176249731e-05,
+      "loss": 0.8275,
+      "step": 2730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990810231634697e-05,
+      "loss": 1.3855,
+      "step": 2731
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907986935668028e-05,
+      "loss": 1.0219,
+      "step": 2732
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99078714829371e-05,
+      "loss": 1.0853,
+      "step": 2733
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990775595815503e-05,
+      "loss": 1.2049,
+      "step": 2734
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990764036132266e-05,
+      "loss": 0.8159,
+      "step": 2735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990752469244083e-05,
+      "loss": 1.153,
+      "step": 2736
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907408951510382e-05,
+      "loss": 0.9963,
+      "step": 2737
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990729313853215e-05,
+      "loss": 1.2062,
+      "step": 2738
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907177253506985e-05,
+      "loss": 1.1705,
+      "step": 2739
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990706129643573e-05,
+      "loss": 1.4386,
+      "step": 2740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990694526731922e-05,
+      "loss": 1.0181,
+      "step": 2741
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990682916615831e-05,
+      "loss": 1.0836,
+      "step": 2742
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990671299295383e-05,
+      "loss": 1.0066,
+      "step": 2743
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906596747706638e-05,
+      "loss": 0.8075,
+      "step": 2744
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906480430417575e-05,
+      "loss": 1.0094,
+      "step": 2745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906364041087486e-05,
+      "loss": 1.3169,
+      "step": 2746
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906247579717214e-05,
+      "loss": 1.2262,
+      "step": 2747
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906131046307614e-05,
+      "loss": 1.0605,
+      "step": 2748
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906014440859528e-05,
+      "loss": 1.2013,
+      "step": 2749
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905897763373804e-05,
+      "loss": 1.4695,
+      "step": 2750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905781013851292e-05,
+      "loss": 1.0291,
+      "step": 2751
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905664192292837e-05,
+      "loss": 2.2342,
+      "step": 2752
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905547298699294e-05,
+      "loss": 1.1771,
+      "step": 2753
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905430333071515e-05,
+      "loss": 1.0852,
+      "step": 2754
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905313295410343e-05,
+      "loss": 1.1659,
+      "step": 2755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905196185716632e-05,
+      "loss": 1.3492,
+      "step": 2756
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905079003991236e-05,
+      "loss": 1.1221,
+      "step": 2757
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904961750235005e-05,
+      "loss": 1.2579,
+      "step": 2758
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904844424448796e-05,
+      "loss": 0.7128,
+      "step": 2759
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904727026633453e-05,
+      "loss": 1.1839,
+      "step": 2760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904609556789837e-05,
+      "loss": 1.0264,
+      "step": 2761
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904492014918802e-05,
+      "loss": 0.9634,
+      "step": 2762
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904374401021197e-05,
+      "loss": 1.1919,
+      "step": 2763
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904256715097884e-05,
+      "loss": 1.1749,
+      "step": 2764
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990413895714972e-05,
+      "loss": 1.0147,
+      "step": 2765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904021127177554e-05,
+      "loss": 1.2808,
+      "step": 2766
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903903225182245e-05,
+      "loss": 0.8509,
+      "step": 2767
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903785251164652e-05,
+      "loss": 1.2403,
+      "step": 2768
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903667205125637e-05,
+      "loss": 1.1116,
+      "step": 2769
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903549087066053e-05,
+      "loss": 0.9128,
+      "step": 2770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903430896986758e-05,
+      "loss": 1.3026,
+      "step": 2771
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903312634888615e-05,
+      "loss": 1.2192,
+      "step": 2772
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903194300772482e-05,
+      "loss": 0.8289,
+      "step": 2773
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903075894639222e-05,
+      "loss": 1.1687,
+      "step": 2774
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902957416489693e-05,
+      "loss": 1.0159,
+      "step": 2775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990283886632476e-05,
+      "loss": 1.2317,
+      "step": 2776
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990272024414528e-05,
+      "loss": 1.1146,
+      "step": 2777
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990260154995212e-05,
+      "loss": 0.9783,
+      "step": 2778
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902482783746144e-05,
+      "loss": 1.2252,
+      "step": 2779
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990236394552821e-05,
+      "loss": 0.8486,
+      "step": 2780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902245035299187e-05,
+      "loss": 1.1187,
+      "step": 2781
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990212605305994e-05,
+      "loss": 1.2109,
+      "step": 2782
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990200699881133e-05,
+      "loss": 1.0366,
+      "step": 2783
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901887872554227e-05,
+      "loss": 1.2123,
+      "step": 2784
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901768674289497e-05,
+      "loss": 0.8876,
+      "step": 2785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901649404018006e-05,
+      "loss": 1.2294,
+      "step": 2786
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990153006174062e-05,
+      "loss": 1.2526,
+      "step": 2787
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901410647458205e-05,
+      "loss": 1.316,
+      "step": 2788
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901291161171638e-05,
+      "loss": 0.9957,
+      "step": 2789
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901171602881778e-05,
+      "loss": 1.0543,
+      "step": 2790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99010519725895e-05,
+      "loss": 1.1431,
+      "step": 2791
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900932270295673e-05,
+      "loss": 1.0035,
+      "step": 2792
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900812496001163e-05,
+      "loss": 1.0711,
+      "step": 2793
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900692649706846e-05,
+      "loss": 1.2419,
+      "step": 2794
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900572731413598e-05,
+      "loss": 1.0098,
+      "step": 2795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990045274112228e-05,
+      "loss": 1.0792,
+      "step": 2796
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900332678833774e-05,
+      "loss": 0.9145,
+      "step": 2797
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900212544548946e-05,
+      "loss": 1.0957,
+      "step": 2798
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900092338268677e-05,
+      "loss": 1.1751,
+      "step": 2799
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989997205999384e-05,
+      "loss": 1.1623,
+      "step": 2800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899851709725296e-05,
+      "loss": 1.2637,
+      "step": 2801
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899731287463938e-05,
+      "loss": 0.946,
+      "step": 2802
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899610793210633e-05,
+      "loss": 1.0803,
+      "step": 2803
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899490226966264e-05,
+      "loss": 0.8901,
+      "step": 2804
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899369588731697e-05,
+      "loss": 1.029,
+      "step": 2805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899248878507816e-05,
+      "loss": 1.0333,
+      "step": 2806
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899128096295497e-05,
+      "loss": 1.1686,
+      "step": 2807
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899007242095623e-05,
+      "loss": 1.1579,
+      "step": 2808
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898886315909064e-05,
+      "loss": 0.9135,
+      "step": 2809
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898765317736706e-05,
+      "loss": 0.8757,
+      "step": 2810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989864424757943e-05,
+      "loss": 1.2534,
+      "step": 2811
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989852310543811e-05,
+      "loss": 1.0457,
+      "step": 2812
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898401891313636e-05,
+      "loss": 1.0877,
+      "step": 2813
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989828060520688e-05,
+      "loss": 1.2368,
+      "step": 2814
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989815924711873e-05,
+      "loss": 1.1233,
+      "step": 2815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898037817050067e-05,
+      "loss": 1.1181,
+      "step": 2816
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897916315001773e-05,
+      "loss": 1.0413,
+      "step": 2817
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897794740974734e-05,
+      "loss": 1.0925,
+      "step": 2818
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989767309496983e-05,
+      "loss": 1.1163,
+      "step": 2819
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897551376987948e-05,
+      "loss": 0.9593,
+      "step": 2820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897429587029977e-05,
+      "loss": 1.273,
+      "step": 2821
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897307725096797e-05,
+      "loss": 1.015,
+      "step": 2822
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897185791189294e-05,
+      "loss": 1.1975,
+      "step": 2823
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989706378530836e-05,
+      "loss": 1.221,
+      "step": 2824
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989694170745488e-05,
+      "loss": 0.8131,
+      "step": 2825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896819557629735e-05,
+      "loss": 0.9438,
+      "step": 2826
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896697335833822e-05,
+      "loss": 0.8588,
+      "step": 2827
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989657504206803e-05,
+      "loss": 1.7278,
+      "step": 2828
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896452676333244e-05,
+      "loss": 1.3237,
+      "step": 2829
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896330238630354e-05,
+      "loss": 1.0913,
+      "step": 2830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989620772896025e-05,
+      "loss": 1.192,
+      "step": 2831
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896085147323827e-05,
+      "loss": 1.3035,
+      "step": 2832
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895962493721973e-05,
+      "loss": 1.0687,
+      "step": 2833
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989583976815558e-05,
+      "loss": 1.713,
+      "step": 2834
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895716970625544e-05,
+      "loss": 0.8462,
+      "step": 2835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895594101132752e-05,
+      "loss": 2.8152,
+      "step": 2836
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98954711596781e-05,
+      "loss": 1.0686,
+      "step": 2837
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989534814626249e-05,
+      "loss": 1.4801,
+      "step": 2838
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98952250608868e-05,
+      "loss": 1.0883,
+      "step": 2839
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989510190355194e-05,
+      "loss": 2.012,
+      "step": 2840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98949786742588e-05,
+      "loss": 1.0578,
+      "step": 2841
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894855373008273e-05,
+      "loss": 1.1709,
+      "step": 2842
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989473199980126e-05,
+      "loss": 1.1242,
+      "step": 2843
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894608554638664e-05,
+      "loss": 1.1451,
+      "step": 2844
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989448503752137e-05,
+      "loss": 0.8578,
+      "step": 2845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989436144845028e-05,
+      "loss": 1.2181,
+      "step": 2846
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894237787426298e-05,
+      "loss": 1.2672,
+      "step": 2847
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989411405445032e-05,
+      "loss": 0.9444,
+      "step": 2848
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893990249523244e-05,
+      "loss": 0.6499,
+      "step": 2849
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893866372645975e-05,
+      "loss": 0.8269,
+      "step": 2850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989374242381941e-05,
+      "loss": 0.8719,
+      "step": 2851
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989361840304445e-05,
+      "loss": 1.2278,
+      "step": 2852
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893494310321998e-05,
+      "loss": 1.1641,
+      "step": 2853
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893370145652956e-05,
+      "loss": 1.0878,
+      "step": 2854
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893245909038233e-05,
+      "loss": 1.099,
+      "step": 2855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893121600478724e-05,
+      "loss": 1.1221,
+      "step": 2856
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892997219975333e-05,
+      "loss": 1.2399,
+      "step": 2857
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892872767528972e-05,
+      "loss": 0.8975,
+      "step": 2858
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989274824314054e-05,
+      "loss": 0.9247,
+      "step": 2859
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892623646810943e-05,
+      "loss": 1.2367,
+      "step": 2860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989249897854109e-05,
+      "loss": 1.2433,
+      "step": 2861
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892374238331885e-05,
+      "loss": 1.196,
+      "step": 2862
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892249426184238e-05,
+      "loss": 1.1247,
+      "step": 2863
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989212454209905e-05,
+      "loss": 1.1383,
+      "step": 2864
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989199958607724e-05,
+      "loss": 1.1144,
+      "step": 2865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989187455811971e-05,
+      "loss": 1.8944,
+      "step": 2866
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891749458227364e-05,
+      "loss": 0.8941,
+      "step": 2867
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989162428640112e-05,
+      "loss": 0.9833,
+      "step": 2868
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989149904264189e-05,
+      "loss": 1.5035,
+      "step": 2869
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891373726950577e-05,
+      "loss": 1.1189,
+      "step": 2870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891248339328095e-05,
+      "loss": 0.9967,
+      "step": 2871
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989112287977536e-05,
+      "loss": 1.0905,
+      "step": 2872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890997348293275e-05,
+      "loss": 1.0302,
+      "step": 2873
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890871744882763e-05,
+      "loss": 0.9034,
+      "step": 2874
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989074606954473e-05,
+      "loss": 1.1498,
+      "step": 2875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98906203222801e-05,
+      "loss": 1.0962,
+      "step": 2876
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890494503089773e-05,
+      "loss": 1.3475,
+      "step": 2877
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890368611974678e-05,
+      "loss": 0.818,
+      "step": 2878
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890242648935722e-05,
+      "loss": 0.9104,
+      "step": 2879
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890116613973822e-05,
+      "loss": 1.1754,
+      "step": 2880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889990507089898e-05,
+      "loss": 1.0203,
+      "step": 2881
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889864328284863e-05,
+      "loss": 1.3142,
+      "step": 2882
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889738077559633e-05,
+      "loss": 0.9376,
+      "step": 2883
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889611754915136e-05,
+      "loss": 1.1165,
+      "step": 2884
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988948536035228e-05,
+      "loss": 0.784,
+      "step": 2885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889358893871987e-05,
+      "loss": 0.9936,
+      "step": 2886
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889232355475184e-05,
+      "loss": 1.2316,
+      "step": 2887
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988910574516278e-05,
+      "loss": 0.8897,
+      "step": 2888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98889790629357e-05,
+      "loss": 0.955,
+      "step": 2889
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988885230879487e-05,
+      "loss": 1.1559,
+      "step": 2890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888725482741207e-05,
+      "loss": 1.279,
+      "step": 2891
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888598584775635e-05,
+      "loss": 1.4376,
+      "step": 2892
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888471614899074e-05,
+      "loss": 1.154,
+      "step": 2893
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988834457311245e-05,
+      "loss": 0.9292,
+      "step": 2894
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888217459416685e-05,
+      "loss": 0.8675,
+      "step": 2895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888090273812707e-05,
+      "loss": 1.0299,
+      "step": 2896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887963016301436e-05,
+      "loss": 1.0069,
+      "step": 2897
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887835686883803e-05,
+      "loss": 1.3246,
+      "step": 2898
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887708285560727e-05,
+      "loss": 0.9197,
+      "step": 2899
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988758081233314e-05,
+      "loss": 1.2187,
+      "step": 2900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887453267201968e-05,
+      "loss": 1.2558,
+      "step": 2901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887325650168135e-05,
+      "loss": 1.3043,
+      "step": 2902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988719796123258e-05,
+      "loss": 0.8455,
+      "step": 2903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887070200396214e-05,
+      "loss": 0.9956,
+      "step": 2904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988694236765998e-05,
+      "loss": 1.175,
+      "step": 2905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886814463024803e-05,
+      "loss": 1.212,
+      "step": 2906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886686486491614e-05,
+      "loss": 1.0607,
+      "step": 2907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988655843806134e-05,
+      "loss": 0.9818,
+      "step": 2908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988643031773492e-05,
+      "loss": 1.2711,
+      "step": 2909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886302125513276e-05,
+      "loss": 1.5912,
+      "step": 2910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886173861397348e-05,
+      "loss": 1.1083,
+      "step": 2911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886045525388064e-05,
+      "loss": 1.4889,
+      "step": 2912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988591711748636e-05,
+      "loss": 1.0795,
+      "step": 2913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885788637693165e-05,
+      "loss": 1.0808,
+      "step": 2914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885660086009425e-05,
+      "loss": 1.2933,
+      "step": 2915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988553146243606e-05,
+      "loss": 0.9164,
+      "step": 2916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988540276697402e-05,
+      "loss": 0.7906,
+      "step": 2917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988527399962423e-05,
+      "loss": 0.9515,
+      "step": 2918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885145160387623e-05,
+      "loss": 1.0242,
+      "step": 2919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988501624926515e-05,
+      "loss": 1.2757,
+      "step": 2920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988488726625774e-05,
+      "loss": 1.2596,
+      "step": 2921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884758211366333e-05,
+      "loss": 1.297,
+      "step": 2922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884629084591864e-05,
+      "loss": 0.9926,
+      "step": 2923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884499885935276e-05,
+      "loss": 1.1721,
+      "step": 2924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884370615397507e-05,
+      "loss": 1.2282,
+      "step": 2925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98842412729795e-05,
+      "loss": 0.966,
+      "step": 2926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988411185868219e-05,
+      "loss": 0.9799,
+      "step": 2927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988398237250652e-05,
+      "loss": 1.0856,
+      "step": 2928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883852814453434e-05,
+      "loss": 1.1161,
+      "step": 2929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988372318452387e-05,
+      "loss": 1.2509,
+      "step": 2930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883593482718776e-05,
+      "loss": 1.1169,
+      "step": 2931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883463709039095e-05,
+      "loss": 2.4092,
+      "step": 2932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883333863485763e-05,
+      "loss": 1.0932,
+      "step": 2933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883203946059733e-05,
+      "loss": 0.9239,
+      "step": 2934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883073956761948e-05,
+      "loss": 1.1916,
+      "step": 2935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988294389559335e-05,
+      "loss": 1.0085,
+      "step": 2936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882813762554883e-05,
+      "loss": 0.7865,
+      "step": 2937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98826835576475e-05,
+      "loss": 1.1244,
+      "step": 2938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882553280872144e-05,
+      "loss": 1.0591,
+      "step": 2939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882422932229765e-05,
+      "loss": 1.0589,
+      "step": 2940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882292511721307e-05,
+      "loss": 0.9762,
+      "step": 2941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988216201934772e-05,
+      "loss": 1.256,
+      "step": 2942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882031455109957e-05,
+      "loss": 1.4891,
+      "step": 2943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881900819008964e-05,
+      "loss": 1.6556,
+      "step": 2944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988177011104569e-05,
+      "loss": 1.2496,
+      "step": 2945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881639331221083e-05,
+      "loss": 1.0731,
+      "step": 2946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881508479536097e-05,
+      "loss": 1.132,
+      "step": 2947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881377555991687e-05,
+      "loss": 1.2326,
+      "step": 2948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98812465605888e-05,
+      "loss": 1.15,
+      "step": 2949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881115493328393e-05,
+      "loss": 0.7843,
+      "step": 2950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880984354211416e-05,
+      "loss": 1.029,
+      "step": 2951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880853143238824e-05,
+      "loss": 1.169,
+      "step": 2952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988072186041157e-05,
+      "loss": 1.0182,
+      "step": 2953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988059050573061e-05,
+      "loss": 1.2756,
+      "step": 2954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880459079196898e-05,
+      "loss": 0.8112,
+      "step": 2955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988032758081139e-05,
+      "loss": 0.7253,
+      "step": 2956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880196010575043e-05,
+      "loss": 1.0365,
+      "step": 2957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880064368488817e-05,
+      "loss": 1.1987,
+      "step": 2958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987993265455366e-05,
+      "loss": 1.0326,
+      "step": 2959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987980086877054e-05,
+      "loss": 0.8638,
+      "step": 2960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987966901114041e-05,
+      "loss": 0.9965,
+      "step": 2961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879537081664225e-05,
+      "loss": 1.1682,
+      "step": 2962
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879405080342957e-05,
+      "loss": 1.291,
+      "step": 2963
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879273007177555e-05,
+      "loss": 1.0697,
+      "step": 2964
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987914086216898e-05,
+      "loss": 1.3152,
+      "step": 2965
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98790086453182e-05,
+      "loss": 1.0259,
+      "step": 2966
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878876356626168e-05,
+      "loss": 1.3335,
+      "step": 2967
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987874399609385e-05,
+      "loss": 1.0599,
+      "step": 2968
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878611563722212e-05,
+      "loss": 1.1129,
+      "step": 2969
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878479059512212e-05,
+      "loss": 1.0636,
+      "step": 2970
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878346483464815e-05,
+      "loss": 1.0252,
+      "step": 2971
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878213835580985e-05,
+      "loss": 1.004,
+      "step": 2972
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878081115861686e-05,
+      "loss": 1.1078,
+      "step": 2973
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877948324307888e-05,
+      "loss": 0.9584,
+      "step": 2974
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877815460920553e-05,
+      "loss": 1.298,
+      "step": 2975
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877682525700643e-05,
+      "loss": 1.1336,
+      "step": 2976
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987754951864913e-05,
+      "loss": 1.2612,
+      "step": 2977
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987741643976698e-05,
+      "loss": 1.2936,
+      "step": 2978
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877283289055164e-05,
+      "loss": 1.224,
+      "step": 2979
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877150066514644e-05,
+      "loss": 0.9073,
+      "step": 2980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877016772146395e-05,
+      "loss": 2.1325,
+      "step": 2981
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876883405951378e-05,
+      "loss": 1.2313,
+      "step": 2982
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876749967930573e-05,
+      "loss": 1.1213,
+      "step": 2983
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987661645808494e-05,
+      "loss": 1.1978,
+      "step": 2984
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987648287641546e-05,
+      "loss": 3.9549,
+      "step": 2985
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98763492229231e-05,
+      "loss": 1.091,
+      "step": 2986
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987621549760883e-05,
+      "loss": 0.9322,
+      "step": 2987
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876081700473624e-05,
+      "loss": 0.8208,
+      "step": 2988
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875947831518454e-05,
+      "loss": 1.4588,
+      "step": 2989
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98758138907443e-05,
+      "loss": 1.8823,
+      "step": 2990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875679878152124e-05,
+      "loss": 0.9553,
+      "step": 2991
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987554579374291e-05,
+      "loss": 0.8885,
+      "step": 2992
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987541163751763e-05,
+      "loss": 1.5221,
+      "step": 2993
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987527740947726e-05,
+      "loss": 1.5255,
+      "step": 2994
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987514310962278e-05,
+      "loss": 5.4677,
+      "step": 2995
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875008737955155e-05,
+      "loss": 1.3346,
+      "step": 2996
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874874294475375e-05,
+      "loss": 1.2859,
+      "step": 2997
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987473977918441e-05,
+      "loss": 0.9807,
+      "step": 2998
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874605192083245e-05,
+      "loss": 1.2618,
+      "step": 2999
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987447053317285e-05,
+      "loss": 1.2357,
+      "step": 3000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987433580245421e-05,
+      "loss": 1.3005,
+      "step": 3001
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874200999928303e-05,
+      "loss": 1.1337,
+      "step": 3002
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987406612559611e-05,
+      "loss": 1.1306,
+      "step": 3003
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873931179458615e-05,
+      "loss": 1.2284,
+      "step": 3004
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873796161516794e-05,
+      "loss": 1.0554,
+      "step": 3005
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987366107177163e-05,
+      "loss": 0.8793,
+      "step": 3006
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873525910224106e-05,
+      "loss": 2.4062,
+      "step": 3007
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873390676875207e-05,
+      "loss": 0.9888,
+      "step": 3008
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873255371725913e-05,
+      "loss": 1.0928,
+      "step": 3009
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987311999477721e-05,
+      "loss": 1.3638,
+      "step": 3010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872984546030083e-05,
+      "loss": 1.5402,
+      "step": 3011
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872849025485515e-05,
+      "loss": 1.1428,
+      "step": 3012
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872713433144493e-05,
+      "loss": 1.0913,
+      "step": 3013
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872577769008e-05,
+      "loss": 1.3833,
+      "step": 3014
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872442033077027e-05,
+      "loss": 1.1551,
+      "step": 3015
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987230622535256e-05,
+      "loss": 1.3586,
+      "step": 3016
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872170345835587e-05,
+      "loss": 1.1561,
+      "step": 3017
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987203439452709e-05,
+      "loss": 1.026,
+      "step": 3018
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871898371428067e-05,
+      "loss": 1.2798,
+      "step": 3019
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98717622765395e-05,
+      "loss": 0.9809,
+      "step": 3020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987162610986238e-05,
+      "loss": 1.0937,
+      "step": 3021
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871489871397705e-05,
+      "loss": 1.2714,
+      "step": 3022
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987135356114646e-05,
+      "loss": 1.0381,
+      "step": 3023
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987121717910963e-05,
+      "loss": 1.1102,
+      "step": 3024
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871080725288213e-05,
+      "loss": 1.0307,
+      "step": 3025
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870944199683203e-05,
+      "loss": 1.0281,
+      "step": 3026
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987080760229559e-05,
+      "loss": 1.8901,
+      "step": 3027
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987067093312637e-05,
+      "loss": 1.2935,
+      "step": 3028
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987053419217653e-05,
+      "loss": 0.9155,
+      "step": 3029
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870397379447074e-05,
+      "loss": 1.0058,
+      "step": 3030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987026049493899e-05,
+      "loss": 1.0549,
+      "step": 3031
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870123538653273e-05,
+      "loss": 1.2282,
+      "step": 3032
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869986510590925e-05,
+      "loss": 1.2839,
+      "step": 3033
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986984941075294e-05,
+      "loss": 0.828,
+      "step": 3034
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986971223914031e-05,
+      "loss": 1.0262,
+      "step": 3035
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869574995754043e-05,
+      "loss": 0.9235,
+      "step": 3036
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869437680595125e-05,
+      "loss": 1.3565,
+      "step": 3037
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986930029366456e-05,
+      "loss": 1.2012,
+      "step": 3038
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986916283496335e-05,
+      "loss": 0.9056,
+      "step": 3039
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869025304492492e-05,
+      "loss": 1.2112,
+      "step": 3040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868887702252987e-05,
+      "loss": 1.2927,
+      "step": 3041
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986875002824583e-05,
+      "loss": 1.2004,
+      "step": 3042
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986861228247203e-05,
+      "loss": 0.9922,
+      "step": 3043
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868474464932587e-05,
+      "loss": 1.1185,
+      "step": 3044
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98683365756285e-05,
+      "loss": 0.8019,
+      "step": 3045
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868198614560776e-05,
+      "loss": 1.3784,
+      "step": 3046
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986806058173041e-05,
+      "loss": 1.1176,
+      "step": 3047
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867922477138416e-05,
+      "loss": 1.1806,
+      "step": 3048
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867784300785793e-05,
+      "loss": 0.9842,
+      "step": 3049
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986764605267355e-05,
+      "loss": 0.9479,
+      "step": 3050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867507732802685e-05,
+      "loss": 1.3587,
+      "step": 3051
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986736934117421e-05,
+      "loss": 1.5282,
+      "step": 3052
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867230877789127e-05,
+      "loss": 0.843,
+      "step": 3053
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867092342648446e-05,
+      "loss": 1.9497,
+      "step": 3054
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866953735753177e-05,
+      "loss": 0.8874,
+      "step": 3055
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986681505710432e-05,
+      "loss": 0.913,
+      "step": 3056
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866676306702893e-05,
+      "loss": 1.0614,
+      "step": 3057
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98665374845499e-05,
+      "loss": 1.0516,
+      "step": 3058
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866398590646345e-05,
+      "loss": 0.901,
+      "step": 3059
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866259624993246e-05,
+      "loss": 1.191,
+      "step": 3060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986612058759161e-05,
+      "loss": 1.2029,
+      "step": 3061
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865981478442454e-05,
+      "loss": 1.315,
+      "step": 3062
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865842297546784e-05,
+      "loss": 0.8033,
+      "step": 3063
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865703044905614e-05,
+      "loss": 1.1093,
+      "step": 3064
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865563720519954e-05,
+      "loss": 1.1874,
+      "step": 3065
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986542432439082e-05,
+      "loss": 1.1781,
+      "step": 3066
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865284856519226e-05,
+      "loss": 1.1613,
+      "step": 3067
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986514531690618e-05,
+      "loss": 1.2869,
+      "step": 3068
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865005705552706e-05,
+      "loss": 1.1546,
+      "step": 3069
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864866022459814e-05,
+      "loss": 1.031,
+      "step": 3070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986472626762852e-05,
+      "loss": 0.8686,
+      "step": 3071
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864586441059845e-05,
+      "loss": 0.863,
+      "step": 3072
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864446542754797e-05,
+      "loss": 1.0581,
+      "step": 3073
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98643065727144e-05,
+      "loss": 1.063,
+      "step": 3074
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986416653093967e-05,
+      "loss": 1.0552,
+      "step": 3075
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986402641743163e-05,
+      "loss": 0.9139,
+      "step": 3076
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863886232191287e-05,
+      "loss": 0.8411,
+      "step": 3077
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863745975219672e-05,
+      "loss": 0.8595,
+      "step": 3078
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98636056465178e-05,
+      "loss": 1.2407,
+      "step": 3079
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863465246086695e-05,
+      "loss": 1.1656,
+      "step": 3080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863324773927372e-05,
+      "loss": 1.0226,
+      "step": 3081
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986318423004086e-05,
+      "loss": 1.228,
+      "step": 3082
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863043614428173e-05,
+      "loss": 0.8986,
+      "step": 3083
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986290292709034e-05,
+      "loss": 0.8688,
+      "step": 3084
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862762168028377e-05,
+      "loss": 1.0469,
+      "step": 3085
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862621337243318e-05,
+      "loss": 1.2178,
+      "step": 3086
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986248043473618e-05,
+      "loss": 1.3519,
+      "step": 3087
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862339460507986e-05,
+      "loss": 1.2173,
+      "step": 3088
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862198414559767e-05,
+      "loss": 1.0278,
+      "step": 3089
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862057296892546e-05,
+      "loss": 1.1262,
+      "step": 3090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861916107507347e-05,
+      "loss": 1.1426,
+      "step": 3091
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861774846405203e-05,
+      "loss": 2.6294,
+      "step": 3092
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861633513587135e-05,
+      "loss": 0.7474,
+      "step": 3093
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861492109054175e-05,
+      "loss": 1.2891,
+      "step": 3094
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861350632807347e-05,
+      "loss": 0.999,
+      "step": 3095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986120908484768e-05,
+      "loss": 0.9952,
+      "step": 3096
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986106746517621e-05,
+      "loss": 0.9343,
+      "step": 3097
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986092577379396e-05,
+      "loss": 0.9168,
+      "step": 3098
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986078401070196e-05,
+      "loss": 0.8954,
+      "step": 3099
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860642175901248e-05,
+      "loss": 1.214,
+      "step": 3100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986050026939285e-05,
+      "loss": 1.029,
+      "step": 3101
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860358291177796e-05,
+      "loss": 1.0755,
+      "step": 3102
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860216241257125e-05,
+      "loss": 0.808,
+      "step": 3103
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860074119631866e-05,
+      "loss": 1.2635,
+      "step": 3104
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985993192630305e-05,
+      "loss": 1.0014,
+      "step": 3105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985978966127172e-05,
+      "loss": 1.2098,
+      "step": 3106
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859647324538902e-05,
+      "loss": 1.1184,
+      "step": 3107
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859504916105634e-05,
+      "loss": 1.2061,
+      "step": 3108
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985936243597295e-05,
+      "loss": 1.2887,
+      "step": 3109
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859219884141887e-05,
+      "loss": 1.2242,
+      "step": 3110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859077260613486e-05,
+      "loss": 0.9695,
+      "step": 3111
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858934565388777e-05,
+      "loss": 2.1086,
+      "step": 3112
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98587917984688e-05,
+      "loss": 1.0691,
+      "step": 3113
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858648959854596e-05,
+      "loss": 0.8521,
+      "step": 3114
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858506049547205e-05,
+      "loss": 1.137,
+      "step": 3115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985836306754766e-05,
+      "loss": 1.0816,
+      "step": 3116
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858220013857002e-05,
+      "loss": 0.8868,
+      "step": 3117
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858076888476277e-05,
+      "loss": 1.1564,
+      "step": 3118
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857933691406522e-05,
+      "loss": 1.0341,
+      "step": 3119
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857790422648774e-05,
+      "loss": 1.064,
+      "step": 3120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857647082204083e-05,
+      "loss": 0.9601,
+      "step": 3121
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985750367007349e-05,
+      "loss": 1.3277,
+      "step": 3122
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985736018625803e-05,
+      "loss": 2.5101,
+      "step": 3123
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857216630758755e-05,
+      "loss": 1.0247,
+      "step": 3124
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857073003576708e-05,
+      "loss": 1.0612,
+      "step": 3125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985692930471293e-05,
+      "loss": 1.1355,
+      "step": 3126
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985678553416847e-05,
+      "loss": 0.9507,
+      "step": 3127
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856641691944363e-05,
+      "loss": 0.6984,
+      "step": 3128
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985649777804167e-05,
+      "loss": 1.8873,
+      "step": 3129
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985635379246143e-05,
+      "loss": 1.2062,
+      "step": 3130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985620973520469e-05,
+      "loss": 0.879,
+      "step": 3131
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856065606272498e-05,
+      "loss": 1.1538,
+      "step": 3132
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98559214056659e-05,
+      "loss": 1.0789,
+      "step": 3133
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855777133385952e-05,
+      "loss": 1.1267,
+      "step": 3134
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855632789433695e-05,
+      "loss": 0.9684,
+      "step": 3135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855488373810182e-05,
+      "loss": 1.1765,
+      "step": 3136
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855343886516464e-05,
+      "loss": 1.3416,
+      "step": 3137
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985519932755359e-05,
+      "loss": 0.8407,
+      "step": 3138
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855054696922612e-05,
+      "loss": 1.0718,
+      "step": 3139
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854909994624582e-05,
+      "loss": 1.108,
+      "step": 3140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854765220660552e-05,
+      "loss": 1.2631,
+      "step": 3141
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985462037503158e-05,
+      "loss": 0.8428,
+      "step": 3142
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854475457738708e-05,
+      "loss": 1.001,
+      "step": 3143
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854330468782998e-05,
+      "loss": 0.9311,
+      "step": 3144
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854185408165502e-05,
+      "loss": 1.0857,
+      "step": 3145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854040275887273e-05,
+      "loss": 1.1437,
+      "step": 3146
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853895071949373e-05,
+      "loss": 0.8652,
+      "step": 3147
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853749796352852e-05,
+      "loss": 1.4706,
+      "step": 3148
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853604449098773e-05,
+      "loss": 1.0511,
+      "step": 3149
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853459030188183e-05,
+      "loss": 1.2339,
+      "step": 3150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853313539622143e-05,
+      "loss": 1.1527,
+      "step": 3151
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853167977401717e-05,
+      "loss": 1.3468,
+      "step": 3152
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985302234352796e-05,
+      "loss": 1.1415,
+      "step": 3153
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852876638001928e-05,
+      "loss": 1.0459,
+      "step": 3154
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852730860824683e-05,
+      "loss": 1.3772,
+      "step": 3155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852585011997286e-05,
+      "loss": 1.1276,
+      "step": 3156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852439091520795e-05,
+      "loss": 1.1403,
+      "step": 3157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852293099396274e-05,
+      "loss": 1.1014,
+      "step": 3158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852147035624784e-05,
+      "loss": 1.423,
+      "step": 3159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852000900207384e-05,
+      "loss": 0.865,
+      "step": 3160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985185469314514e-05,
+      "loss": 0.8969,
+      "step": 3161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851708414439117e-05,
+      "loss": 1.1349,
+      "step": 3162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851562064090376e-05,
+      "loss": 1.6774,
+      "step": 3163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851415642099982e-05,
+      "loss": 1.7158,
+      "step": 3164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851269148468998e-05,
+      "loss": 1.0821,
+      "step": 3165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851122583198492e-05,
+      "loss": 1.2326,
+      "step": 3166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850975946289527e-05,
+      "loss": 0.8216,
+      "step": 3167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850829237743174e-05,
+      "loss": 1.5046,
+      "step": 3168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850682457560493e-05,
+      "loss": 1.1853,
+      "step": 3169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985053560574256e-05,
+      "loss": 0.9993,
+      "step": 3170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850388682290432e-05,
+      "loss": 1.18,
+      "step": 3171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850241687205186e-05,
+      "loss": 1.0004,
+      "step": 3172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850094620487892e-05,
+      "loss": 1.0722,
+      "step": 3173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849947482139613e-05,
+      "loss": 1.2703,
+      "step": 3174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849800272161424e-05,
+      "loss": 0.8912,
+      "step": 3175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849652990554392e-05,
+      "loss": 0.9812,
+      "step": 3176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984950563731959e-05,
+      "loss": 1.2088,
+      "step": 3177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984935821245809e-05,
+      "loss": 0.9562,
+      "step": 3178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849210715970964e-05,
+      "loss": 1.1848,
+      "step": 3179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849063147859282e-05,
+      "loss": 1.0626,
+      "step": 3180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848915508124118e-05,
+      "loss": 0.8781,
+      "step": 3181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848767796766548e-05,
+      "loss": 1.2608,
+      "step": 3182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848620013787646e-05,
+      "loss": 0.9711,
+      "step": 3183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848472159188487e-05,
+      "loss": 1.1759,
+      "step": 3184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984832423297014e-05,
+      "loss": 0.924,
+      "step": 3185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984817623513369e-05,
+      "loss": 1.197,
+      "step": 3186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848028165680207e-05,
+      "loss": 1.1335,
+      "step": 3187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847880024610768e-05,
+      "loss": 1.1037,
+      "step": 3188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847731811926455e-05,
+      "loss": 1.0923,
+      "step": 3189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984758352762834e-05,
+      "loss": 0.7632,
+      "step": 3190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847435171717507e-05,
+      "loss": 1.2966,
+      "step": 3191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984728674419503e-05,
+      "loss": 1.0313,
+      "step": 3192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984713824506199e-05,
+      "loss": 0.8883,
+      "step": 3193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846989674319468e-05,
+      "loss": 1.0037,
+      "step": 3194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846841031968545e-05,
+      "loss": 1.2702,
+      "step": 3195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846692318010296e-05,
+      "loss": 1.2388,
+      "step": 3196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846543532445813e-05,
+      "loss": 1.0541,
+      "step": 3197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846394675276167e-05,
+      "loss": 1.0519,
+      "step": 3198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846245746502445e-05,
+      "loss": 1.0991,
+      "step": 3199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846096746125735e-05,
+      "loss": 1.2512,
+      "step": 3200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845947674147113e-05,
+      "loss": 0.8671,
+      "step": 3201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984579853056767e-05,
+      "loss": 1.1075,
+      "step": 3202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845649315388482e-05,
+      "loss": 0.793,
+      "step": 3203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984550002861064e-05,
+      "loss": 1.2401,
+      "step": 3204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984535067023523e-05,
+      "loss": 0.9452,
+      "step": 3205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845201240263335e-05,
+      "loss": 1.208,
+      "step": 3206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845051738696048e-05,
+      "loss": 1.0067,
+      "step": 3207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844902165534442e-05,
+      "loss": 1.2817,
+      "step": 3208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844752520779623e-05,
+      "loss": 1.2877,
+      "step": 3209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844602804432667e-05,
+      "loss": 1.1988,
+      "step": 3210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844453016494663e-05,
+      "loss": 1.1884,
+      "step": 3211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984430315696671e-05,
+      "loss": 1.173,
+      "step": 3212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984415322584989e-05,
+      "loss": 0.6746,
+      "step": 3213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984400322314529e-05,
+      "loss": 1.1975,
+      "step": 3214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984385314885401e-05,
+      "loss": 1.1373,
+      "step": 3215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843703002977134e-05,
+      "loss": 1.155,
+      "step": 3216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843552785515756e-05,
+      "loss": 2.1534,
+      "step": 3217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984340249647097e-05,
+      "loss": 1.2456,
+      "step": 3218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984325213584387e-05,
+      "loss": 1.0809,
+      "step": 3219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843101703635547e-05,
+      "loss": 0.8999,
+      "step": 3220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842951199847092e-05,
+      "loss": 0.98,
+      "step": 3221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842800624479607e-05,
+      "loss": 1.2644,
+      "step": 3222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984264997753418e-05,
+      "loss": 0.779,
+      "step": 3223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984249925901191e-05,
+      "loss": 1.2341,
+      "step": 3224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842348468913895e-05,
+      "loss": 1.0826,
+      "step": 3225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842197607241226e-05,
+      "loss": 1.0668,
+      "step": 3226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842046673995002e-05,
+      "loss": 1.0174,
+      "step": 3227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9841895669176327e-05,
+      "loss": 1.2794,
+      "step": 3228
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841744592786288e-05,
+      "loss": 1.0201,
+      "step": 3229
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984159344482599e-05,
+      "loss": 1.2046,
+      "step": 3230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841442225296538e-05,
+      "loss": 1.072,
+      "step": 3231
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984129093419902e-05,
+      "loss": 1.355,
+      "step": 3232
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841139571534538e-05,
+      "loss": 1.0328,
+      "step": 3233
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98409881373042e-05,
+      "loss": 1.2077,
+      "step": 3234
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840836631509103e-05,
+      "loss": 1.0579,
+      "step": 3235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840685054150352e-05,
+      "loss": 1.3966,
+      "step": 3236
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840533405229047e-05,
+      "loss": 2.0316,
+      "step": 3237
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840381684746285e-05,
+      "loss": 1.5216,
+      "step": 3238
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840229892703178e-05,
+      "loss": 0.9555,
+      "step": 3239
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840078029100826e-05,
+      "loss": 1.3658,
+      "step": 3240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839926093940335e-05,
+      "loss": 1.1985,
+      "step": 3241
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839774087222806e-05,
+      "loss": 0.6774,
+      "step": 3242
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839622008949346e-05,
+      "loss": 0.9614,
+      "step": 3243
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839469859121067e-05,
+      "loss": 1.0556,
+      "step": 3244
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839317637739066e-05,
+      "loss": 0.8215,
+      "step": 3245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983916534480446e-05,
+      "loss": 0.943,
+      "step": 3246
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839012980318346e-05,
+      "loss": 1.0933,
+      "step": 3247
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838860544281838e-05,
+      "loss": 0.9659,
+      "step": 3248
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838708036696044e-05,
+      "loss": 1.0368,
+      "step": 3249
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838555457562072e-05,
+      "loss": 1.0975,
+      "step": 3250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838402806881033e-05,
+      "loss": 1.4587,
+      "step": 3251
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838250084654037e-05,
+      "loss": 1.1876,
+      "step": 3252
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838097290882194e-05,
+      "loss": 1.0335,
+      "step": 3253
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837944425566614e-05,
+      "loss": 1.1456,
+      "step": 3254
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983779148870841e-05,
+      "loss": 1.0241,
+      "step": 3255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837638480308693e-05,
+      "loss": 1.0732,
+      "step": 3256
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837485400368578e-05,
+      "loss": 1.0586,
+      "step": 3257
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837332248889174e-05,
+      "loss": 0.8825,
+      "step": 3258
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98371790258716e-05,
+      "loss": 1.24,
+      "step": 3259
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983702573131697e-05,
+      "loss": 1.1116,
+      "step": 3260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983687236522639e-05,
+      "loss": 1.5546,
+      "step": 3261
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836718927600985e-05,
+      "loss": 1.214,
+      "step": 3262
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983656541844187e-05,
+      "loss": 1.1217,
+      "step": 3263
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983641183775016e-05,
+      "loss": 1.2564,
+      "step": 3264
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836258185526965e-05,
+      "loss": 1.1253,
+      "step": 3265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983610446177341e-05,
+      "loss": 0.8357,
+      "step": 3266
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835950666490613e-05,
+      "loss": 0.9817,
+      "step": 3267
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983579679967969e-05,
+      "loss": 1.0939,
+      "step": 3268
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835642861341758e-05,
+      "loss": 1.0785,
+      "step": 3269
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835488851477943e-05,
+      "loss": 1.2824,
+      "step": 3270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835334770089355e-05,
+      "loss": 1.1412,
+      "step": 3271
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835180617177124e-05,
+      "loss": 1.0162,
+      "step": 3272
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983502639274237e-05,
+      "loss": 1.345,
+      "step": 3273
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834872096786212e-05,
+      "loss": 1.2701,
+      "step": 3274
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983471772930977e-05,
+      "loss": 1.185,
+      "step": 3275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834563290314163e-05,
+      "loss": 1.1775,
+      "step": 3276
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834408779800527e-05,
+      "loss": 1.1265,
+      "step": 3277
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834254197769974e-05,
+      "loss": 1.0188,
+      "step": 3278
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834099544223637e-05,
+      "loss": 0.8388,
+      "step": 3279
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833944819162633e-05,
+      "loss": 0.7979,
+      "step": 3280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983379002258809e-05,
+      "loss": 1.1958,
+      "step": 3281
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833635154501136e-05,
+      "loss": 2.6844,
+      "step": 3282
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833480214902892e-05,
+      "loss": 1.2344,
+      "step": 3283
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833325203794488e-05,
+      "loss": 0.9338,
+      "step": 3284
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983317012117705e-05,
+      "loss": 1.091,
+      "step": 3285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833014967051715e-05,
+      "loss": 1.1505,
+      "step": 3286
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832859741419596e-05,
+      "loss": 1.406,
+      "step": 3287
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983270444428183e-05,
+      "loss": 1.2518,
+      "step": 3288
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983254907563955e-05,
+      "loss": 1.079,
+      "step": 3289
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832393635493873e-05,
+      "loss": 1.1154,
+      "step": 3290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832238123845944e-05,
+      "loss": 0.9478,
+      "step": 3291
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832082540696885e-05,
+      "loss": 1.1282,
+      "step": 3292
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983192688604783e-05,
+      "loss": 0.9119,
+      "step": 3293
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831771159899915e-05,
+      "loss": 1.2392,
+      "step": 3294
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831615362254263e-05,
+      "loss": 0.9311,
+      "step": 3295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831459493112013e-05,
+      "loss": 1.2552,
+      "step": 3296
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831303552474298e-05,
+      "loss": 1.2742,
+      "step": 3297
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831147540342255e-05,
+      "loss": 1.2241,
+      "step": 3298
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830991456717013e-05,
+      "loss": 1.1391,
+      "step": 3299
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983083530159971e-05,
+      "loss": 1.1006,
+      "step": 3300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983067907499148e-05,
+      "loss": 1.0988,
+      "step": 3301
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830522776893462e-05,
+      "loss": 1.5648,
+      "step": 3302
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830366407306787e-05,
+      "loss": 1.0479,
+      "step": 3303
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830209966232597e-05,
+      "loss": 1.3058,
+      "step": 3304
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830053453672032e-05,
+      "loss": 1.055,
+      "step": 3305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829896869626223e-05,
+      "loss": 1.407,
+      "step": 3306
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829740214096316e-05,
+      "loss": 0.9729,
+      "step": 3307
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829583487083444e-05,
+      "loss": 1.2356,
+      "step": 3308
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829426688588748e-05,
+      "loss": 1.2515,
+      "step": 3309
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982926981861337e-05,
+      "loss": 1.1987,
+      "step": 3310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829112877158452e-05,
+      "loss": 1.1228,
+      "step": 3311
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828955864225134e-05,
+      "loss": 1.309,
+      "step": 3312
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828798779814557e-05,
+      "loss": 0.8555,
+      "step": 3313
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828641623927865e-05,
+      "loss": 1.0847,
+      "step": 3314
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828484396566197e-05,
+      "loss": 1.2329,
+      "step": 3315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98283270977307e-05,
+      "loss": 1.3463,
+      "step": 3316
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982816972742252e-05,
+      "loss": 0.8386,
+      "step": 3317
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828012285642793e-05,
+      "loss": 0.8535,
+      "step": 3318
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827854772392673e-05,
+      "loss": 1.1658,
+      "step": 3319
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98276971876733e-05,
+      "loss": 1.1656,
+      "step": 3320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982753953148582e-05,
+      "loss": 1.1464,
+      "step": 3321
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827381803831388e-05,
+      "loss": 1.065,
+      "step": 3322
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827224004711137e-05,
+      "loss": 0.9311,
+      "step": 3323
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827066134126226e-05,
+      "loss": 1.29,
+      "step": 3324
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826908192077796e-05,
+      "loss": 1.1459,
+      "step": 3325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826750178567e-05,
+      "loss": 1.2488,
+      "step": 3326
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826592093594985e-05,
+      "loss": 0.8399,
+      "step": 3327
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98264339371629e-05,
+      "loss": 1.0852,
+      "step": 3328
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98262757092719e-05,
+      "loss": 1.1862,
+      "step": 3329
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982611740992313e-05,
+      "loss": 1.1133,
+      "step": 3330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982595903911774e-05,
+      "loss": 0.7969,
+      "step": 3331
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825800596856887e-05,
+      "loss": 1.0705,
+      "step": 3332
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825642083141722e-05,
+      "loss": 1.117,
+      "step": 3333
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825483497973396e-05,
+      "loss": 0.8436,
+      "step": 3334
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825324841353063e-05,
+      "loss": 1.3118,
+      "step": 3335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825166113281876e-05,
+      "loss": 1.0992,
+      "step": 3336
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982500731376099e-05,
+      "loss": 1.067,
+      "step": 3337
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982484844279156e-05,
+      "loss": 1.159,
+      "step": 3338
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824689500374742e-05,
+      "loss": 1.2285,
+      "step": 3339
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824530486511688e-05,
+      "loss": 1.1841,
+      "step": 3340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982437140120356e-05,
+      "loss": 1.0167,
+      "step": 3341
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982421224445151e-05,
+      "loss": 1.1344,
+      "step": 3342
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98240530162567e-05,
+      "loss": 1.2684,
+      "step": 3343
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823893716620284e-05,
+      "loss": 1.5993,
+      "step": 3344
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823734345543422e-05,
+      "loss": 1.0702,
+      "step": 3345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823574903027273e-05,
+      "loss": 1.0858,
+      "step": 3346
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823415389073e-05,
+      "loss": 1.3129,
+      "step": 3347
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823255803681754e-05,
+      "loss": 0.9841,
+      "step": 3348
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823096146854706e-05,
+      "loss": 1.2359,
+      "step": 3349
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822936418593008e-05,
+      "loss": 1.2077,
+      "step": 3350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822776618897827e-05,
+      "loss": 0.8534,
+      "step": 3351
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822616747770324e-05,
+      "loss": 1.0545,
+      "step": 3352
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982245680521166e-05,
+      "loss": 0.9131,
+      "step": 3353
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822296791223e-05,
+      "loss": 1.2784,
+      "step": 3354
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822136705805507e-05,
+      "loss": 0.9388,
+      "step": 3355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821976548960347e-05,
+      "loss": 1.2117,
+      "step": 3356
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821816320688677e-05,
+      "loss": 1.3353,
+      "step": 3357
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821656020991672e-05,
+      "loss": 0.9343,
+      "step": 3358
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821495649870497e-05,
+      "loss": 1.081,
+      "step": 3359
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982133520732631e-05,
+      "loss": 0.9058,
+      "step": 3360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821174693360284e-05,
+      "loss": 1.0437,
+      "step": 3361
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821014107973587e-05,
+      "loss": 1.2229,
+      "step": 3362
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820853451167386e-05,
+      "loss": 0.9923,
+      "step": 3363
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820692722942844e-05,
+      "loss": 1.0089,
+      "step": 3364
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820531923301135e-05,
+      "loss": 1.3979,
+      "step": 3365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982037105224343e-05,
+      "loss": 1.133,
+      "step": 3366
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820210109770894e-05,
+      "loss": 0.967,
+      "step": 3367
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820049095884696e-05,
+      "loss": 1.0378,
+      "step": 3368
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819888010586018e-05,
+      "loss": 1.1319,
+      "step": 3369
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981972685387602e-05,
+      "loss": 1.7249,
+      "step": 3370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819565625755873e-05,
+      "loss": 1.0161,
+      "step": 3371
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981940432622676e-05,
+      "loss": 1.2555,
+      "step": 3372
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981924295528985e-05,
+      "loss": 0.9193,
+      "step": 3373
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981908151294631e-05,
+      "loss": 1.0119,
+      "step": 3374
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981891999919732e-05,
+      "loss": 1.0621,
+      "step": 3375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818758414044056e-05,
+      "loss": 1.1904,
+      "step": 3376
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818596757487693e-05,
+      "loss": 1.3223,
+      "step": 3377
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818435029529397e-05,
+      "loss": 0.6885,
+      "step": 3378
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818273230170354e-05,
+      "loss": 1.0436,
+      "step": 3379
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981811135941174e-05,
+      "loss": 1.0171,
+      "step": 3380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817949417254728e-05,
+      "loss": 1.0609,
+      "step": 3381
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817787403700497e-05,
+      "loss": 1.3482,
+      "step": 3382
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817625318750228e-05,
+      "loss": 0.8797,
+      "step": 3383
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817463162405093e-05,
+      "loss": 1.6058,
+      "step": 3384
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817300934666275e-05,
+      "loss": 1.5514,
+      "step": 3385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981713863553496e-05,
+      "loss": 1.2248,
+      "step": 3386
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981697626501232e-05,
+      "loss": 1.1362,
+      "step": 3387
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816813823099542e-05,
+      "loss": 1.487,
+      "step": 3388
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98166513097978e-05,
+      "loss": 1.1052,
+      "step": 3389
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981648872510828e-05,
+      "loss": 1.1009,
+      "step": 3390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816326069032165e-05,
+      "loss": 1.0057,
+      "step": 3391
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816163341570634e-05,
+      "loss": 1.2086,
+      "step": 3392
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816000542724877e-05,
+      "loss": 1.0492,
+      "step": 3393
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981583767249607e-05,
+      "loss": 1.1245,
+      "step": 3394
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815674730885404e-05,
+      "loss": 1.4041,
+      "step": 3395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981551171789406e-05,
+      "loss": 0.6742,
+      "step": 3396
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815348633523227e-05,
+      "loss": 0.9539,
+      "step": 3397
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981518547777409e-05,
+      "loss": 1.1285,
+      "step": 3398
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981502225064783e-05,
+      "loss": 0.7578,
+      "step": 3399
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814858952145637e-05,
+      "loss": 0.9209,
+      "step": 3400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814695582268705e-05,
+      "loss": 1.1171,
+      "step": 3401
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814532141018215e-05,
+      "loss": 1.1024,
+      "step": 3402
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814368628395357e-05,
+      "loss": 1.1763,
+      "step": 3403
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981420504440132e-05,
+      "loss": 0.949,
+      "step": 3404
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814041389037292e-05,
+      "loss": 1.3421,
+      "step": 3405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813877662304467e-05,
+      "loss": 0.9712,
+      "step": 3406
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813713864204035e-05,
+      "loss": 1.3015,
+      "step": 3407
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813549994737187e-05,
+      "loss": 1.1753,
+      "step": 3408
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813386053905113e-05,
+      "loss": 1.0394,
+      "step": 3409
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813222041709002e-05,
+      "loss": 1.0953,
+      "step": 3410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813057958150054e-05,
+      "loss": 1.2553,
+      "step": 3411
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981289380322946e-05,
+      "loss": 1.2553,
+      "step": 3412
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981272957694841e-05,
+      "loss": 1.1247,
+      "step": 3413
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812565279308103e-05,
+      "loss": 0.9901,
+      "step": 3414
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981240091030973e-05,
+      "loss": 1.2896,
+      "step": 3415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812236469954487e-05,
+      "loss": 1.1382,
+      "step": 3416
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812071958243573e-05,
+      "loss": 1.0717,
+      "step": 3417
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811907375178182e-05,
+      "loss": 1.088,
+      "step": 3418
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811742720759512e-05,
+      "loss": 0.7935,
+      "step": 3419
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811577994988755e-05,
+      "loss": 1.5049,
+      "step": 3420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811413197867117e-05,
+      "loss": 1.0532,
+      "step": 3421
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981124832939579e-05,
+      "loss": 1.3346,
+      "step": 3422
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981108338957598e-05,
+      "loss": 0.9757,
+      "step": 3423
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810918378408875e-05,
+      "loss": 1.237,
+      "step": 3424
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810753295895688e-05,
+      "loss": 1.2839,
+      "step": 3425
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810588142037613e-05,
+      "loss": 0.8444,
+      "step": 3426
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981042291683585e-05,
+      "loss": 0.7821,
+      "step": 3427
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98102576202916e-05,
+      "loss": 0.983,
+      "step": 3428
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981009225240607e-05,
+      "loss": 1.4233,
+      "step": 3429
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980992681318046e-05,
+      "loss": 0.989,
+      "step": 3430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809761302615976e-05,
+      "loss": 1.0199,
+      "step": 3431
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809595720713813e-05,
+      "loss": 1.0441,
+      "step": 3432
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809430067475182e-05,
+      "loss": 1.0975,
+      "step": 3433
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809264342901287e-05,
+      "loss": 1.066,
+      "step": 3434
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809098546993333e-05,
+      "loss": 1.1068,
+      "step": 3435
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808932679752523e-05,
+      "loss": 1.2469,
+      "step": 3436
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980876674118007e-05,
+      "loss": 1.2956,
+      "step": 3437
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980860073127717e-05,
+      "loss": 1.123,
+      "step": 3438
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808434650045042e-05,
+      "loss": 1.1033,
+      "step": 3439
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980826849748488e-05,
+      "loss": 0.9879,
+      "step": 3440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808102273597906e-05,
+      "loss": 1.0205,
+      "step": 3441
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807935978385324e-05,
+      "loss": 1.0261,
+      "step": 3442
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980776961184834e-05,
+      "loss": 1.2471,
+      "step": 3443
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980760317398817e-05,
+      "loss": 1.0463,
+      "step": 3444
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807436664806014e-05,
+      "loss": 1.2007,
+      "step": 3445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807270084303093e-05,
+      "loss": 1.1842,
+      "step": 3446
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807103432480613e-05,
+      "loss": 1.0419,
+      "step": 3447
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806936709339793e-05,
+      "loss": 0.8352,
+      "step": 3448
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806769914881835e-05,
+      "loss": 1.0141,
+      "step": 3449
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980660304910796e-05,
+      "loss": 1.1152,
+      "step": 3450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806436112019378e-05,
+      "loss": 1.0189,
+      "step": 3451
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806269103617302e-05,
+      "loss": 1.0278,
+      "step": 3452
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980610202390295e-05,
+      "loss": 1.452,
+      "step": 3453
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805934872877534e-05,
+      "loss": 0.9768,
+      "step": 3454
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805767650542275e-05,
+      "loss": 1.0699,
+      "step": 3455
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805600356898378e-05,
+      "loss": 1.2197,
+      "step": 3456
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980543299194707e-05,
+      "loss": 1.0581,
+      "step": 3457
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805265555689564e-05,
+      "loss": 1.1719,
+      "step": 3458
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980509804812708e-05,
+      "loss": 0.9098,
+      "step": 3459
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980493046926083e-05,
+      "loss": 1.3231,
+      "step": 3460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804762819092037e-05,
+      "loss": 0.9326,
+      "step": 3461
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804595097621922e-05,
+      "loss": 1.1016,
+      "step": 3462
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804427304851705e-05,
+      "loss": 1.056,
+      "step": 3463
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804259440782597e-05,
+      "loss": 1.5471,
+      "step": 3464
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804091505415833e-05,
+      "loss": 1.1145,
+      "step": 3465
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803923498752625e-05,
+      "loss": 0.9741,
+      "step": 3466
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803755420794193e-05,
+      "loss": 1.217,
+      "step": 3467
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803587271541764e-05,
+      "loss": 1.3019,
+      "step": 3468
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803419050996563e-05,
+      "loss": 1.1441,
+      "step": 3469
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803250759159805e-05,
+      "loss": 1.0492,
+      "step": 3470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803082396032725e-05,
+      "loss": 0.7752,
+      "step": 3471
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802913961616537e-05,
+      "loss": 0.845,
+      "step": 3472
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980274545591247e-05,
+      "loss": 1.0741,
+      "step": 3473
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802576878921752e-05,
+      "loss": 0.8303,
+      "step": 3474
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802408230645605e-05,
+      "loss": 1.0152,
+      "step": 3475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802239511085255e-05,
+      "loss": 1.5367,
+      "step": 3476
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802070720241937e-05,
+      "loss": 1.0408,
+      "step": 3477
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801901858116865e-05,
+      "loss": 1.2921,
+      "step": 3478
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801732924711277e-05,
+      "loss": 0.9487,
+      "step": 3479
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98015639200264e-05,
+      "loss": 1.3043,
+      "step": 3480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801394844063457e-05,
+      "loss": 1.3217,
+      "step": 3481
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801225696823686e-05,
+      "loss": 1.3754,
+      "step": 3482
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801056478308313e-05,
+      "loss": 1.2964,
+      "step": 3483
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800887188518573e-05,
+      "loss": 1.3549,
+      "step": 3484
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800717827455686e-05,
+      "loss": 1.2077,
+      "step": 3485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800548395120893e-05,
+      "loss": 0.7088,
+      "step": 3486
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800378891515425e-05,
+      "loss": 0.9584,
+      "step": 3487
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800209316640515e-05,
+      "loss": 0.8436,
+      "step": 3488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980003967049739e-05,
+      "loss": 1.1537,
+      "step": 3489
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979986995308729e-05,
+      "loss": 1.0506,
+      "step": 3490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979970016441145e-05,
+      "loss": 1.0772,
+      "step": 3491
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97995303044711e-05,
+      "loss": 1.2072,
+      "step": 3492
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799360373267477e-05,
+      "loss": 1.2912,
+      "step": 3493
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799190370801816e-05,
+      "loss": 1.2008,
+      "step": 3494
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979902029707536e-05,
+      "loss": 1.0174,
+      "step": 3495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798850152089334e-05,
+      "loss": 1.0577,
+      "step": 3496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798679935844985e-05,
+      "loss": 1.0689,
+      "step": 3497
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798509648343546e-05,
+      "loss": 1.1688,
+      "step": 3498
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798339289586253e-05,
+      "loss": 1.0534,
+      "step": 3499
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798168859574356e-05,
+      "loss": 1.2939,
+      "step": 3500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797998358309083e-05,
+      "loss": 1.2545,
+      "step": 3501
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979782778579168e-05,
+      "loss": 1.096,
+      "step": 3502
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797657142023384e-05,
+      "loss": 1.0812,
+      "step": 3503
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797486427005436e-05,
+      "loss": 0.9871,
+      "step": 3504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797315640739078e-05,
+      "loss": 1.2509,
+      "step": 3505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797144783225553e-05,
+      "loss": 1.226,
+      "step": 3506
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979697385446611e-05,
+      "loss": 1.1284,
+      "step": 3507
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796802854461977e-05,
+      "loss": 1.1801,
+      "step": 3508
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979663178321441e-05,
+      "loss": 0.9116,
+      "step": 3509
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796460640724646e-05,
+      "loss": 1.3064,
+      "step": 3510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796289426993936e-05,
+      "loss": 1.0387,
+      "step": 3511
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796118142023516e-05,
+      "loss": 0.7129,
+      "step": 3512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979594678581464e-05,
+      "loss": 0.9736,
+      "step": 3513
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795775358368553e-05,
+      "loss": 1.5374,
+      "step": 3514
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795603859686497e-05,
+      "loss": 1.0547,
+      "step": 3515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979543228976973e-05,
+      "loss": 1.2116,
+      "step": 3516
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979526064861948e-05,
+      "loss": 1.1544,
+      "step": 3517
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795088936237013e-05,
+      "loss": 1.4156,
+      "step": 3518
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979491715262357e-05,
+      "loss": 0.924,
+      "step": 3519
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97947452977804e-05,
+      "loss": 0.8578,
+      "step": 3520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979457337170876e-05,
+      "loss": 1.053,
+      "step": 3521
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979440137440989e-05,
+      "loss": 1.0855,
+      "step": 3522
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794229305885048e-05,
+      "loss": 1.1668,
+      "step": 3523
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979405716613548e-05,
+      "loss": 0.9008,
+      "step": 3524
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793884955162442e-05,
+      "loss": 1.2242,
+      "step": 3525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793712672967188e-05,
+      "loss": 1.3489,
+      "step": 3526
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793540319550964e-05,
+      "loss": 1.0186,
+      "step": 3527
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793367894915028e-05,
+      "loss": 1.09,
+      "step": 3528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793195399060635e-05,
+      "loss": 1.0454,
+      "step": 3529
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793022831989034e-05,
+      "loss": 1.1405,
+      "step": 3530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792850193701488e-05,
+      "loss": 1.0062,
+      "step": 3531
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792677484199244e-05,
+      "loss": 1.1369,
+      "step": 3532
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792504703483565e-05,
+      "loss": 1.6285,
+      "step": 3533
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97923318515557e-05,
+      "loss": 1.0827,
+      "step": 3534
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792158928416915e-05,
+      "loss": 1.0711,
+      "step": 3535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979198593406846e-05,
+      "loss": 0.8392,
+      "step": 3536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791812868511597e-05,
+      "loss": 1.2121,
+      "step": 3537
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791639731747584e-05,
+      "loss": 0.9809,
+      "step": 3538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791466523777682e-05,
+      "loss": 1.1701,
+      "step": 3539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979129324460314e-05,
+      "loss": 1.1296,
+      "step": 3540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979111989422523e-05,
+      "loss": 1.0331,
+      "step": 3541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790946472645214e-05,
+      "loss": 1.1408,
+      "step": 3542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790772979864343e-05,
+      "loss": 0.9103,
+      "step": 3543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790599415883887e-05,
+      "loss": 1.0654,
+      "step": 3544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790425780705102e-05,
+      "loss": 1.1242,
+      "step": 3545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790252074329253e-05,
+      "loss": 1.2527,
+      "step": 3546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790078296757605e-05,
+      "loss": 0.8896,
+      "step": 3547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789904447991422e-05,
+      "loss": 1.2544,
+      "step": 3548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789730528031964e-05,
+      "loss": 1.0299,
+      "step": 3549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97895565368805e-05,
+      "loss": 1.0505,
+      "step": 3550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789382474538294e-05,
+      "loss": 1.1504,
+      "step": 3551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978920834100661e-05,
+      "loss": 1.3921,
+      "step": 3552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789034136286716e-05,
+      "loss": 1.1305,
+      "step": 3553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978885986037988e-05,
+      "loss": 0.9207,
+      "step": 3554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788685513287368e-05,
+      "loss": 0.9943,
+      "step": 3555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788511095010448e-05,
+      "loss": 1.154,
+      "step": 3556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788336605550386e-05,
+      "loss": 0.8969,
+      "step": 3557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788162044908458e-05,
+      "loss": 0.8707,
+      "step": 3558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787987413085922e-05,
+      "loss": 1.2129,
+      "step": 3559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978781271008406e-05,
+      "loss": 1.0892,
+      "step": 3560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787637935904137e-05,
+      "loss": 0.8901,
+      "step": 3561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978746309054742e-05,
+      "loss": 0.9743,
+      "step": 3562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978728817401519e-05,
+      "loss": 1.0461,
+      "step": 3563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978711318630871e-05,
+      "loss": 1.1002,
+      "step": 3564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786938127429255e-05,
+      "loss": 1.0547,
+      "step": 3565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786762997378098e-05,
+      "loss": 1.1544,
+      "step": 3566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978658779615652e-05,
+      "loss": 1.2448,
+      "step": 3567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786412523765784e-05,
+      "loss": 1.1764,
+      "step": 3568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978623718020717e-05,
+      "loss": 0.9303,
+      "step": 3569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786061765481954e-05,
+      "loss": 0.9468,
+      "step": 3570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978588627959141e-05,
+      "loss": 1.0761,
+      "step": 3571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785710722536814e-05,
+      "loss": 1.4667,
+      "step": 3572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978553509431944e-05,
+      "loss": 0.9462,
+      "step": 3573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785359394940574e-05,
+      "loss": 1.1817,
+      "step": 3574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978518362440148e-05,
+      "loss": 1.6134,
+      "step": 3575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978500778270345e-05,
+      "loss": 0.9477,
+      "step": 3576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784831869847755e-05,
+      "loss": 1.3771,
+      "step": 3577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784655885835676e-05,
+      "loss": 0.9578,
+      "step": 3578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784479830668494e-05,
+      "loss": 0.8731,
+      "step": 3579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784303704347487e-05,
+      "loss": 1.4237,
+      "step": 3580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784127506873938e-05,
+      "loss": 1.0727,
+      "step": 3581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783951238249126e-05,
+      "loss": 1.5226,
+      "step": 3582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783774898474336e-05,
+      "loss": 1.0812,
+      "step": 3583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783598487550846e-05,
+      "loss": 1.041,
+      "step": 3584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783422005479942e-05,
+      "loss": 0.9924,
+      "step": 3585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783245452262905e-05,
+      "loss": 1.1829,
+      "step": 3586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783068827901024e-05,
+      "loss": 0.7341,
+      "step": 3587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782892132395578e-05,
+      "loss": 1.2945,
+      "step": 3588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782715365747858e-05,
+      "loss": 1.2274,
+      "step": 3589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782538527959142e-05,
+      "loss": 1.406,
+      "step": 3590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782361619030718e-05,
+      "loss": 0.8313,
+      "step": 3591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782184638963876e-05,
+      "loss": 1.0627,
+      "step": 3592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782007587759903e-05,
+      "loss": 1.0787,
+      "step": 3593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978183046542008e-05,
+      "loss": 1.1352,
+      "step": 3594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781653271945703e-05,
+      "loss": 0.9735,
+      "step": 3595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781476007338058e-05,
+      "loss": 0.8053,
+      "step": 3596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781298671598432e-05,
+      "loss": 1.342,
+      "step": 3597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781121264728113e-05,
+      "loss": 1.1296,
+      "step": 3598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97809437867284e-05,
+      "loss": 0.7812,
+      "step": 3599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780766237600574e-05,
+      "loss": 0.9757,
+      "step": 3600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978058861734593e-05,
+      "loss": 1.0099,
+      "step": 3601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780410925965757e-05,
+      "loss": 1.03,
+      "step": 3602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780233163461354e-05,
+      "loss": 1.089,
+      "step": 3603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780055329834007e-05,
+      "loss": 0.8258,
+      "step": 3604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779877425085013e-05,
+      "loss": 1.2677,
+      "step": 3605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779699449215664e-05,
+      "loss": 0.8841,
+      "step": 3606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779521402227253e-05,
+      "loss": 1.265,
+      "step": 3607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779343284121078e-05,
+      "loss": 0.8503,
+      "step": 3608
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977916509489843e-05,
+      "loss": 0.9997,
+      "step": 3609
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977898683456061e-05,
+      "loss": 1.009,
+      "step": 3610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977880850310891e-05,
+      "loss": 1.1258,
+      "step": 3611
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977863010054463e-05,
+      "loss": 1.078,
+      "step": 3612
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778451626869067e-05,
+      "loss": 0.787,
+      "step": 3613
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778273082083516e-05,
+      "loss": 0.9382,
+      "step": 3614
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977809446618928e-05,
+      "loss": 1.1066,
+      "step": 3615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977791577918765e-05,
+      "loss": 1.8857,
+      "step": 3616
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777737021079936e-05,
+      "loss": 0.8721,
+      "step": 3617
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777558191867434e-05,
+      "loss": 1.2001,
+      "step": 3618
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777379291551435e-05,
+      "loss": 1.2355,
+      "step": 3619
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777200320133254e-05,
+      "loss": 1.2468,
+      "step": 3620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777021277614186e-05,
+      "loss": 0.8491,
+      "step": 3621
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977684216399553e-05,
+      "loss": 1.7395,
+      "step": 3622
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776662979278595e-05,
+      "loss": 1.1704,
+      "step": 3623
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776483723464684e-05,
+      "loss": 0.9747,
+      "step": 3624
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776304396555096e-05,
+      "loss": 1.1626,
+      "step": 3625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977612499855113e-05,
+      "loss": 0.8601,
+      "step": 3626
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775945529454106e-05,
+      "loss": 0.8283,
+      "step": 3627
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775765989265316e-05,
+      "loss": 0.8817,
+      "step": 3628
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775586377986073e-05,
+      "loss": 1.1913,
+      "step": 3629
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775406695617677e-05,
+      "loss": 1.0444,
+      "step": 3630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775226942161443e-05,
+      "loss": 0.9599,
+      "step": 3631
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775047117618668e-05,
+      "loss": 1.1193,
+      "step": 3632
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977486722199067e-05,
+      "loss": 1.068,
+      "step": 3633
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774687255278746e-05,
+      "loss": 1.2028,
+      "step": 3634
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774507217484214e-05,
+      "loss": 1.0255,
+      "step": 3635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977432710860838e-05,
+      "loss": 0.9,
+      "step": 3636
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774146928652554e-05,
+      "loss": 1.0745,
+      "step": 3637
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773966677618048e-05,
+      "loss": 1.615,
+      "step": 3638
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773786355506167e-05,
+      "loss": 0.7445,
+      "step": 3639
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773605962318228e-05,
+      "loss": 1.1608,
+      "step": 3640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977342549805554e-05,
+      "loss": 1.1613,
+      "step": 3641
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977324496271942e-05,
+      "loss": 1.2048,
+      "step": 3642
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773064356311172e-05,
+      "loss": 0.8304,
+      "step": 3643
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977288367883212e-05,
+      "loss": 1.1642,
+      "step": 3644
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977270293028357e-05,
+      "loss": 1.0336,
+      "step": 3645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772522110666838e-05,
+      "loss": 1.1802,
+      "step": 3646
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977234121998324e-05,
+      "loss": 1.3209,
+      "step": 3647
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772160258234095e-05,
+      "loss": 1.2375,
+      "step": 3648
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771979225420715e-05,
+      "loss": 1.059,
+      "step": 3649
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771798121544414e-05,
+      "loss": 1.2025,
+      "step": 3650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771616946606515e-05,
+      "loss": 0.9332,
+      "step": 3651
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771435700608328e-05,
+      "loss": 0.9211,
+      "step": 3652
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771254383551178e-05,
+      "loss": 1.0015,
+      "step": 3653
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771072995436384e-05,
+      "loss": 1.151,
+      "step": 3654
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977089153626526e-05,
+      "loss": 1.3275,
+      "step": 3655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770710006039126e-05,
+      "loss": 1.5438,
+      "step": 3656
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770528404759305e-05,
+      "loss": 0.8157,
+      "step": 3657
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770346732427115e-05,
+      "loss": 1.2891,
+      "step": 3658
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977016498904388e-05,
+      "loss": 1.0091,
+      "step": 3659
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769983174610918e-05,
+      "loss": 0.8342,
+      "step": 3660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769801289129558e-05,
+      "loss": 1.0024,
+      "step": 3661
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769619332601114e-05,
+      "loss": 0.9619,
+      "step": 3662
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769437305026917e-05,
+      "loss": 0.9656,
+      "step": 3663
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769255206408283e-05,
+      "loss": 0.9567,
+      "step": 3664
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769073036746542e-05,
+      "loss": 1.1468,
+      "step": 3665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976889079604302e-05,
+      "loss": 1.2042,
+      "step": 3666
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768708484299034e-05,
+      "loss": 1.0037,
+      "step": 3667
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976852610151592e-05,
+      "loss": 0.8387,
+      "step": 3668
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768343647694998e-05,
+      "loss": 1.2168,
+      "step": 3669
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768161122837594e-05,
+      "loss": 1.8033,
+      "step": 3670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976797852694504e-05,
+      "loss": 0.6196,
+      "step": 3671
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976779586001866e-05,
+      "loss": 1.2376,
+      "step": 3672
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767613122059785e-05,
+      "loss": 1.3003,
+      "step": 3673
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976743031306974e-05,
+      "loss": 0.8409,
+      "step": 3674
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767247433049858e-05,
+      "loss": 0.9469,
+      "step": 3675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976706448200147e-05,
+      "loss": 2.1718,
+      "step": 3676
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766881459925904e-05,
+      "loss": 0.9767,
+      "step": 3677
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976669836682449e-05,
+      "loss": 0.819,
+      "step": 3678
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976651520269856e-05,
+      "loss": 1.0357,
+      "step": 3679
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976633196754945e-05,
+      "loss": 1.0585,
+      "step": 3680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976614866137849e-05,
+      "loss": 2.7406,
+      "step": 3681
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765965284187012e-05,
+      "loss": 1.2012,
+      "step": 3682
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765781835976346e-05,
+      "loss": 1.1695,
+      "step": 3683
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765598316747832e-05,
+      "loss": 1.3433,
+      "step": 3684
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765414726502805e-05,
+      "loss": 0.821,
+      "step": 3685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765231065242597e-05,
+      "loss": 0.9543,
+      "step": 3686
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765047332968543e-05,
+      "loss": 0.9901,
+      "step": 3687
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976486352968198e-05,
+      "loss": 1.0539,
+      "step": 3688
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764679655384248e-05,
+      "loss": 0.9762,
+      "step": 3689
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764495710076678e-05,
+      "loss": 0.8897,
+      "step": 3690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976431169376061e-05,
+      "loss": 1.0094,
+      "step": 3691
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764127606437386e-05,
+      "loss": 1.0397,
+      "step": 3692
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763943448108344e-05,
+      "loss": 1.1482,
+      "step": 3693
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763759218774814e-05,
+      "loss": 1.1736,
+      "step": 3694
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763574918438147e-05,
+      "loss": 1.1244,
+      "step": 3695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763390547099676e-05,
+      "loss": 1.7032,
+      "step": 3696
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976320610476075e-05,
+      "loss": 1.3206,
+      "step": 3697
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97630215914227e-05,
+      "loss": 1.1663,
+      "step": 3698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976283700708688e-05,
+      "loss": 1.0135,
+      "step": 3699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762652351754618e-05,
+      "loss": 0.9683,
+      "step": 3700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762467625427264e-05,
+      "loss": 1.2095,
+      "step": 3701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976228282810616e-05,
+      "loss": 0.8941,
+      "step": 3702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976209795979266e-05,
+      "loss": 1.0938,
+      "step": 3703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761913020488094e-05,
+      "loss": 0.7988,
+      "step": 3704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761728010193812e-05,
+      "loss": 0.9712,
+      "step": 3705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976154292891116e-05,
+      "loss": 2.221,
+      "step": 3706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761357776641486e-05,
+      "loss": 1.5021,
+      "step": 3707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976117255338613e-05,
+      "loss": 1.1793,
+      "step": 3708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760987259146448e-05,
+      "loss": 0.9942,
+      "step": 3709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976080189392378e-05,
+      "loss": 1.0424,
+      "step": 3710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760616457719477e-05,
+      "loss": 0.7866,
+      "step": 3711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760430950534885e-05,
+      "loss": 1.0844,
+      "step": 3712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976024537237136e-05,
+      "loss": 1.1818,
+      "step": 3713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760059723230244e-05,
+      "loss": 1.6191,
+      "step": 3714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759874003112893e-05,
+      "loss": 1.1241,
+      "step": 3715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975968821202065e-05,
+      "loss": 0.942,
+      "step": 3716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759502349954872e-05,
+      "loss": 0.8905,
+      "step": 3717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975931641691691e-05,
+      "loss": 1.3947,
+      "step": 3718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759130412908115e-05,
+      "loss": 0.986,
+      "step": 3719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975894433792984e-05,
+      "loss": 0.8201,
+      "step": 3720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975875819198344e-05,
+      "loss": 1.3644,
+      "step": 3721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758571975070263e-05,
+      "loss": 0.8535,
+      "step": 3722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758385687191666e-05,
+      "loss": 0.5739,
+      "step": 3723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975819932834901e-05,
+      "loss": 1.0124,
+      "step": 3724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975801289854364e-05,
+      "loss": 1.0918,
+      "step": 3725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975782639777692e-05,
+      "loss": 1.4131,
+      "step": 3726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757639826050202e-05,
+      "loss": 1.0741,
+      "step": 3727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757453183364842e-05,
+      "loss": 1.067,
+      "step": 3728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97572664697222e-05,
+      "loss": 1.0624,
+      "step": 3729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757079685123633e-05,
+      "loss": 0.9061,
+      "step": 3730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97568928295705e-05,
+      "loss": 0.7532,
+      "step": 3731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756705903064153e-05,
+      "loss": 1.0176,
+      "step": 3732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756518905605966e-05,
+      "loss": 0.9539,
+      "step": 3733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756331837197286e-05,
+      "loss": 0.7839,
+      "step": 3734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756144697839477e-05,
+      "loss": 1.18,
+      "step": 3735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97559574875339e-05,
+      "loss": 1.1383,
+      "step": 3736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755770206281917e-05,
+      "loss": 0.9792,
+      "step": 3737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755582854084888e-05,
+      "loss": 1.2026,
+      "step": 3738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975539543094418e-05,
+      "loss": 1.1407,
+      "step": 3739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755207936861157e-05,
+      "loss": 1.7168,
+      "step": 3740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755020371837172e-05,
+      "loss": 1.281,
+      "step": 3741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97548327358736e-05,
+      "loss": 1.0115,
+      "step": 3742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97546450289718e-05,
+      "loss": 1.0695,
+      "step": 3743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754457251133134e-05,
+      "loss": 0.9575,
+      "step": 3744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754269402358973e-05,
+      "loss": 0.764,
+      "step": 3745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754081482650685e-05,
+      "loss": 1.1657,
+      "step": 3746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975389349200963e-05,
+      "loss": 1.0728,
+      "step": 3747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975370543043718e-05,
+      "loss": 1.1956,
+      "step": 3748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97535172979347e-05,
+      "loss": 1.3663,
+      "step": 3749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753329094503563e-05,
+      "loss": 0.9443,
+      "step": 3750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753140820145128e-05,
+      "loss": 1.2239,
+      "step": 3751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752952474860772e-05,
+      "loss": 1.0926,
+      "step": 3752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752764058651865e-05,
+      "loss": 0.9475,
+      "step": 3753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752575571519773e-05,
+      "loss": 1.0247,
+      "step": 3754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752387013465867e-05,
+      "loss": 1.1216,
+      "step": 3755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752198384491523e-05,
+      "loss": 1.0492,
+      "step": 3756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752009684598108e-05,
+      "loss": 1.1192,
+      "step": 3757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751820913786995e-05,
+      "loss": 1.1294,
+      "step": 3758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751632072059556e-05,
+      "loss": 0.9614,
+      "step": 3759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751443159417167e-05,
+      "loss": 1.1473,
+      "step": 3760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751254175861203e-05,
+      "loss": 0.852,
+      "step": 3761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751065121393034e-05,
+      "loss": 1.0694,
+      "step": 3762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750875996014034e-05,
+      "loss": 1.4437,
+      "step": 3763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750686799725583e-05,
+      "loss": 1.3928,
+      "step": 3764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750497532529053e-05,
+      "loss": 0.7406,
+      "step": 3765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750308194425827e-05,
+      "loss": 1.0891,
+      "step": 3766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750118785417275e-05,
+      "loss": 1.3824,
+      "step": 3767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749929305504775e-05,
+      "loss": 1.2085,
+      "step": 3768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749739754689705e-05,
+      "loss": 1.1228,
+      "step": 3769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749550132973447e-05,
+      "loss": 1.2979,
+      "step": 3770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749360440357377e-05,
+      "loss": 1.1768,
+      "step": 3771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749170676842878e-05,
+      "loss": 0.8423,
+      "step": 3772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974898084243132e-05,
+      "loss": 1.4177,
+      "step": 3773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97487909371241e-05,
+      "loss": 1.1304,
+      "step": 3774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748600960922584e-05,
+      "loss": 1.0124,
+      "step": 3775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748410913828163e-05,
+      "loss": 1.2393,
+      "step": 3776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748220795842213e-05,
+      "loss": 1.0387,
+      "step": 3777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748030606966117e-05,
+      "loss": 1.3799,
+      "step": 3778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747840347201262e-05,
+      "loss": 1.0891,
+      "step": 3779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974765001654903e-05,
+      "loss": 1.3246,
+      "step": 3780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747459615010804e-05,
+      "loss": 0.7431,
+      "step": 3781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974726914258797e-05,
+      "loss": 1.2962,
+      "step": 3782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974707859928191e-05,
+      "loss": 1.3408,
+      "step": 3783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746887985094015e-05,
+      "loss": 0.9265,
+      "step": 3784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746697300025667e-05,
+      "loss": 1.0807,
+      "step": 3785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746506544078252e-05,
+      "loss": 0.9307,
+      "step": 3786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746315717253163e-05,
+      "loss": 0.9558,
+      "step": 3787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974612481955178e-05,
+      "loss": 1.0913,
+      "step": 3788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745933850975496e-05,
+      "loss": 1.1036,
+      "step": 3789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97457428115257e-05,
+      "loss": 1.2691,
+      "step": 3790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974555170120378e-05,
+      "loss": 0.9968,
+      "step": 3791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745360520011124e-05,
+      "loss": 1.0304,
+      "step": 3792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745169267949126e-05,
+      "loss": 1.1366,
+      "step": 3793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744977945019174e-05,
+      "loss": 0.9867,
+      "step": 3794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744786551222658e-05,
+      "loss": 1.1062,
+      "step": 3795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744595086560977e-05,
+      "loss": 0.8407,
+      "step": 3796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744403551035517e-05,
+      "loss": 1.237,
+      "step": 3797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974421194464767e-05,
+      "loss": 1.1745,
+      "step": 3798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744020267398832e-05,
+      "loss": 0.97,
+      "step": 3799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743828519290396e-05,
+      "loss": 1.0668,
+      "step": 3800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974363670032376e-05,
+      "loss": 1.0208,
+      "step": 3801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743444810500313e-05,
+      "loss": 1.2813,
+      "step": 3802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743252849821458e-05,
+      "loss": 1.3111,
+      "step": 3803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974306081828858e-05,
+      "loss": 0.6652,
+      "step": 3804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974286871590309e-05,
+      "loss": 1.0083,
+      "step": 3805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742676542666366e-05,
+      "loss": 1.4691,
+      "step": 3806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742484298579824e-05,
+      "loss": 1.2286,
+      "step": 3807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974229198364485e-05,
+      "loss": 0.9704,
+      "step": 3808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974209959786285e-05,
+      "loss": 1.1509,
+      "step": 3809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974190714123522e-05,
+      "loss": 1.0084,
+      "step": 3810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974171461376336e-05,
+      "loss": 0.7886,
+      "step": 3811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741522015448665e-05,
+      "loss": 0.803,
+      "step": 3812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741329346292547e-05,
+      "loss": 0.9632,
+      "step": 3813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741136606296394e-05,
+      "loss": 1.2124,
+      "step": 3814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740943795461614e-05,
+      "loss": 0.9622,
+      "step": 3815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740750913789612e-05,
+      "loss": 0.9185,
+      "step": 3816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974055796128179e-05,
+      "loss": 0.916,
+      "step": 3817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740364937939542e-05,
+      "loss": 0.9758,
+      "step": 3818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740171843764283e-05,
+      "loss": 1.1435,
+      "step": 3819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973997867875741e-05,
+      "loss": 1.1029,
+      "step": 3820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739785442920338e-05,
+      "loss": 1.01,
+      "step": 3821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739592136254456e-05,
+      "loss": 1.3318,
+      "step": 3822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973939875876118e-05,
+      "loss": 0.9304,
+      "step": 3823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739205310441918e-05,
+      "loss": 1.2548,
+      "step": 3824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739011791298073e-05,
+      "loss": 0.7669,
+      "step": 3825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973881820133105e-05,
+      "loss": 1.1612,
+      "step": 3826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973862454054226e-05,
+      "loss": 1.0716,
+      "step": 3827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738430808933113e-05,
+      "loss": 1.0124,
+      "step": 3828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738237006505015e-05,
+      "loss": 1.1211,
+      "step": 3829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738043133259374e-05,
+      "loss": 0.9265,
+      "step": 3830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737849189197606e-05,
+      "loss": 1.0385,
+      "step": 3831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737655174321115e-05,
+      "loss": 1.3053,
+      "step": 3832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737461088631314e-05,
+      "loss": 1.0267,
+      "step": 3833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737266932129616e-05,
+      "loss": 1.1479,
+      "step": 3834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737072704817428e-05,
+      "loss": 0.7312,
+      "step": 3835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973687840669617e-05,
+      "loss": 0.871,
+      "step": 3836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736684037767247e-05,
+      "loss": 1.1406,
+      "step": 3837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973648959803208e-05,
+      "loss": 1.0024,
+      "step": 3838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736295087492075e-05,
+      "loss": 1.1229,
+      "step": 3839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736100506148657e-05,
+      "loss": 1.0899,
+      "step": 3840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973590585400323e-05,
+      "loss": 0.8788,
+      "step": 3841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735711131057222e-05,
+      "loss": 1.2138,
+      "step": 3842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973551633731203e-05,
+      "loss": 1.2193,
+      "step": 3843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973532147276909e-05,
+      "loss": 1.1963,
+      "step": 3844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735126537429813e-05,
+      "loss": 1.0251,
+      "step": 3845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973493153129561e-05,
+      "loss": 0.9122,
+      "step": 3846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734736454367905e-05,
+      "loss": 0.9584,
+      "step": 3847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734541306648113e-05,
+      "loss": 1.1112,
+      "step": 3848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973434608813766e-05,
+      "loss": 1.2209,
+      "step": 3849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973415079883796e-05,
+      "loss": 1.1551,
+      "step": 3850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733955438750434e-05,
+      "loss": 0.914,
+      "step": 3851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97337600078765e-05,
+      "loss": 1.2396,
+      "step": 3852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733564506217588e-05,
+      "loss": 1.1188,
+      "step": 3853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973336893377511e-05,
+      "loss": 1.4181,
+      "step": 3854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733173290550494e-05,
+      "loss": 0.8641,
+      "step": 3855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732977576545162e-05,
+      "loss": 1.0075,
+      "step": 3856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732781791760534e-05,
+      "loss": 1.1743,
+      "step": 3857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732585936198036e-05,
+      "loss": 0.7981,
+      "step": 3858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973239000985909e-05,
+      "loss": 1.1231,
+      "step": 3859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973219401274513e-05,
+      "loss": 1.3923,
+      "step": 3860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973199794485757e-05,
+      "loss": 1.0707,
+      "step": 3861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973180180619784e-05,
+      "loss": 1.0829,
+      "step": 3862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973160559676737e-05,
+      "loss": 1.1779,
+      "step": 3863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973140931656758e-05,
+      "loss": 1.0851,
+      "step": 3864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731212965599902e-05,
+      "loss": 1.117,
+      "step": 3865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731016543865763e-05,
+      "loss": 0.9253,
+      "step": 3866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973082005136659e-05,
+      "loss": 1.479,
+      "step": 3867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730623488103813e-05,
+      "loss": 1.2305,
+      "step": 3868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973042685407886e-05,
+      "loss": 0.8478,
+      "step": 3869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730230149293167e-05,
+      "loss": 1.3518,
+      "step": 3870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730033373748158e-05,
+      "loss": 1.4068,
+      "step": 3871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729836527445262e-05,
+      "loss": 1.2203,
+      "step": 3872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729639610385916e-05,
+      "loss": 1.1326,
+      "step": 3873
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729442622571553e-05,
+      "loss": 1.1008,
+      "step": 3874
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97292455640036e-05,
+      "loss": 0.9983,
+      "step": 3875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729048434683497e-05,
+      "loss": 1.0944,
+      "step": 3876
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972885123461267e-05,
+      "loss": 0.8629,
+      "step": 3877
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728653963792557e-05,
+      "loss": 1.2791,
+      "step": 3878
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972845662222459e-05,
+      "loss": 1.1733,
+      "step": 3879
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972825920991021e-05,
+      "loss": 0.772,
+      "step": 3880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728061726850846e-05,
+      "loss": 1.095,
+      "step": 3881
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727864173047936e-05,
+      "loss": 0.8627,
+      "step": 3882
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727666548502922e-05,
+      "loss": 0.9604,
+      "step": 3883
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727468853217236e-05,
+      "loss": 1.0899,
+      "step": 3884
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727271087192312e-05,
+      "loss": 1.144,
+      "step": 3885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727073250429596e-05,
+      "loss": 0.9647,
+      "step": 3886
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726875342930522e-05,
+      "loss": 1.1655,
+      "step": 3887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972667736469653e-05,
+      "loss": 1.0899,
+      "step": 3888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972647931572906e-05,
+      "loss": 1.0657,
+      "step": 3889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726281196029552e-05,
+      "loss": 1.4015,
+      "step": 3890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726083005599448e-05,
+      "loss": 1.1805,
+      "step": 3891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725884744440188e-05,
+      "loss": 1.0575,
+      "step": 3892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725686412553213e-05,
+      "loss": 1.0741,
+      "step": 3893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972548800993997e-05,
+      "loss": 1.1369,
+      "step": 3894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725289536601893e-05,
+      "loss": 1.1895,
+      "step": 3895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725090992540436e-05,
+      "loss": 0.9768,
+      "step": 3896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724892377757033e-05,
+      "loss": 1.0922,
+      "step": 3897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724693692253135e-05,
+      "loss": 1.0965,
+      "step": 3898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972449493603018e-05,
+      "loss": 1.1804,
+      "step": 3899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724296109089623e-05,
+      "loss": 1.1859,
+      "step": 3900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724097211432903e-05,
+      "loss": 1.271,
+      "step": 3901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972389824306147e-05,
+      "loss": 1.3724,
+      "step": 3902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723699203976768e-05,
+      "loss": 1.1276,
+      "step": 3903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723500094180247e-05,
+      "loss": 1.0447,
+      "step": 3904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972330091367335e-05,
+      "loss": 0.7372,
+      "step": 3905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972310166245753e-05,
+      "loss": 1.0522,
+      "step": 3906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722902340534237e-05,
+      "loss": 0.9873,
+      "step": 3907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722702947904917e-05,
+      "loss": 1.0531,
+      "step": 3908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722503484571023e-05,
+      "loss": 1.1534,
+      "step": 3909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722303950534004e-05,
+      "loss": 0.9638,
+      "step": 3910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972210434579531e-05,
+      "loss": 1.1008,
+      "step": 3911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721904670356392e-05,
+      "loss": 1.1686,
+      "step": 3912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721704924218708e-05,
+      "loss": 0.9594,
+      "step": 3913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97215051073837e-05,
+      "loss": 1.094,
+      "step": 3914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721305219852833e-05,
+      "loss": 1.176,
+      "step": 3915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721105261627553e-05,
+      "loss": 1.2844,
+      "step": 3916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720905232709315e-05,
+      "loss": 0.9807,
+      "step": 3917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720705133099575e-05,
+      "loss": 1.0513,
+      "step": 3918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972050496279979e-05,
+      "loss": 1.0016,
+      "step": 3919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972030472181141e-05,
+      "loss": 1.1824,
+      "step": 3920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97201044101359e-05,
+      "loss": 1.2856,
+      "step": 3921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719904027774707e-05,
+      "loss": 0.8707,
+      "step": 3922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719703574729292e-05,
+      "loss": 0.7584,
+      "step": 3923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719503051001116e-05,
+      "loss": 1.1001,
+      "step": 3924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971930245659163e-05,
+      "loss": 1.2113,
+      "step": 3925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719101791502302e-05,
+      "loss": 0.905,
+      "step": 3926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718901055734583e-05,
+      "loss": 0.9644,
+      "step": 3927
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971870024928994e-05,
+      "loss": 1.2982,
+      "step": 3928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718499372169826e-05,
+      "loss": 1.2534,
+      "step": 3929
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971829842437571e-05,
+      "loss": 1.2754,
+      "step": 3930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718097405909042e-05,
+      "loss": 0.9811,
+      "step": 3931
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717896316771297e-05,
+      "loss": 0.8598,
+      "step": 3932
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717695156963925e-05,
+      "loss": 0.7085,
+      "step": 3933
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717493926488398e-05,
+      "loss": 0.824,
+      "step": 3934
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717292625346176e-05,
+      "loss": 1.0619,
+      "step": 3935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717091253538727e-05,
+      "loss": 1.068,
+      "step": 3936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716889811067502e-05,
+      "loss": 1.235,
+      "step": 3937
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971668829793398e-05,
+      "loss": 0.7711,
+      "step": 3938
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716486714139626e-05,
+      "loss": 1.0497,
+      "step": 3939
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716285059685894e-05,
+      "loss": 1.0579,
+      "step": 3940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716083334574263e-05,
+      "loss": 1.2064,
+      "step": 3941
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715881538806193e-05,
+      "loss": 1.2137,
+      "step": 3942
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715679672383154e-05,
+      "loss": 1.2636,
+      "step": 3943
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715477735306613e-05,
+      "loss": 1.255,
+      "step": 3944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971527572757804e-05,
+      "loss": 1.3746,
+      "step": 3945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715073649198905e-05,
+      "loss": 1.0985,
+      "step": 3946
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714871500170672e-05,
+      "loss": 0.8435,
+      "step": 3947
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971466928049482e-05,
+      "loss": 1.2163,
+      "step": 3948
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714466990172807e-05,
+      "loss": 1.4075,
+      "step": 3949
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714264629206116e-05,
+      "loss": 1.155,
+      "step": 3950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971406219759621e-05,
+      "loss": 0.9575,
+      "step": 3951
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713859695344568e-05,
+      "loss": 1.0527,
+      "step": 3952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971365712245266e-05,
+      "loss": 0.969,
+      "step": 3953
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713454478921957e-05,
+      "loss": 1.0425,
+      "step": 3954
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713251764753934e-05,
+      "loss": 1.0781,
+      "step": 3955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713048979950066e-05,
+      "loss": 0.8935,
+      "step": 3956
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712846124511832e-05,
+      "loss": 1.7291,
+      "step": 3957
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712643198440696e-05,
+      "loss": 1.0846,
+      "step": 3958
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712440201738143e-05,
+      "loss": 1.423,
+      "step": 3959
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971223713440564e-05,
+      "loss": 0.9578,
+      "step": 3960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971203399644468e-05,
+      "loss": 0.9963,
+      "step": 3961
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711830787856727e-05,
+      "loss": 0.9308,
+      "step": 3962
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971162750864326e-05,
+      "loss": 1.1067,
+      "step": 3963
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971142415880576e-05,
+      "loss": 0.918,
+      "step": 3964
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711220738345705e-05,
+      "loss": 1.119,
+      "step": 3965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711017247264575e-05,
+      "loss": 0.8899,
+      "step": 3966
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971081368556385e-05,
+      "loss": 0.9461,
+      "step": 3967
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710610053245007e-05,
+      "loss": 1.1835,
+      "step": 3968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710406350309534e-05,
+      "loss": 1.2525,
+      "step": 3969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710202576758905e-05,
+      "loss": 0.993,
+      "step": 3970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709998732594602e-05,
+      "loss": 1.2066,
+      "step": 3971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970979481781811e-05,
+      "loss": 0.8457,
+      "step": 3972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709590832430914e-05,
+      "loss": 1.1716,
+      "step": 3973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97093867764345e-05,
+      "loss": 0.9179,
+      "step": 3974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970918264983034e-05,
+      "loss": 1.0804,
+      "step": 3975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708978452619926e-05,
+      "loss": 1.2587,
+      "step": 3976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708774184804746e-05,
+      "loss": 1.0601,
+      "step": 3977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970856984638628e-05,
+      "loss": 0.9937,
+      "step": 3978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708365437366015e-05,
+      "loss": 1.0799,
+      "step": 3979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708160957745442e-05,
+      "loss": 0.8325,
+      "step": 3980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970795640752604e-05,
+      "loss": 0.9885,
+      "step": 3981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707751786709304e-05,
+      "loss": 0.8781,
+      "step": 3982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707547095296718e-05,
+      "loss": 1.2265,
+      "step": 3983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970734233328977e-05,
+      "loss": 1.3634,
+      "step": 3984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707137500689955e-05,
+      "loss": 0.9293,
+      "step": 3985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706932597498752e-05,
+      "loss": 1.3334,
+      "step": 3986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9706727623717664e-05,
+      "loss": 1.4826,
+      "step": 3987
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970652257934817e-05,
+      "loss": 1.1848,
+      "step": 3988
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706317464391767e-05,
+      "loss": 0.8229,
+      "step": 3989
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970611227884995e-05,
+      "loss": 1.1679,
+      "step": 3990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.97059070227242e-05,
+      "loss": 1.3234,
+      "step": 3991
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970570169601602e-05,
+      "loss": 1.1869,
+      "step": 3992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.97054962987269e-05,
+      "loss": 0.8909,
+      "step": 3993
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970529083085833e-05,
+      "loss": 1.0959,
+      "step": 3994
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970508529241181e-05,
+      "loss": 0.9992,
+      "step": 3995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970487968338883e-05,
+      "loss": 0.799,
+      "step": 3996
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704674003790886e-05,
+      "loss": 0.7662,
+      "step": 3997
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970446825361948e-05,
+      "loss": 1.0075,
+      "step": 3998
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.97042624328761e-05,
+      "loss": 1.0535,
+      "step": 3999
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704056541562244e-05,
+      "loss": 1.319,
+      "step": 4000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703850579679414e-05,
+      "loss": 1.0884,
+      "step": 4001
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703644547229108e-05,
+      "loss": 1.7016,
+      "step": 4002
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703438444212815e-05,
+      "loss": 1.3529,
+      "step": 4003
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703232270632043e-05,
+      "loss": 1.25,
+      "step": 4004
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703026026488288e-05,
+      "loss": 0.9987,
+      "step": 4005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970281971178305e-05,
+      "loss": 0.755,
+      "step": 4006
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970261332651783e-05,
+      "loss": 1.2661,
+      "step": 4007
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970240687069413e-05,
+      "loss": 1.118,
+      "step": 4008
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702200344313447e-05,
+      "loss": 1.0678,
+      "step": 4009
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701993747377288e-05,
+      "loss": 0.9681,
+      "step": 4010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701787079887153e-05,
+      "loss": 1.266,
+      "step": 4011
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970158034184454e-05,
+      "loss": 1.1551,
+      "step": 4012
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701373533250964e-05,
+      "loss": 1.4065,
+      "step": 4013
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970116665410792e-05,
+      "loss": 0.9273,
+      "step": 4014
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700959704416917e-05,
+      "loss": 1.0276,
+      "step": 4015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700752684179453e-05,
+      "loss": 1.1703,
+      "step": 4016
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970054559339704e-05,
+      "loss": 1.2206,
+      "step": 4017
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700338432071184e-05,
+      "loss": 1.1667,
+      "step": 4018
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970013120020339e-05,
+      "loss": 0.8601,
+      "step": 4019
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699923897795165e-05,
+      "loss": 1.076,
+      "step": 4020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699716524848012e-05,
+      "loss": 1.2138,
+      "step": 4021
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699509081363445e-05,
+      "loss": 0.995,
+      "step": 4022
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699301567342975e-05,
+      "loss": 0.7207,
+      "step": 4023
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699093982788104e-05,
+      "loss": 0.8274,
+      "step": 4024
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698886327700344e-05,
+      "loss": 0.9099,
+      "step": 4025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698678602081202e-05,
+      "loss": 0.9419,
+      "step": 4026
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96984708059322e-05,
+      "loss": 1.0342,
+      "step": 4027
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698262939254835e-05,
+      "loss": 0.7484,
+      "step": 4028
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969805500205063e-05,
+      "loss": 1.1752,
+      "step": 4029
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697846994321086e-05,
+      "loss": 1.2253,
+      "step": 4030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697638916067725e-05,
+      "loss": 1.2039,
+      "step": 4031
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697430767292057e-05,
+      "loss": 1.1722,
+      "step": 4032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697222547995597e-05,
+      "loss": 1.0839,
+      "step": 4033
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697014258179857e-05,
+      "loss": 1.1513,
+      "step": 4034
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696805897846353e-05,
+      "loss": 0.988,
+      "step": 4035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696597466996598e-05,
+      "loss": 0.9467,
+      "step": 4036
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696388965632106e-05,
+      "loss": 1.1378,
+      "step": 4037
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696180393754404e-05,
+      "loss": 1.3603,
+      "step": 4038
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695971751364997e-05,
+      "loss": 1.2505,
+      "step": 4039
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969576303846541e-05,
+      "loss": 1.0621,
+      "step": 4040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695554255057154e-05,
+      "loss": 1.1017,
+      "step": 4041
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695345401141753e-05,
+      "loss": 0.9605,
+      "step": 4042
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695136476720723e-05,
+      "loss": 1.032,
+      "step": 4043
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694927481795583e-05,
+      "loss": 1.1783,
+      "step": 4044
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694718416367853e-05,
+      "loss": 0.9864,
+      "step": 4045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694509280439057e-05,
+      "loss": 0.8692,
+      "step": 4046
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694300074010712e-05,
+      "loss": 0.9135,
+      "step": 4047
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969409079708434e-05,
+      "loss": 0.8195,
+      "step": 4048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693881449661463e-05,
+      "loss": 1.3516,
+      "step": 4049
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693672031743604e-05,
+      "loss": 1.3566,
+      "step": 4050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693462543332282e-05,
+      "loss": 1.1929,
+      "step": 4051
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969325298442903e-05,
+      "loss": 1.1781,
+      "step": 4052
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969304335503536e-05,
+      "loss": 1.042,
+      "step": 4053
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692833655152808e-05,
+      "loss": 1.0182,
+      "step": 4054
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969262388478289e-05,
+      "loss": 0.7247,
+      "step": 4055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692414043927135e-05,
+      "loss": 1.2185,
+      "step": 4056
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969220413258707e-05,
+      "loss": 1.347,
+      "step": 4057
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969199415076422e-05,
+      "loss": 0.8743,
+      "step": 4058
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969178409846011e-05,
+      "loss": 1.0225,
+      "step": 4059
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691573975676272e-05,
+      "loss": 1.257,
+      "step": 4060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969136378241423e-05,
+      "loss": 1.2848,
+      "step": 4061
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691153518675516e-05,
+      "loss": 1.2449,
+      "step": 4062
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690943184461657e-05,
+      "loss": 0.8256,
+      "step": 4063
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690732779774183e-05,
+      "loss": 0.8188,
+      "step": 4064
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690522304614624e-05,
+      "loss": 0.9645,
+      "step": 4065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690311758984513e-05,
+      "loss": 1.0799,
+      "step": 4066
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690101142885377e-05,
+      "loss": 0.9255,
+      "step": 4067
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689890456318748e-05,
+      "loss": 1.0093,
+      "step": 4068
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968967969928616e-05,
+      "loss": 1.0356,
+      "step": 4069
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968946887178915e-05,
+      "loss": 0.9179,
+      "step": 4070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968925797382924e-05,
+      "loss": 1.0843,
+      "step": 4071
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968904700540797e-05,
+      "loss": 0.9591,
+      "step": 4072
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968883596652688e-05,
+      "loss": 1.3339,
+      "step": 4073
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688624857187496e-05,
+      "loss": 0.8213,
+      "step": 4074
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688413677391352e-05,
+      "loss": 1.0424,
+      "step": 4075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688202427139994e-05,
+      "loss": 1.0228,
+      "step": 4076
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968799110643495e-05,
+      "loss": 0.9692,
+      "step": 4077
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687779715277756e-05,
+      "loss": 1.0884,
+      "step": 4078
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687568253669956e-05,
+      "loss": 1.1652,
+      "step": 4079
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687356721613084e-05,
+      "loss": 1.0777,
+      "step": 4080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687145119108675e-05,
+      "loss": 0.9439,
+      "step": 4081
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968693344615827e-05,
+      "loss": 1.1852,
+      "step": 4082
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686721702763412e-05,
+      "loss": 1.5023,
+      "step": 4083
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968650988892564e-05,
+      "loss": 1.2333,
+      "step": 4084
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968629800464649e-05,
+      "loss": 1.0257,
+      "step": 4085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686086049927503e-05,
+      "loss": 1.2789,
+      "step": 4086
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685874024770228e-05,
+      "loss": 0.9883,
+      "step": 4087
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685661929176197e-05,
+      "loss": 1.3456,
+      "step": 4088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968544976314696e-05,
+      "loss": 1.0488,
+      "step": 4089
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685237526684057e-05,
+      "loss": 1.0711,
+      "step": 4090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685025219789028e-05,
+      "loss": 1.0961,
+      "step": 4091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684812842463423e-05,
+      "loss": 1.0893,
+      "step": 4092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684600394708783e-05,
+      "loss": 0.8551,
+      "step": 4093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968438787652665e-05,
+      "loss": 1.2237,
+      "step": 4094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684175287918576e-05,
+      "loss": 1.1464,
+      "step": 4095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683962628886103e-05,
+      "loss": 0.943,
+      "step": 4096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968374989943078e-05,
+      "loss": 0.9469,
+      "step": 4097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968353709955415e-05,
+      "loss": 1.1348,
+      "step": 4098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683324229257765e-05,
+      "loss": 1.0953,
+      "step": 4099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968311128854317e-05,
+      "loss": 1.107,
+      "step": 4100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968289827741192e-05,
+      "loss": 0.884,
+      "step": 4101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968268519586555e-05,
+      "loss": 0.8434,
+      "step": 4102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682472043905617e-05,
+      "loss": 1.0806,
+      "step": 4103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968225882153368e-05,
+      "loss": 1.276,
+      "step": 4104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682045528751276e-05,
+      "loss": 1.0312,
+      "step": 4105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681832165559965e-05,
+      "loss": 0.7965,
+      "step": 4106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681618731961293e-05,
+      "loss": 1.0492,
+      "step": 4107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681405227956815e-05,
+      "loss": 1.1845,
+      "step": 4108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9681191653548083e-05,
+      "loss": 0.9595,
+      "step": 4109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968097800873665e-05,
+      "loss": 1.2289,
+      "step": 4110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968076429352407e-05,
+      "loss": 1.0519,
+      "step": 4111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680550507911898e-05,
+      "loss": 1.1245,
+      "step": 4112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968033665190169e-05,
+      "loss": 1.0665,
+      "step": 4113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680122725494993e-05,
+      "loss": 1.1584,
+      "step": 4114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679908728693374e-05,
+      "loss": 1.088,
+      "step": 4115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967969466149838e-05,
+      "loss": 0.9493,
+      "step": 4116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679480523911574e-05,
+      "loss": 1.1117,
+      "step": 4117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679266315934507e-05,
+      "loss": 0.8099,
+      "step": 4118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9679052037568745e-05,
+      "loss": 1.0441,
+      "step": 4119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678837688815843e-05,
+      "loss": 1.1505,
+      "step": 4120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678623269677355e-05,
+      "loss": 0.7566,
+      "step": 4121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678408780154844e-05,
+      "loss": 0.9116,
+      "step": 4122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967819422024987e-05,
+      "loss": 1.0231,
+      "step": 4123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677979589963994e-05,
+      "loss": 1.401,
+      "step": 4124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677764889298775e-05,
+      "loss": 0.89,
+      "step": 4125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967755011825577e-05,
+      "loss": 0.976,
+      "step": 4126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677335276836552e-05,
+      "loss": 1.3226,
+      "step": 4127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677120365042675e-05,
+      "loss": 1.2141,
+      "step": 4128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676905382875706e-05,
+      "loss": 1.3163,
+      "step": 4129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967669033033721e-05,
+      "loss": 1.1328,
+      "step": 4130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967647520742874e-05,
+      "loss": 1.163,
+      "step": 4131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676260014151868e-05,
+      "loss": 1.2208,
+      "step": 4132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676044750508162e-05,
+      "loss": 1.0568,
+      "step": 4133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675829416499183e-05,
+      "loss": 1.012,
+      "step": 4134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96756140121265e-05,
+      "loss": 1.0215,
+      "step": 4135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675398537391674e-05,
+      "loss": 0.9227,
+      "step": 4136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9675182992296277e-05,
+      "loss": 0.7946,
+      "step": 4137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674967376841875e-05,
+      "loss": 0.8677,
+      "step": 4138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674751691030034e-05,
+      "loss": 1.3481,
+      "step": 4139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9674535934862327e-05,
+      "loss": 1.0846,
+      "step": 4140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967432010834032e-05,
+      "loss": 1.0542,
+      "step": 4141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967410421146558e-05,
+      "loss": 1.1609,
+      "step": 4142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673888244239682e-05,
+      "loss": 1.0224,
+      "step": 4143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673672206664195e-05,
+      "loss": 0.9953,
+      "step": 4144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967345609874069e-05,
+      "loss": 1.2325,
+      "step": 4145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673239920470734e-05,
+      "loss": 1.0617,
+      "step": 4146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673023671855908e-05,
+      "loss": 1.1916,
+      "step": 4147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672807352897776e-05,
+      "loss": 1.043,
+      "step": 4148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967259096359792e-05,
+      "loss": 1.1856,
+      "step": 4149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672374503957906e-05,
+      "loss": 1.249,
+      "step": 4150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967215797397931e-05,
+      "loss": 1.2188,
+      "step": 4151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671941373663705e-05,
+      "loss": 0.7497,
+      "step": 4152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967172470301267e-05,
+      "loss": 0.8731,
+      "step": 4153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967150796202778e-05,
+      "loss": 1.0981,
+      "step": 4154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967129115071061e-05,
+      "loss": 0.938,
+      "step": 4155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671074269062734e-05,
+      "loss": 0.94,
+      "step": 4156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670857317085734e-05,
+      "loss": 1.1197,
+      "step": 4157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670640294781185e-05,
+      "loss": 1.0376,
+      "step": 4158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670423202150664e-05,
+      "loss": 0.8634,
+      "step": 4159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670206039195755e-05,
+      "loss": 0.8965,
+      "step": 4160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669988805918036e-05,
+      "loss": 1.0647,
+      "step": 4161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966977150231908e-05,
+      "loss": 1.3223,
+      "step": 4162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966955412840047e-05,
+      "loss": 0.9942,
+      "step": 4163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669336684163792e-05,
+      "loss": 1.3473,
+      "step": 4164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669119169610626e-05,
+      "loss": 0.7373,
+      "step": 4165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668901584742548e-05,
+      "loss": 1.0608,
+      "step": 4166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668683929561146e-05,
+      "loss": 1.2456,
+      "step": 4167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668466204068e-05,
+      "loss": 0.9628,
+      "step": 4168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668248408264694e-05,
+      "loss": 1.1711,
+      "step": 4169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966803054215281e-05,
+      "loss": 1.2766,
+      "step": 4170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966781260573394e-05,
+      "loss": 0.7257,
+      "step": 4171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667594599009657e-05,
+      "loss": 0.9039,
+      "step": 4172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667376521981555e-05,
+      "loss": 1.0769,
+      "step": 4173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966715837465122e-05,
+      "loss": 1.1942,
+      "step": 4174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666940157020233e-05,
+      "loss": 1.5343,
+      "step": 4175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666721869090183e-05,
+      "loss": 1.2729,
+      "step": 4176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966650351086266e-05,
+      "loss": 1.0033,
+      "step": 4177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966628508233925e-05,
+      "loss": 0.787,
+      "step": 4178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666066583521538e-05,
+      "loss": 1.1839,
+      "step": 4179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966584801441112e-05,
+      "loss": 1.1226,
+      "step": 4180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966562937500958e-05,
+      "loss": 1.1672,
+      "step": 4181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665410665318512e-05,
+      "loss": 0.9738,
+      "step": 4182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665191885339502e-05,
+      "loss": 0.975,
+      "step": 4183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664973035074145e-05,
+      "loss": 0.9559,
+      "step": 4184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966475411452403e-05,
+      "loss": 1.2497,
+      "step": 4185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664535123690748e-05,
+      "loss": 1.1139,
+      "step": 4186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664316062575895e-05,
+      "loss": 1.072,
+      "step": 4187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664096931181062e-05,
+      "loss": 1.282,
+      "step": 4188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663877729507844e-05,
+      "loss": 1.0837,
+      "step": 4189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663658457557835e-05,
+      "loss": 0.956,
+      "step": 4190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966343911533263e-05,
+      "loss": 1.0694,
+      "step": 4191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663219702833818e-05,
+      "loss": 0.8997,
+      "step": 4192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663000220063e-05,
+      "loss": 1.1203,
+      "step": 4193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662780667021772e-05,
+      "loss": 1.3181,
+      "step": 4194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662561043711732e-05,
+      "loss": 1.0216,
+      "step": 4195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662341350134474e-05,
+      "loss": 1.0499,
+      "step": 4196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662121586291595e-05,
+      "loss": 1.1268,
+      "step": 4197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661901752184697e-05,
+      "loss": 1.0397,
+      "step": 4198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661681847815375e-05,
+      "loss": 1.1768,
+      "step": 4199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966146187318523e-05,
+      "loss": 1.0798,
+      "step": 4200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966124182829586e-05,
+      "loss": 1.3714,
+      "step": 4201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966102171314887e-05,
+      "loss": 1.0504,
+      "step": 4202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660801527745856e-05,
+      "loss": 1.2779,
+      "step": 4203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966058127208842e-05,
+      "loss": 1.0303,
+      "step": 4204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660360946178163e-05,
+      "loss": 1.0259,
+      "step": 4205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966014055001669e-05,
+      "loss": 0.9812,
+      "step": 4206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659920083605604e-05,
+      "loss": 1.1605,
+      "step": 4207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659699546946506e-05,
+      "loss": 0.9377,
+      "step": 4208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659478940040997e-05,
+      "loss": 1.1581,
+      "step": 4209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659258262890683e-05,
+      "loss": 1.2797,
+      "step": 4210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659037515497174e-05,
+      "loss": 1.1365,
+      "step": 4211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658816697862072e-05,
+      "loss": 0.958,
+      "step": 4212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658595809986984e-05,
+      "loss": 1.1957,
+      "step": 4213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965837485187351e-05,
+      "loss": 1.0637,
+      "step": 4214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658153823523262e-05,
+      "loss": 0.9018,
+      "step": 4215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965793272493785e-05,
+      "loss": 0.945,
+      "step": 4216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965771155611888e-05,
+      "loss": 1.1011,
+      "step": 4217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657490317067955e-05,
+      "loss": 1.1313,
+      "step": 4218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657269007786687e-05,
+      "loss": 1.285,
+      "step": 4219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657047628276687e-05,
+      "loss": 0.8095,
+      "step": 4220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965682617853957e-05,
+      "loss": 1.1436,
+      "step": 4221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965660465857693e-05,
+      "loss": 1.1688,
+      "step": 4222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96563830683904e-05,
+      "loss": 1.1039,
+      "step": 4223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656161407981574e-05,
+      "loss": 0.9512,
+      "step": 4224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965593967735207e-05,
+      "loss": 0.9487,
+      "step": 4225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655717876503502e-05,
+      "loss": 0.9576,
+      "step": 4226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965549600543748e-05,
+      "loss": 0.8671,
+      "step": 4227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965527406415562e-05,
+      "loss": 1.3355,
+      "step": 4228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655052052659536e-05,
+      "loss": 1.3317,
+      "step": 4229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654829970950838e-05,
+      "loss": 0.7836,
+      "step": 4230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654607819031145e-05,
+      "loss": 0.9673,
+      "step": 4231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654385596902073e-05,
+      "loss": 1.1877,
+      "step": 4232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965416330456524e-05,
+      "loss": 1.0664,
+      "step": 4233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653940942022255e-05,
+      "loss": 1.2094,
+      "step": 4234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653718509274738e-05,
+      "loss": 1.0259,
+      "step": 4235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653496006324314e-05,
+      "loss": 1.2045,
+      "step": 4236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965327343317259e-05,
+      "loss": 0.8687,
+      "step": 4237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965305078982119e-05,
+      "loss": 1.1758,
+      "step": 4238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652828076271735e-05,
+      "loss": 1.0303,
+      "step": 4239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965260529252584e-05,
+      "loss": 1.1593,
+      "step": 4240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652382438585128e-05,
+      "loss": 1.2532,
+      "step": 4241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965215951445122e-05,
+      "loss": 1.2084,
+      "step": 4242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651936520125737e-05,
+      "loss": 1.065,
+      "step": 4243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96517134556103e-05,
+      "loss": 0.8783,
+      "step": 4244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965149032090653e-05,
+      "loss": 1.0615,
+      "step": 4245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965126711601605e-05,
+      "loss": 0.9014,
+      "step": 4246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651043840940483e-05,
+      "loss": 0.9729,
+      "step": 4247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650820495681454e-05,
+      "loss": 1.2129,
+      "step": 4248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650597080240586e-05,
+      "loss": 1.1067,
+      "step": 4249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650373594619508e-05,
+      "loss": 0.7158,
+      "step": 4250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965015003881984e-05,
+      "loss": 1.1109,
+      "step": 4251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649926412843205e-05,
+      "loss": 1.1344,
+      "step": 4252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649702716691238e-05,
+      "loss": 1.0992,
+      "step": 4253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649478950365558e-05,
+      "loss": 1.0981,
+      "step": 4254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649255113867796e-05,
+      "loss": 1.2253,
+      "step": 4255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649031207199583e-05,
+      "loss": 1.033,
+      "step": 4256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964880723036254e-05,
+      "loss": 0.9706,
+      "step": 4257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648583183358303e-05,
+      "loss": 0.6409,
+      "step": 4258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648359066188497e-05,
+      "loss": 0.8061,
+      "step": 4259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648134878854747e-05,
+      "loss": 1.0295,
+      "step": 4260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647910621358694e-05,
+      "loss": 1.1215,
+      "step": 4261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647686293701963e-05,
+      "loss": 0.9913,
+      "step": 4262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647461895886187e-05,
+      "loss": 1.0849,
+      "step": 4263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647237427912993e-05,
+      "loss": 1.046,
+      "step": 4264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964701288978402e-05,
+      "loss": 0.8839,
+      "step": 4265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646788281500897e-05,
+      "loss": 1.2782,
+      "step": 4266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646563603065257e-05,
+      "loss": 1.0675,
+      "step": 4267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646338854478735e-05,
+      "loss": 0.8826,
+      "step": 4268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964611403574297e-05,
+      "loss": 1.3964,
+      "step": 4269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964588914685959e-05,
+      "loss": 1.2207,
+      "step": 4270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645664187830232e-05,
+      "loss": 1.1678,
+      "step": 4271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645439158656535e-05,
+      "loss": 1.2265,
+      "step": 4272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964521405934013e-05,
+      "loss": 0.8875,
+      "step": 4273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964498888988266e-05,
+      "loss": 0.8476,
+      "step": 4274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644763650285758e-05,
+      "loss": 1.2511,
+      "step": 4275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644538340551066e-05,
+      "loss": 1.1407,
+      "step": 4276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644312960680222e-05,
+      "loss": 0.9447,
+      "step": 4277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644087510674858e-05,
+      "loss": 1.3391,
+      "step": 4278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643861990536622e-05,
+      "loss": 0.8681,
+      "step": 4279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964363640026715e-05,
+      "loss": 1.2938,
+      "step": 4280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964341073986808e-05,
+      "loss": 1.0977,
+      "step": 4281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964318500934106e-05,
+      "loss": 1.0614,
+      "step": 4282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642959208687725e-05,
+      "loss": 1.0227,
+      "step": 4283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642733337909725e-05,
+      "loss": 0.8653,
+      "step": 4284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642507397008692e-05,
+      "loss": 0.9431,
+      "step": 4285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642281385986278e-05,
+      "loss": 1.0916,
+      "step": 4286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964205530484412e-05,
+      "loss": 1.0885,
+      "step": 4287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641829153583866e-05,
+      "loss": 0.9797,
+      "step": 4288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964160293220716e-05,
+      "loss": 1.1707,
+      "step": 4289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641376640715646e-05,
+      "loss": 1.086,
+      "step": 4290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964115027911097e-05,
+      "loss": 1.1237,
+      "step": 4291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640923847394783e-05,
+      "loss": 1.0939,
+      "step": 4292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640697345568722e-05,
+      "loss": 1.1689,
+      "step": 4293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640470773634443e-05,
+      "loss": 1.2262,
+      "step": 4294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640244131593588e-05,
+      "loss": 0.7106,
+      "step": 4295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640017419447807e-05,
+      "loss": 1.0224,
+      "step": 4296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963979063719875e-05,
+      "loss": 1.0778,
+      "step": 4297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639563784848064e-05,
+      "loss": 1.0866,
+      "step": 4298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639336862397402e-05,
+      "loss": 1.1366,
+      "step": 4299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963910986984841e-05,
+      "loss": 1.0631,
+      "step": 4300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963888280720274e-05,
+      "loss": 0.6127,
+      "step": 4301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638655674462045e-05,
+      "loss": 1.0879,
+      "step": 4302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638428471627974e-05,
+      "loss": 1.2281,
+      "step": 4303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963820119870218e-05,
+      "loss": 1.5671,
+      "step": 4304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963797385568632e-05,
+      "loss": 1.191,
+      "step": 4305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637746442582043e-05,
+      "loss": 1.0261,
+      "step": 4306
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637518959391003e-05,
+      "loss": 1.0953,
+      "step": 4307
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637291406114857e-05,
+      "loss": 1.055,
+      "step": 4308
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637063782755253e-05,
+      "loss": 0.9593,
+      "step": 4309
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636836089313854e-05,
+      "loss": 1.022,
+      "step": 4310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636608325792314e-05,
+      "loss": 1.2614,
+      "step": 4311
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636380492192285e-05,
+      "loss": 1.039,
+      "step": 4312
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636152588515428e-05,
+      "loss": 0.7307,
+      "step": 4313
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96359246147634e-05,
+      "loss": 1.2174,
+      "step": 4314
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963569657093786e-05,
+      "loss": 0.9845,
+      "step": 4315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963546845704046e-05,
+      "loss": 1.0658,
+      "step": 4316
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635240273072866e-05,
+      "loss": 0.9647,
+      "step": 4317
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635012019036734e-05,
+      "loss": 0.9549,
+      "step": 4318
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634783694933725e-05,
+      "loss": 0.8527,
+      "step": 4319
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96345553007655e-05,
+      "loss": 1.1138,
+      "step": 4320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963432683653372e-05,
+      "loss": 1.2226,
+      "step": 4321
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634098302240045e-05,
+      "loss": 1.312,
+      "step": 4322
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633869697886133e-05,
+      "loss": 1.1499,
+      "step": 4323
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963364102347365e-05,
+      "loss": 1.0943,
+      "step": 4324
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633412279004267e-05,
+      "loss": 1.1947,
+      "step": 4325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633183464479638e-05,
+      "loss": 1.2253,
+      "step": 4326
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632954579901426e-05,
+      "loss": 1.2922,
+      "step": 4327
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96327256252713e-05,
+      "loss": 0.9841,
+      "step": 4328
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632496600590925e-05,
+      "loss": 1.0836,
+      "step": 4329
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632267505861966e-05,
+      "loss": 0.808,
+      "step": 4330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632038341086085e-05,
+      "loss": 0.9688,
+      "step": 4331
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963180910626495e-05,
+      "loss": 0.895,
+      "step": 4332
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631579801400232e-05,
+      "loss": 0.9685,
+      "step": 4333
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631350426493598e-05,
+      "loss": 0.9487,
+      "step": 4334
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631120981546713e-05,
+      "loss": 1.2096,
+      "step": 4335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630891466561245e-05,
+      "loss": 0.9367,
+      "step": 4336
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630661881538865e-05,
+      "loss": 0.6904,
+      "step": 4337
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630432226481243e-05,
+      "loss": 0.7386,
+      "step": 4338
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963020250139005e-05,
+      "loss": 1.1074,
+      "step": 4339
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629972706266954e-05,
+      "loss": 1.0424,
+      "step": 4340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629742841113625e-05,
+      "loss": 0.868,
+      "step": 4341
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962951290593174e-05,
+      "loss": 1.0679,
+      "step": 4342
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629282900722963e-05,
+      "loss": 1.0772,
+      "step": 4343
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629052825488976e-05,
+      "loss": 1.0513,
+      "step": 4344
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628822680231446e-05,
+      "loss": 0.9378,
+      "step": 4345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628592464952047e-05,
+      "loss": 1.2346,
+      "step": 4346
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962836217965246e-05,
+      "loss": 1.1018,
+      "step": 4347
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962813182433435e-05,
+      "loss": 1.0675,
+      "step": 4348
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627901398999395e-05,
+      "loss": 1.308,
+      "step": 4349
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627670903649273e-05,
+      "loss": 1.0559,
+      "step": 4350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9627440338285658e-05,
+      "loss": 1.0884,
+      "step": 4351
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962720970291023e-05,
+      "loss": 1.1397,
+      "step": 4352
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962697899752466e-05,
+      "loss": 1.239,
+      "step": 4353
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626748222130634e-05,
+      "loss": 1.1116,
+      "step": 4354
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626517376729822e-05,
+      "loss": 1.1164,
+      "step": 4355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626286461323908e-05,
+      "loss": 1.374,
+      "step": 4356
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962605547591457e-05,
+      "loss": 1.2435,
+      "step": 4357
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625824420503493e-05,
+      "loss": 0.8855,
+      "step": 4358
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625593295092345e-05,
+      "loss": 1.0786,
+      "step": 4359
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625362099682818e-05,
+      "loss": 1.0341,
+      "step": 4360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625130834276585e-05,
+      "loss": 0.9622,
+      "step": 4361
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624899498875334e-05,
+      "loss": 1.0853,
+      "step": 4362
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624668093480746e-05,
+      "loss": 1.1514,
+      "step": 4363
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96244366180945e-05,
+      "loss": 1.08,
+      "step": 4364
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9624205072718285e-05,
+      "loss": 0.8057,
+      "step": 4365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962397345735378e-05,
+      "loss": 0.9036,
+      "step": 4366
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9623741772002673e-05,
+      "loss": 1.1765,
+      "step": 4367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623510016666644e-05,
+      "loss": 0.9695,
+      "step": 4368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623278191347384e-05,
+      "loss": 1.1205,
+      "step": 4369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623046296046574e-05,
+      "loss": 1.2479,
+      "step": 4370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622814330765908e-05,
+      "loss": 1.2515,
+      "step": 4371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622582295507066e-05,
+      "loss": 1.2476,
+      "step": 4372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622350190271732e-05,
+      "loss": 0.9218,
+      "step": 4373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622118015061603e-05,
+      "loss": 1.0042,
+      "step": 4374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962188576987836e-05,
+      "loss": 1.0405,
+      "step": 4375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96216534547237e-05,
+      "loss": 1.2646,
+      "step": 4376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621421069599304e-05,
+      "loss": 0.9602,
+      "step": 4377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621188614506867e-05,
+      "loss": 1.0648,
+      "step": 4378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962095608944808e-05,
+      "loss": 0.888,
+      "step": 4379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620723494424627e-05,
+      "loss": 0.9648,
+      "step": 4380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962049082943821e-05,
+      "loss": 1.1348,
+      "step": 4381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962025809449051e-05,
+      "loss": 0.9816,
+      "step": 4382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620025289583228e-05,
+      "loss": 0.7409,
+      "step": 4383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619792414718054e-05,
+      "loss": 1.0707,
+      "step": 4384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961955946989668e-05,
+      "loss": 1.096,
+      "step": 4385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96193264551208e-05,
+      "loss": 1.1266,
+      "step": 4386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619093370392113e-05,
+      "loss": 1.1653,
+      "step": 4387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961886021571231e-05,
+      "loss": 1.0025,
+      "step": 4388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618626991083084e-05,
+      "loss": 1.3903,
+      "step": 4389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961839369650614e-05,
+      "loss": 0.873,
+      "step": 4390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618160331983165e-05,
+      "loss": 1.1149,
+      "step": 4391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961792689751586e-05,
+      "loss": 1.152,
+      "step": 4392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617693393105922e-05,
+      "loss": 1.1651,
+      "step": 4393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961745981875505e-05,
+      "loss": 1.0845,
+      "step": 4394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617226174464945e-05,
+      "loss": 1.0675,
+      "step": 4395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616992460237304e-05,
+      "loss": 1.0474,
+      "step": 4396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961675867607382e-05,
+      "loss": 1.0286,
+      "step": 4397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616524821976205e-05,
+      "loss": 1.1346,
+      "step": 4398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961629089794615e-05,
+      "loss": 0.9135,
+      "step": 4399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616056903985362e-05,
+      "loss": 0.9917,
+      "step": 4400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615822840095537e-05,
+      "loss": 1.161,
+      "step": 4401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615588706278385e-05,
+      "loss": 0.8567,
+      "step": 4402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96153545025356e-05,
+      "loss": 1.1181,
+      "step": 4403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615120228868895e-05,
+      "loss": 0.891,
+      "step": 4404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961488588527996e-05,
+      "loss": 1.3162,
+      "step": 4405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961465147177051e-05,
+      "loss": 0.9685,
+      "step": 4406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961441698834225e-05,
+      "loss": 1.1934,
+      "step": 4407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961418243499688e-05,
+      "loss": 1.1599,
+      "step": 4408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613947811736106e-05,
+      "loss": 1.2053,
+      "step": 4409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613713118561638e-05,
+      "loss": 1.1634,
+      "step": 4410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613478355475185e-05,
+      "loss": 1.0495,
+      "step": 4411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961324352247844e-05,
+      "loss": 1.0738,
+      "step": 4412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613008619573127e-05,
+      "loss": 1.158,
+      "step": 4413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612773646760947e-05,
+      "loss": 0.7008,
+      "step": 4414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961253860404361e-05,
+      "loss": 0.8578,
+      "step": 4415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612303491422824e-05,
+      "loss": 1.0676,
+      "step": 4416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96120683089003e-05,
+      "loss": 1.4408,
+      "step": 4417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611833056477746e-05,
+      "loss": 0.7815,
+      "step": 4418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611597734156878e-05,
+      "loss": 0.8349,
+      "step": 4419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96113623419394e-05,
+      "loss": 0.9008,
+      "step": 4420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611126879827028e-05,
+      "loss": 0.9978,
+      "step": 4421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9610891347821477e-05,
+      "loss": 1.3173,
+      "step": 4422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961065574592446e-05,
+      "loss": 0.8429,
+      "step": 4423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961042007413768e-05,
+      "loss": 1.1302,
+      "step": 4424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961018433246286e-05,
+      "loss": 0.8707,
+      "step": 4425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609948520901713e-05,
+      "loss": 1.0511,
+      "step": 4426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609712639455956e-05,
+      "loss": 0.7793,
+      "step": 4427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609476688127297e-05,
+      "loss": 0.9924,
+      "step": 4428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960924066691746e-05,
+      "loss": 1.0582,
+      "step": 4429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609004575828155e-05,
+      "loss": 1.1286,
+      "step": 4430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96087684148611e-05,
+      "loss": 0.8832,
+      "step": 4431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608532184018017e-05,
+      "loss": 0.7571,
+      "step": 4432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608295883300624e-05,
+      "loss": 1.104,
+      "step": 4433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960805951271063e-05,
+      "loss": 1.2072,
+      "step": 4434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607823072249764e-05,
+      "loss": 1.2336,
+      "step": 4435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607586561919743e-05,
+      "loss": 1.3069,
+      "step": 4436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607349981722284e-05,
+      "loss": 0.9293,
+      "step": 4437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607113331659108e-05,
+      "loss": 1.2108,
+      "step": 4438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960687661173194e-05,
+      "loss": 0.9986,
+      "step": 4439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606639821942496e-05,
+      "loss": 0.7977,
+      "step": 4440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.96064029622925e-05,
+      "loss": 1.1935,
+      "step": 4441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606166032783674e-05,
+      "loss": 1.0456,
+      "step": 4442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605929033417744e-05,
+      "loss": 0.956,
+      "step": 4443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605691964196428e-05,
+      "loss": 1.0335,
+      "step": 4444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960545482512146e-05,
+      "loss": 2.0177,
+      "step": 4445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960521761619455e-05,
+      "loss": 1.1054,
+      "step": 4446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604980337417436e-05,
+      "loss": 1.1685,
+      "step": 4447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960474298879184e-05,
+      "loss": 0.9201,
+      "step": 4448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604505570319483e-05,
+      "loss": 1.4959,
+      "step": 4449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604268082002097e-05,
+      "loss": 1.4016,
+      "step": 4450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604030523841403e-05,
+      "loss": 1.0468,
+      "step": 4451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603792895839136e-05,
+      "loss": 1.1522,
+      "step": 4452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960355519799702e-05,
+      "loss": 1.3218,
+      "step": 4453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603317430316785e-05,
+      "loss": 1.1505,
+      "step": 4454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9603079592800157e-05,
+      "loss": 1.4226,
+      "step": 4455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602841685448867e-05,
+      "loss": 0.6104,
+      "step": 4456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960260370826465e-05,
+      "loss": 0.7939,
+      "step": 4457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602365661249228e-05,
+      "loss": 1.0567,
+      "step": 4458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602127544404342e-05,
+      "loss": 1.0589,
+      "step": 4459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601889357731713e-05,
+      "loss": 1.0594,
+      "step": 4460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960165110123308e-05,
+      "loss": 0.9976,
+      "step": 4461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601412774910177e-05,
+      "loss": 1.132,
+      "step": 4462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601174378764728e-05,
+      "loss": 3.179,
+      "step": 4463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960093591279848e-05,
+      "loss": 1.1445,
+      "step": 4464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600697377013154e-05,
+      "loss": 1.3306,
+      "step": 4465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600458771410494e-05,
+      "loss": 1.1346,
+      "step": 4466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960022009599223e-05,
+      "loss": 1.0498,
+      "step": 4467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599981350760098e-05,
+      "loss": 0.9966,
+      "step": 4468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599742535715836e-05,
+      "loss": 1.8802,
+      "step": 4469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599503650861183e-05,
+      "loss": 0.9295,
+      "step": 4470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959926469619787e-05,
+      "loss": 1.6618,
+      "step": 4471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959902567172764e-05,
+      "loss": 1.0766,
+      "step": 4472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959878657745223e-05,
+      "loss": 1.1565,
+      "step": 4473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959854741337338e-05,
+      "loss": 1.3327,
+      "step": 4474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598308179492824e-05,
+      "loss": 1.1827,
+      "step": 4475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598068875812305e-05,
+      "loss": 0.8952,
+      "step": 4476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597829502333564e-05,
+      "loss": 0.8977,
+      "step": 4477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597590059058342e-05,
+      "loss": 1.2894,
+      "step": 4478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597350545988378e-05,
+      "loss": 0.9117,
+      "step": 4479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597110963125416e-05,
+      "loss": 0.5886,
+      "step": 4480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95968713104712e-05,
+      "loss": 0.6834,
+      "step": 4481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596631588027466e-05,
+      "loss": 1.2042,
+      "step": 4482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596391795795958e-05,
+      "loss": 0.9906,
+      "step": 4483
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959615193377843e-05,
+      "loss": 1.0668,
+      "step": 4484
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959591200197662e-05,
+      "loss": 1.2152,
+      "step": 4485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595672000392267e-05,
+      "loss": 0.905,
+      "step": 4486
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595431929027126e-05,
+      "loss": 0.8583,
+      "step": 4487
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595191787882936e-05,
+      "loss": 1.1542,
+      "step": 4488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594951576961446e-05,
+      "loss": 1.0173,
+      "step": 4489
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95947112962644e-05,
+      "loss": 0.9742,
+      "step": 4490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594470945793555e-05,
+      "loss": 0.9499,
+      "step": 4491
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594230525550647e-05,
+      "loss": 1.0908,
+      "step": 4492
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593990035537432e-05,
+      "loss": 1.1874,
+      "step": 4493
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959374947575565e-05,
+      "loss": 1.3538,
+      "step": 4494
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593508846207063e-05,
+      "loss": 0.8997,
+      "step": 4495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593268146893413e-05,
+      "loss": 1.1465,
+      "step": 4496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593027377816448e-05,
+      "loss": 1.1088,
+      "step": 4497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592786538977928e-05,
+      "loss": 0.9024,
+      "step": 4498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592545630379594e-05,
+      "loss": 1.0094,
+      "step": 4499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592304652023208e-05,
+      "loss": 0.8349,
+      "step": 4500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592063603910514e-05,
+      "loss": 1.2367,
+      "step": 4501
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591822486043268e-05,
+      "loss": 1.1073,
+      "step": 4502
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591581298423225e-05,
+      "loss": 1.2437,
+      "step": 4503
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959134004105214e-05,
+      "loss": 1.232,
+      "step": 4504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959109871393176e-05,
+      "loss": 1.218,
+      "step": 4505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590857317063848e-05,
+      "loss": 1.1147,
+      "step": 4506
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590615850450155e-05,
+      "loss": 1.0637,
+      "step": 4507
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959037431409244e-05,
+      "loss": 0.945,
+      "step": 4508
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590132707992457e-05,
+      "loss": 1.2339,
+      "step": 4509
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589891032151964e-05,
+      "loss": 1.2742,
+      "step": 4510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958964928657272e-05,
+      "loss": 0.9367,
+      "step": 4511
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589407471256482e-05,
+      "loss": 1.2167,
+      "step": 4512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589165586205007e-05,
+      "loss": 1.1525,
+      "step": 4513
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588923631420054e-05,
+      "loss": 1.1156,
+      "step": 4514
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588681606903385e-05,
+      "loss": 1.0808,
+      "step": 4515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588439512656758e-05,
+      "loss": 1.0155,
+      "step": 4516
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958819734868193e-05,
+      "loss": 1.9012,
+      "step": 4517
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587955114980673e-05,
+      "loss": 1.034,
+      "step": 4518
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587712811554737e-05,
+      "loss": 1.1532,
+      "step": 4519
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958747043840589e-05,
+      "loss": 0.9241,
+      "step": 4520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587227995535897e-05,
+      "loss": 0.9426,
+      "step": 4521
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586985482946514e-05,
+      "loss": 1.1614,
+      "step": 4522
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586742900639506e-05,
+      "loss": 0.9161,
+      "step": 4523
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958650024861664e-05,
+      "loss": 1.2409,
+      "step": 4524
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586257526879682e-05,
+      "loss": 1.1816,
+      "step": 4525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586014735430393e-05,
+      "loss": 0.8978,
+      "step": 4526
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958577187427054e-05,
+      "loss": 1.2028,
+      "step": 4527
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585528943401892e-05,
+      "loss": 0.8995,
+      "step": 4528
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958528594282621e-05,
+      "loss": 0.7335,
+      "step": 4529
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585042872545266e-05,
+      "loss": 0.8255,
+      "step": 4530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584799732560825e-05,
+      "loss": 1.0034,
+      "step": 4531
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584556522874657e-05,
+      "loss": 1.5359,
+      "step": 4532
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958431324348853e-05,
+      "loss": 1.0103,
+      "step": 4533
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958406989440421e-05,
+      "loss": 0.9486,
+      "step": 4534
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583826475623472e-05,
+      "loss": 1.4347,
+      "step": 4535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958358298714808e-05,
+      "loss": 0.9489,
+      "step": 4536
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583339428979812e-05,
+      "loss": 0.9643,
+      "step": 4537
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583095801120436e-05,
+      "loss": 1.0552,
+      "step": 4538
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958285210357172e-05,
+      "loss": 1.286,
+      "step": 4539
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958260833633544e-05,
+      "loss": 1.0006,
+      "step": 4540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582364499413372e-05,
+      "loss": 0.9196,
+      "step": 4541
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582120592807282e-05,
+      "loss": 0.8276,
+      "step": 4542
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958187661651895e-05,
+      "loss": 1.2236,
+      "step": 4543
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958163257055014e-05,
+      "loss": 1.1734,
+      "step": 4544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958138845490264e-05,
+      "loss": 1.2456,
+      "step": 4545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958114426957822e-05,
+      "loss": 1.0626,
+      "step": 4546
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958090001457865e-05,
+      "loss": 1.369,
+      "step": 4547
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580655689905714e-05,
+      "loss": 1.089,
+      "step": 4548
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580411295561186e-05,
+      "loss": 1.1598,
+      "step": 4549
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580166831546843e-05,
+      "loss": 1.3015,
+      "step": 4550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957992229786446e-05,
+      "loss": 1.0527,
+      "step": 4551
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579677694515824e-05,
+      "loss": 0.9941,
+      "step": 4552
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579433021502705e-05,
+      "loss": 0.9157,
+      "step": 4553
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579188278826886e-05,
+      "loss": 0.8878,
+      "step": 4554
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578943466490146e-05,
+      "loss": 0.8876,
+      "step": 4555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578698584494266e-05,
+      "loss": 1.079,
+      "step": 4556
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578453632841023e-05,
+      "loss": 1.37,
+      "step": 4557
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578208611532204e-05,
+      "loss": 1.0348,
+      "step": 4558
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577963520569587e-05,
+      "loss": 0.9697,
+      "step": 4559
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577718359954955e-05,
+      "loss": 1.0607,
+      "step": 4560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577473129690095e-05,
+      "loss": 1.1447,
+      "step": 4561
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577227829776782e-05,
+      "loss": 0.9786,
+      "step": 4562
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957698246021681e-05,
+      "loss": 0.9943,
+      "step": 4563
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576737021011953e-05,
+      "loss": 1.4622,
+      "step": 4564
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576491512164004e-05,
+      "loss": 1.3358,
+      "step": 4565
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576245933674745e-05,
+      "loss": 1.1656,
+      "step": 4566
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576000285545964e-05,
+      "loss": 0.7824,
+      "step": 4567
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575754567779443e-05,
+      "loss": 0.9841,
+      "step": 4568
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957550878037697e-05,
+      "loss": 1.5667,
+      "step": 4569
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575262923340338e-05,
+      "loss": 1.1094,
+      "step": 4570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575016996671328e-05,
+      "loss": 1.2214,
+      "step": 4571
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574771000371734e-05,
+      "loss": 1.0356,
+      "step": 4572
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574524934443338e-05,
+      "loss": 0.8738,
+      "step": 4573
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574278798887936e-05,
+      "loss": 1.2254,
+      "step": 4574
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574032593707314e-05,
+      "loss": 1.1539,
+      "step": 4575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573786318903263e-05,
+      "loss": 0.9077,
+      "step": 4576
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573539974477577e-05,
+      "loss": 1.2092,
+      "step": 4577
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573293560432043e-05,
+      "loss": 1.157,
+      "step": 4578
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573047076768455e-05,
+      "loss": 1.2309,
+      "step": 4579
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572800523488608e-05,
+      "loss": 0.9272,
+      "step": 4580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957255390059429e-05,
+      "loss": 1.1182,
+      "step": 4581
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95723072080873e-05,
+      "loss": 1.0522,
+      "step": 4582
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957206044596943e-05,
+      "loss": 1.1958,
+      "step": 4583
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957181361424247e-05,
+      "loss": 1.0283,
+      "step": 4584
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571566712908215e-05,
+      "loss": 1.2641,
+      "step": 4585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957131974196847e-05,
+      "loss": 0.9973,
+      "step": 4586
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571072701425026e-05,
+      "loss": 0.7342,
+      "step": 4587
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570825591279674e-05,
+      "loss": 0.9477,
+      "step": 4588
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957057841153422e-05,
+      "loss": 0.89,
+      "step": 4589
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957033116219045e-05,
+      "loss": 1.1238,
+      "step": 4590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570083843250177e-05,
+      "loss": 1.0974,
+      "step": 4591
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569836454715186e-05,
+      "loss": 1.3488,
+      "step": 4592
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569588996587284e-05,
+      "loss": 0.9147,
+      "step": 4593
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956934146886827e-05,
+      "loss": 1.0724,
+      "step": 4594
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569093871559938e-05,
+      "loss": 1.258,
+      "step": 4595
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568846204664093e-05,
+      "loss": 1.1216,
+      "step": 4596
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956859846818254e-05,
+      "loss": 0.8655,
+      "step": 4597
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568350662117073e-05,
+      "loss": 0.8716,
+      "step": 4598
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568102786469497e-05,
+      "loss": 1.2433,
+      "step": 4599
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567854841241617e-05,
+      "loss": 1.0253,
+      "step": 4600
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567606826435232e-05,
+      "loss": 0.8936,
+      "step": 4601
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956735874205215e-05,
+      "loss": 1.1057,
+      "step": 4602
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956711058809417e-05,
+      "loss": 1.3717,
+      "step": 4603
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566862364563102e-05,
+      "loss": 1.0909,
+      "step": 4604
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956661407146075e-05,
+      "loss": 1.0937,
+      "step": 4605
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566365708788914e-05,
+      "loss": 1.1341,
+      "step": 4606
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566117276549405e-05,
+      "loss": 0.9657,
+      "step": 4607
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565868774744033e-05,
+      "loss": 1.1709,
+      "step": 4608
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95656202033746e-05,
+      "loss": 1.0983,
+      "step": 4609
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956537156244291e-05,
+      "loss": 1.0468,
+      "step": 4610
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956512285195078e-05,
+      "loss": 1.1812,
+      "step": 4611
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956487407190001e-05,
+      "loss": 1.0955,
+      "step": 4612
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564625222292417e-05,
+      "loss": 1.0923,
+      "step": 4613
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564376303129808e-05,
+      "loss": 0.9075,
+      "step": 4614
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956412731441399e-05,
+      "loss": 1.1379,
+      "step": 4615
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956387825614678e-05,
+      "loss": 1.1878,
+      "step": 4616
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956362912832998e-05,
+      "loss": 0.8609,
+      "step": 4617
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956337993096541e-05,
+      "loss": 1.2473,
+      "step": 4618
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563130664054875e-05,
+      "loss": 1.0599,
+      "step": 4619
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562881327600197e-05,
+      "loss": 1.2404,
+      "step": 4620
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562631921603183e-05,
+      "loss": 1.0659,
+      "step": 4621
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562382446065648e-05,
+      "loss": 0.9364,
+      "step": 4622
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562132900989405e-05,
+      "loss": 0.9689,
+      "step": 4623
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956188328637627e-05,
+      "loss": 1.1008,
+      "step": 4624
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561633602228058e-05,
+      "loss": 0.9873,
+      "step": 4625
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561383848546587e-05,
+      "loss": 1.1327,
+      "step": 4626
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561134025333666e-05,
+      "loss": 1.0705,
+      "step": 4627
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956088413259112e-05,
+      "loss": 1.3052,
+      "step": 4628
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956063417032076e-05,
+      "loss": 1.0901,
+      "step": 4629
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560384138524408e-05,
+      "loss": 1.278,
+      "step": 4630
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560134037203885e-05,
+      "loss": 0.9068,
+      "step": 4631
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559883866361e-05,
+      "loss": 0.8569,
+      "step": 4632
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955963362599758e-05,
+      "loss": 1.1673,
+      "step": 4633
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559383316115445e-05,
+      "loss": 1.0052,
+      "step": 4634
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955913293671641e-05,
+      "loss": 0.8298,
+      "step": 4635
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558882487802297e-05,
+      "loss": 1.0683,
+      "step": 4636
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558631969374935e-05,
+      "loss": 1.0036,
+      "step": 4637
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558381381436135e-05,
+      "loss": 1.3491,
+      "step": 4638
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558130723987724e-05,
+      "loss": 1.2822,
+      "step": 4639
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955787999703153e-05,
+      "loss": 1.4333,
+      "step": 4640
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557629200569365e-05,
+      "loss": 1.2921,
+      "step": 4641
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557378334603063e-05,
+      "loss": 1.0323,
+      "step": 4642
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955712739913444e-05,
+      "loss": 0.7395,
+      "step": 4643
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556876394165332e-05,
+      "loss": 0.8955,
+      "step": 4644
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556625319697553e-05,
+      "loss": 1.3601,
+      "step": 4645
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556374175732936e-05,
+      "loss": 1.4368,
+      "step": 4646
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556122962273302e-05,
+      "loss": 1.186,
+      "step": 4647
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555871679320482e-05,
+      "loss": 1.0063,
+      "step": 4648
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95556203268763e-05,
+      "loss": 0.9359,
+      "step": 4649
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555368904942593e-05,
+      "loss": 1.2469,
+      "step": 4650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555117413521175e-05,
+      "loss": 0.9622,
+      "step": 4651
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554865852613884e-05,
+      "loss": 1.1392,
+      "step": 4652
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955461422222255e-05,
+      "loss": 1.0654,
+      "step": 4653
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554362522348996e-05,
+      "loss": 1.1052,
+      "step": 4654
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955411075299506e-05,
+      "loss": 1.1121,
+      "step": 4655
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553858914162566e-05,
+      "loss": 1.0069,
+      "step": 4656
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553607005853355e-05,
+      "loss": 1.0215,
+      "step": 4657
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553355028069247e-05,
+      "loss": 1.0524,
+      "step": 4658
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955310298081208e-05,
+      "loss": 0.8535,
+      "step": 4659
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552850864083694e-05,
+      "loss": 0.6967,
+      "step": 4660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552598677885913e-05,
+      "loss": 1.0494,
+      "step": 4661
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955234642222057e-05,
+      "loss": 0.8192,
+      "step": 4662
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552094097089504e-05,
+      "loss": 1.4375,
+      "step": 4663
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955184170249455e-05,
+      "loss": 1.0439,
+      "step": 4664
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551589238437546e-05,
+      "loss": 1.1797,
+      "step": 4665
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955133670492032e-05,
+      "loss": 1.2013,
+      "step": 4666
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551084101944714e-05,
+      "loss": 0.9511,
+      "step": 4667
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550831429512563e-05,
+      "loss": 1.0445,
+      "step": 4668
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955057868762571e-05,
+      "loss": 1.0048,
+      "step": 4669
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550325876285982e-05,
+      "loss": 1.1837,
+      "step": 4670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550072995495225e-05,
+      "loss": 0.888,
+      "step": 4671
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549820045255278e-05,
+      "loss": 0.9422,
+      "step": 4672
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954956702556798e-05,
+      "loss": 1.1416,
+      "step": 4673
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954931393643517e-05,
+      "loss": 0.9917,
+      "step": 4674
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549060777858687e-05,
+      "loss": 0.9397,
+      "step": 4675
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548807549840373e-05,
+      "loss": 0.8378,
+      "step": 4676
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954855425238207e-05,
+      "loss": 0.9996,
+      "step": 4677
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954830088548562e-05,
+      "loss": 0.9179,
+      "step": 4678
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548047449152873e-05,
+      "loss": 0.6833,
+      "step": 4679
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954779394338566e-05,
+      "loss": 0.9704,
+      "step": 4680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547540368185823e-05,
+      "loss": 1.029,
+      "step": 4681
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954728672355522e-05,
+      "loss": 1.2526,
+      "step": 4682
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547033009495685e-05,
+      "loss": 0.9148,
+      "step": 4683
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546779226009065e-05,
+      "loss": 0.8865,
+      "step": 4684
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954652537309721e-05,
+      "loss": 0.8248,
+      "step": 4685
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546271450761956e-05,
+      "loss": 1.5443,
+      "step": 4686
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954601745900516e-05,
+      "loss": 1.0838,
+      "step": 4687
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545763397828662e-05,
+      "loss": 0.9019,
+      "step": 4688
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545509267234317e-05,
+      "loss": 1.1562,
+      "step": 4689
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545255067223964e-05,
+      "loss": 1.1819,
+      "step": 4690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954500079779946e-05,
+      "loss": 1.0048,
+      "step": 4691
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954474645896265e-05,
+      "loss": 1.0897,
+      "step": 4692
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544492050715383e-05,
+      "loss": 1.1239,
+      "step": 4693
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544237573059506e-05,
+      "loss": 1.0418,
+      "step": 4694
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954398302599688e-05,
+      "loss": 1.1033,
+      "step": 4695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543728409529348e-05,
+      "loss": 0.6942,
+      "step": 4696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543473723658763e-05,
+      "loss": 1.13,
+      "step": 4697
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543218968386973e-05,
+      "loss": 1.1241,
+      "step": 4698
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954296414371584e-05,
+      "loss": 1.3224,
+      "step": 4699
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954270924964721e-05,
+      "loss": 1.009,
+      "step": 4700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542454286182938e-05,
+      "loss": 1.0535,
+      "step": 4701
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954219925332488e-05,
+      "loss": 1.0794,
+      "step": 4702
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541944151074886e-05,
+      "loss": 1.3602,
+      "step": 4703
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541688979434822e-05,
+      "loss": 1.1299,
+      "step": 4704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954143373840653e-05,
+      "loss": 0.8926,
+      "step": 4705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541178427991876e-05,
+      "loss": 1.2801,
+      "step": 4706
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954092304819271e-05,
+      "loss": 0.9332,
+      "step": 4707
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540667599010893e-05,
+      "loss": 1.1635,
+      "step": 4708
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540412080448282e-05,
+      "loss": 1.098,
+      "step": 4709
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540156492506734e-05,
+      "loss": 0.8974,
+      "step": 4710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953990083518811e-05,
+      "loss": 1.0321,
+      "step": 4711
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539645108494266e-05,
+      "loss": 1.1916,
+      "step": 4712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539389312427067e-05,
+      "loss": 1.0136,
+      "step": 4713
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539133446988364e-05,
+      "loss": 0.88,
+      "step": 4714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538877512180024e-05,
+      "loss": 1.2896,
+      "step": 4715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538621508003914e-05,
+      "loss": 0.9555,
+      "step": 4716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538365434461884e-05,
+      "loss": 1.065,
+      "step": 4717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95381092915558e-05,
+      "loss": 1.1535,
+      "step": 4718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953785307928753e-05,
+      "loss": 1.0481,
+      "step": 4719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537596797658932e-05,
+      "loss": 1.3942,
+      "step": 4720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953734044667187e-05,
+      "loss": 0.9565,
+      "step": 4721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953708402632821e-05,
+      "loss": 0.9216,
+      "step": 4722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536827536629815e-05,
+      "loss": 1.0136,
+      "step": 4723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536570977578553e-05,
+      "loss": 1.4676,
+      "step": 4724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536314349176288e-05,
+      "loss": 1.0031,
+      "step": 4725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953605765142488e-05,
+      "loss": 1.1381,
+      "step": 4726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953580088432621e-05,
+      "loss": 0.856,
+      "step": 4727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535544047882132e-05,
+      "loss": 0.9129,
+      "step": 4728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953528714209452e-05,
+      "loss": 1.1065,
+      "step": 4729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535030166965245e-05,
+      "loss": 1.0706,
+      "step": 4730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534773122496164e-05,
+      "loss": 1.3595,
+      "step": 4731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953451600868916e-05,
+      "loss": 0.9152,
+      "step": 4732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534258825546095e-05,
+      "loss": 1.0378,
+      "step": 4733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953400157306884e-05,
+      "loss": 0.9395,
+      "step": 4734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953374425125927e-05,
+      "loss": 0.8926,
+      "step": 4735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533486860119247e-05,
+      "loss": 1.1363,
+      "step": 4736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533229399650653e-05,
+      "loss": 1.0688,
+      "step": 4737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532971869855357e-05,
+      "loss": 1.2341,
+      "step": 4738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532714270735232e-05,
+      "loss": 1.2624,
+      "step": 4739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532456602292148e-05,
+      "loss": 1.1206,
+      "step": 4740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532198864527983e-05,
+      "loss": 0.7148,
+      "step": 4741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531941057444606e-05,
+      "loss": 1.0562,
+      "step": 4742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531683181043895e-05,
+      "loss": 1.1078,
+      "step": 4743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953142523532773e-05,
+      "loss": 1.2049,
+      "step": 4744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953116722029798e-05,
+      "loss": 0.9954,
+      "step": 4745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953090913595652e-05,
+      "loss": 0.6419,
+      "step": 4746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9530650982305236e-05,
+      "loss": 1.0439,
+      "step": 4747
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9530392759345996e-05,
+      "loss": 1.1806,
+      "step": 4748
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.953013446708068e-05,
+      "loss": 1.2845,
+      "step": 4749
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529876105511173e-05,
+      "loss": 1.3488,
+      "step": 4750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529617674639344e-05,
+      "loss": 1.1269,
+      "step": 4751
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952935917446708e-05,
+      "loss": 1.1389,
+      "step": 4752
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529100604996256e-05,
+      "loss": 1.2468,
+      "step": 4753
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528841966228756e-05,
+      "loss": 0.76,
+      "step": 4754
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952858325816646e-05,
+      "loss": 1.1483,
+      "step": 4755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528324480811246e-05,
+      "loss": 0.9953,
+      "step": 4756
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528065634165e-05,
+      "loss": 0.9768,
+      "step": 4757
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95278067182296e-05,
+      "loss": 1.0521,
+      "step": 4758
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527547733006935e-05,
+      "loss": 0.87,
+      "step": 4759
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527288678498882e-05,
+      "loss": 0.9888,
+      "step": 4760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952702955470733e-05,
+      "loss": 0.7408,
+      "step": 4761
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952677036163416e-05,
+      "loss": 1.2036,
+      "step": 4762
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526511099281257e-05,
+      "loss": 0.9213,
+      "step": 4763
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526251767650506e-05,
+      "loss": 1.0423,
+      "step": 4764
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95259923667438e-05,
+      "loss": 1.5175,
+      "step": 4765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525732896563015e-05,
+      "loss": 1.3039,
+      "step": 4766
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525473357110043e-05,
+      "loss": 0.8179,
+      "step": 4767
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952521374838677e-05,
+      "loss": 1.0851,
+      "step": 4768
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524954070395087e-05,
+      "loss": 0.9914,
+      "step": 4769
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524694323136883e-05,
+      "loss": 1.3775,
+      "step": 4770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952443450661404e-05,
+      "loss": 1.1077,
+      "step": 4771
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524174620828452e-05,
+      "loss": 0.941,
+      "step": 4772
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952391466578201e-05,
+      "loss": 1.0522,
+      "step": 4773
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523654641476603e-05,
+      "loss": 1.3116,
+      "step": 4774
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952339454791412e-05,
+      "loss": 1.1831,
+      "step": 4775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523134385096458e-05,
+      "loss": 1.1391,
+      "step": 4776
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95228741530255e-05,
+      "loss": 1.1159,
+      "step": 4777
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952261385170315e-05,
+      "loss": 1.1175,
+      "step": 4778
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522353481131285e-05,
+      "loss": 1.0093,
+      "step": 4779
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522093041311814e-05,
+      "loss": 0.7925,
+      "step": 4780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521832532246628e-05,
+      "loss": 0.941,
+      "step": 4781
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521571953937616e-05,
+      "loss": 0.9858,
+      "step": 4782
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521311306386673e-05,
+      "loss": 0.9968,
+      "step": 4783
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95210505895957e-05,
+      "loss": 1.2006,
+      "step": 4784
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952078980356659e-05,
+      "loss": 1.0454,
+      "step": 4785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520528948301236e-05,
+      "loss": 1.2503,
+      "step": 4786
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952026802380154e-05,
+      "loss": 0.943,
+      "step": 4787
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95200070300694e-05,
+      "loss": 1.0311,
+      "step": 4788
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519745967106707e-05,
+      "loss": 1.4352,
+      "step": 4789
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519484834915366e-05,
+      "loss": 0.9281,
+      "step": 4790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519223633497273e-05,
+      "loss": 0.8217,
+      "step": 4791
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518962362854332e-05,
+      "loss": 1.0026,
+      "step": 4792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518701022988436e-05,
+      "loss": 1.1593,
+      "step": 4793
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518439613901494e-05,
+      "loss": 1.0861,
+      "step": 4794
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95181781355954e-05,
+      "loss": 0.8759,
+      "step": 4795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517916588072057e-05,
+      "loss": 0.9064,
+      "step": 4796
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517654971333367e-05,
+      "loss": 1.3071,
+      "step": 4797
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517393285381234e-05,
+      "loss": 1.1992,
+      "step": 4798
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951713153021756e-05,
+      "loss": 1.1796,
+      "step": 4799
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951686970584425e-05,
+      "loss": 0.9328,
+      "step": 4800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516607812263205e-05,
+      "loss": 1.0272,
+      "step": 4801
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951634584947633e-05,
+      "loss": 1.2689,
+      "step": 4802
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516083817485532e-05,
+      "loss": 0.6874,
+      "step": 4803
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951582171629272e-05,
+      "loss": 0.8797,
+      "step": 4804
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951555954589979e-05,
+      "loss": 1.0999,
+      "step": 4805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515297306308657e-05,
+      "loss": 1.0343,
+      "step": 4806
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515034997521224e-05,
+      "loss": 1.2241,
+      "step": 4807
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95147726195394e-05,
+      "loss": 0.9405,
+      "step": 4808
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514510172365094e-05,
+      "loss": 1.079,
+      "step": 4809
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514247656000213e-05,
+      "loss": 0.9987,
+      "step": 4810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513985070446665e-05,
+      "loss": 1.3549,
+      "step": 4811
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513722415706365e-05,
+      "loss": 1.2965,
+      "step": 4812
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513459691781217e-05,
+      "loss": 0.8123,
+      "step": 4813
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513196898673135e-05,
+      "loss": 1.0216,
+      "step": 4814
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512934036384026e-05,
+      "loss": 1.0655,
+      "step": 4815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512671104915806e-05,
+      "loss": 1.0919,
+      "step": 4816
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512408104270384e-05,
+      "loss": 0.9159,
+      "step": 4817
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512145034449676e-05,
+      "loss": 0.913,
+      "step": 4818
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951188189545559e-05,
+      "loss": 1.2366,
+      "step": 4819
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511618687290042e-05,
+      "loss": 0.8972,
+      "step": 4820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511355409954952e-05,
+      "loss": 1.1236,
+      "step": 4821
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511092063452226e-05,
+      "loss": 1.1224,
+      "step": 4822
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510828647783784e-05,
+      "loss": 1.1142,
+      "step": 4823
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510565162951538e-05,
+      "loss": 1.2977,
+      "step": 4824
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510301608957408e-05,
+      "loss": 1.1716,
+      "step": 4825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510037985803306e-05,
+      "loss": 1.0012,
+      "step": 4826
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509774293491153e-05,
+      "loss": 1.3303,
+      "step": 4827
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509510532022865e-05,
+      "loss": 0.8474,
+      "step": 4828
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950924670140036e-05,
+      "loss": 1.5529,
+      "step": 4829
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508982801625557e-05,
+      "loss": 0.961,
+      "step": 4830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508718832700375e-05,
+      "loss": 0.8202,
+      "step": 4831
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508454794626733e-05,
+      "loss": 1.0218,
+      "step": 4832
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508190687406556e-05,
+      "loss": 1.0478,
+      "step": 4833
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507926511041758e-05,
+      "loss": 1.0194,
+      "step": 4834
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950766226553426e-05,
+      "loss": 0.8473,
+      "step": 4835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950739795088599e-05,
+      "loss": 1.2282,
+      "step": 4836
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507133567098867e-05,
+      "loss": 0.8462,
+      "step": 4837
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950686911417481e-05,
+      "loss": 0.8895,
+      "step": 4838
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506604592115744e-05,
+      "loss": 0.9683,
+      "step": 4839
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506340000923597e-05,
+      "loss": 0.9018,
+      "step": 4840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950607534060029e-05,
+      "loss": 1.1058,
+      "step": 4841
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505810611147746e-05,
+      "loss": 1.1616,
+      "step": 4842
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950554581256789e-05,
+      "loss": 1.1122,
+      "step": 4843
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950528094486265e-05,
+      "loss": 0.7733,
+      "step": 4844
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505016008033953e-05,
+      "loss": 0.8454,
+      "step": 4845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504751002083724e-05,
+      "loss": 1.251,
+      "step": 4846
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950448592701389e-05,
+      "loss": 0.7443,
+      "step": 4847
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950422078282638e-05,
+      "loss": 0.9666,
+      "step": 4848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950395556952312e-05,
+      "loss": 1.0403,
+      "step": 4849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503690287106036e-05,
+      "loss": 1.2654,
+      "step": 4850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503424935577065e-05,
+      "loss": 0.9863,
+      "step": 4851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503159514938133e-05,
+      "loss": 1.2284,
+      "step": 4852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950289402519117e-05,
+      "loss": 1.0509,
+      "step": 4853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502628466338104e-05,
+      "loss": 1.2594,
+      "step": 4854
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950236283838087e-05,
+      "loss": 1.067,
+      "step": 4855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502097141321395e-05,
+      "loss": 0.8473,
+      "step": 4856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501831375161617e-05,
+      "loss": 1.2503,
+      "step": 4857
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501565539903467e-05,
+      "loss": 1.1468,
+      "step": 4858
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501299635548874e-05,
+      "loss": 1.3344,
+      "step": 4859
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950103366209978e-05,
+      "loss": 1.1332,
+      "step": 4860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500767619558113e-05,
+      "loss": 0.8846,
+      "step": 4861
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500501507925804e-05,
+      "loss": 1.0666,
+      "step": 4862
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500235327204798e-05,
+      "loss": 1.049,
+      "step": 4863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499969077397024e-05,
+      "loss": 0.7441,
+      "step": 4864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949970275850442e-05,
+      "loss": 0.9269,
+      "step": 4865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499436370528922e-05,
+      "loss": 1.0901,
+      "step": 4866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949916991347247e-05,
+      "loss": 1.1298,
+      "step": 4867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498903387336998e-05,
+      "loss": 1.1664,
+      "step": 4868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498636792124445e-05,
+      "loss": 1.0451,
+      "step": 4869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949837012783675e-05,
+      "loss": 1.3018,
+      "step": 4870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498103394475855e-05,
+      "loss": 1.1333,
+      "step": 4871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497836592043694e-05,
+      "loss": 1.1201,
+      "step": 4872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497569720542217e-05,
+      "loss": 1.1157,
+      "step": 4873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949730277997335e-05,
+      "loss": 1.3097,
+      "step": 4874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949703577033905e-05,
+      "loss": 0.931,
+      "step": 4875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496768691641245e-05,
+      "loss": 1.284,
+      "step": 4876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496501543881886e-05,
+      "loss": 1.2938,
+      "step": 4877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496234327062915e-05,
+      "loss": 1.0264,
+      "step": 4878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495967041186273e-05,
+      "loss": 1.0013,
+      "step": 4879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94956996862539e-05,
+      "loss": 1.132,
+      "step": 4880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949543226226775e-05,
+      "loss": 0.9522,
+      "step": 4881
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949516476922976e-05,
+      "loss": 0.9858,
+      "step": 4882
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494897207141877e-05,
+      "loss": 1.0013,
+      "step": 4883
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494629576006047e-05,
+      "loss": 1.1533,
+      "step": 4884
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494361875824218e-05,
+      "loss": 0.9166,
+      "step": 4885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949409410659833e-05,
+      "loss": 1.0559,
+      "step": 4886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949382626833034e-05,
+      "loss": 1.1097,
+      "step": 4887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949355836102219e-05,
+      "loss": 0.9126,
+      "step": 4888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949329038467583e-05,
+      "loss": 1.0464,
+      "step": 4889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493022339293207e-05,
+      "loss": 0.9328,
+      "step": 4890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492754224876275e-05,
+      "loss": 0.8994,
+      "step": 4891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492486041426977e-05,
+      "loss": 1.0054,
+      "step": 4892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492217788947266e-05,
+      "loss": 1.286,
+      "step": 4893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491949467439092e-05,
+      "loss": 1.294,
+      "step": 4894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491681076904408e-05,
+      "loss": 1.2954,
+      "step": 4895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491412617345168e-05,
+      "loss": 1.335,
+      "step": 4896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949114408876332e-05,
+      "loss": 1.0122,
+      "step": 4897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490875491160818e-05,
+      "loss": 1.1924,
+      "step": 4898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490606824539617e-05,
+      "loss": 1.0435,
+      "step": 4899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490338088901667e-05,
+      "loss": 0.852,
+      "step": 4900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490069284248926e-05,
+      "loss": 1.0023,
+      "step": 4901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489800410583345e-05,
+      "loss": 1.0662,
+      "step": 4902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489531467906887e-05,
+      "loss": 0.831,
+      "step": 4903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489262456221496e-05,
+      "loss": 1.0192,
+      "step": 4904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488993375529137e-05,
+      "loss": 0.8596,
+      "step": 4905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488724225831765e-05,
+      "loss": 1.0998,
+      "step": 4906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488455007131338e-05,
+      "loss": 0.9553,
+      "step": 4907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948818571942981e-05,
+      "loss": 0.9684,
+      "step": 4908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948791636272914e-05,
+      "loss": 1.099,
+      "step": 4909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487646937031294e-05,
+      "loss": 1.2315,
+      "step": 4910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487377442338223e-05,
+      "loss": 1.0068,
+      "step": 4911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948710787865189e-05,
+      "loss": 1.2403,
+      "step": 4912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486838245974257e-05,
+      "loss": 0.9395,
+      "step": 4913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486568544307278e-05,
+      "loss": 1.1649,
+      "step": 4914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486298773652922e-05,
+      "loss": 1.1761,
+      "step": 4915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486028934013147e-05,
+      "loss": 1.0296,
+      "step": 4916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485759025389918e-05,
+      "loss": 1.2673,
+      "step": 4917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485489047785196e-05,
+      "loss": 0.7548,
+      "step": 4918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948521900120094e-05,
+      "loss": 1.1398,
+      "step": 4919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484948885639122e-05,
+      "loss": 1.0908,
+      "step": 4920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484678701101703e-05,
+      "loss": 1.0278,
+      "step": 4921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484408447590645e-05,
+      "loss": 1.1323,
+      "step": 4922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484138125107918e-05,
+      "loss": 1.0648,
+      "step": 4923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483867733655485e-05,
+      "loss": 1.252,
+      "step": 4924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483597273235308e-05,
+      "loss": 0.9776,
+      "step": 4925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483326743849364e-05,
+      "loss": 1.0583,
+      "step": 4926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483056145499613e-05,
+      "loss": 1.0059,
+      "step": 4927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482785478188024e-05,
+      "loss": 0.9683,
+      "step": 4928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948251474191657e-05,
+      "loss": 0.8862,
+      "step": 4929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948224393668721e-05,
+      "loss": 1.1363,
+      "step": 4930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948197306250192e-05,
+      "loss": 0.9578,
+      "step": 4931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481702119362672e-05,
+      "loss": 0.8588,
+      "step": 4932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948143110727143e-05,
+      "loss": 0.749,
+      "step": 4933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948116002623017e-05,
+      "loss": 0.9413,
+      "step": 4934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948088887624086e-05,
+      "loss": 1.2891,
+      "step": 4935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480617657305476e-05,
+      "loss": 0.8907,
+      "step": 4936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480346369425986e-05,
+      "loss": 0.8878,
+      "step": 4937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9480075012604364e-05,
+      "loss": 0.8787,
+      "step": 4938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479803586842584e-05,
+      "loss": 1.0625,
+      "step": 4939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479532092142622e-05,
+      "loss": 1.1659,
+      "step": 4940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479260528506447e-05,
+      "loss": 1.2241,
+      "step": 4941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478988895936037e-05,
+      "loss": 0.9674,
+      "step": 4942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478717194433367e-05,
+      "loss": 1.0012,
+      "step": 4943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478445424000414e-05,
+      "loss": 0.8481,
+      "step": 4944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478173584639154e-05,
+      "loss": 0.6929,
+      "step": 4945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947790167635156e-05,
+      "loss": 0.8655,
+      "step": 4946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477629699139616e-05,
+      "loss": 0.7733,
+      "step": 4947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477357653005295e-05,
+      "loss": 0.8687,
+      "step": 4948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477085537950577e-05,
+      "loss": 0.9831,
+      "step": 4949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476813353977442e-05,
+      "loss": 0.8658,
+      "step": 4950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476541101087864e-05,
+      "loss": 1.2427,
+      "step": 4951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476268779283828e-05,
+      "loss": 1.1579,
+      "step": 4952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475996388567315e-05,
+      "loss": 1.2543,
+      "step": 4953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475723928940305e-05,
+      "loss": 0.9352,
+      "step": 4954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475451400404776e-05,
+      "loss": 1.0754,
+      "step": 4955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947517880296271e-05,
+      "loss": 1.0817,
+      "step": 4956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474906136616096e-05,
+      "loss": 1.0697,
+      "step": 4957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947463340136691e-05,
+      "loss": 1.1614,
+      "step": 4958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947436059721714e-05,
+      "loss": 0.8762,
+      "step": 4959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474087724168767e-05,
+      "loss": 0.9263,
+      "step": 4960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473814782223775e-05,
+      "loss": 0.8883,
+      "step": 4961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473541771384148e-05,
+      "loss": 0.9875,
+      "step": 4962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473268691651873e-05,
+      "loss": 0.8479,
+      "step": 4963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947299554302894e-05,
+      "loss": 0.9365,
+      "step": 4964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947272232551733e-05,
+      "loss": 1.4421,
+      "step": 4965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947244903911903e-05,
+      "loss": 1.0603,
+      "step": 4966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947217568383603e-05,
+      "loss": 1.0553,
+      "step": 4967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471902259670318e-05,
+      "loss": 0.8944,
+      "step": 4968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471628766623877e-05,
+      "loss": 1.2102,
+      "step": 4969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471355204698703e-05,
+      "loss": 0.9489,
+      "step": 4970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947108157389678e-05,
+      "loss": 1.4459,
+      "step": 4971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470807874220102e-05,
+      "loss": 0.9239,
+      "step": 4972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947053410567066e-05,
+      "loss": 0.9928,
+      "step": 4973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470260268250435e-05,
+      "loss": 1.0117,
+      "step": 4974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946998636196143e-05,
+      "loss": 1.0314,
+      "step": 4975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469712386805632e-05,
+      "loss": 1.0596,
+      "step": 4976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469438342785033e-05,
+      "loss": 1.1083,
+      "step": 4977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946916422990163e-05,
+      "loss": 1.0281,
+      "step": 4978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946889004815741e-05,
+      "loss": 1.2156,
+      "step": 4979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468615797554374e-05,
+      "loss": 1.0378,
+      "step": 4980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468341478094506e-05,
+      "loss": 1.1218,
+      "step": 4981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946806708977981e-05,
+      "loss": 1.0057,
+      "step": 4982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467792632612284e-05,
+      "loss": 0.7008,
+      "step": 4983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946751810659391e-05,
+      "loss": 0.9287,
+      "step": 4984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467243511726698e-05,
+      "loss": 1.2564,
+      "step": 4985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466968848012643e-05,
+      "loss": 0.924,
+      "step": 4986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466694115453736e-05,
+      "loss": 0.975,
+      "step": 4987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466419314051976e-05,
+      "loss": 1.0267,
+      "step": 4988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466144443809364e-05,
+      "loss": 1.0329,
+      "step": 4989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94658695047279e-05,
+      "loss": 1.1914,
+      "step": 4990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946559449680958e-05,
+      "loss": 1.3332,
+      "step": 4991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946531942005641e-05,
+      "loss": 0.9546,
+      "step": 4992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946504427447038e-05,
+      "loss": 1.0632,
+      "step": 4993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94647690600535e-05,
+      "loss": 0.8559,
+      "step": 4994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946449377680777e-05,
+      "loss": 1.3187,
+      "step": 4995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946421842473519e-05,
+      "loss": 1.3224,
+      "step": 4996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946394300383776e-05,
+      "loss": 1.7894,
+      "step": 4997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463667514117487e-05,
+      "loss": 1.0329,
+      "step": 4998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463391955576373e-05,
+      "loss": 1.0548,
+      "step": 4999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946311632821642e-05,
+      "loss": 1.3118,
+      "step": 5000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462840632039633e-05,
+      "loss": 1.109,
+      "step": 5001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462564867048025e-05,
+      "loss": 1.0319,
+      "step": 5002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462289033243592e-05,
+      "loss": 1.0992,
+      "step": 5003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946201313062834e-05,
+      "loss": 0.8437,
+      "step": 5004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461737159204276e-05,
+      "loss": 1.1191,
+      "step": 5005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946146111897341e-05,
+      "loss": 1.0709,
+      "step": 5006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946118500993775e-05,
+      "loss": 0.8553,
+      "step": 5007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94609088320993e-05,
+      "loss": 1.0553,
+      "step": 5008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460632585460075e-05,
+      "loss": 1.0859,
+      "step": 5009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460356270022073e-05,
+      "loss": 1.1167,
+      "step": 5010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460079885787313e-05,
+      "loss": 0.9404,
+      "step": 5011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94598034327578e-05,
+      "loss": 1.1789,
+      "step": 5012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459526910935552e-05,
+      "loss": 1.2224,
+      "step": 5013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459250320322567e-05,
+      "loss": 1.0975,
+      "step": 5014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458973660920866e-05,
+      "loss": 1.2387,
+      "step": 5015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458696932732457e-05,
+      "loss": 1.2323,
+      "step": 5016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458420135759352e-05,
+      "loss": 0.9883,
+      "step": 5017
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458143270003567e-05,
+      "loss": 1.3296,
+      "step": 5018
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457866335467113e-05,
+      "loss": 0.9349,
+      "step": 5019
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457589332152008e-05,
+      "loss": 0.8797,
+      "step": 5020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945731226006026e-05,
+      "loss": 0.8715,
+      "step": 5021
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457035119193886e-05,
+      "loss": 0.9229,
+      "step": 5022
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456757909554908e-05,
+      "loss": 1.1045,
+      "step": 5023
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945648063114533e-05,
+      "loss": 0.7916,
+      "step": 5024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945620328396718e-05,
+      "loss": 1.1448,
+      "step": 5025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455925868022464e-05,
+      "loss": 1.0918,
+      "step": 5026
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455648383313212e-05,
+      "loss": 1.1062,
+      "step": 5027
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455370829841432e-05,
+      "loss": 0.992,
+      "step": 5028
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455093207609144e-05,
+      "loss": 1.1831,
+      "step": 5029
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454815516618368e-05,
+      "loss": 1.0563,
+      "step": 5030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454537756871124e-05,
+      "loss": 1.2635,
+      "step": 5031
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945425992836943e-05,
+      "loss": 0.8927,
+      "step": 5032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945398203111531e-05,
+      "loss": 1.1217,
+      "step": 5033
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453704065110783e-05,
+      "loss": 1.2491,
+      "step": 5034
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453426030357868e-05,
+      "loss": 1.6281,
+      "step": 5035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453147926858593e-05,
+      "loss": 1.0044,
+      "step": 5036
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452869754614975e-05,
+      "loss": 1.1212,
+      "step": 5037
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945259151362904e-05,
+      "loss": 1.1982,
+      "step": 5038
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452313203902805e-05,
+      "loss": 1.1816,
+      "step": 5039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452034825438302e-05,
+      "loss": 1.2722,
+      "step": 5040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945175637823755e-05,
+      "loss": 1.2588,
+      "step": 5041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945147786230258e-05,
+      "loss": 1.2294,
+      "step": 5042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945119927763541e-05,
+      "loss": 1.0059,
+      "step": 5043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945092062423807e-05,
+      "loss": 1.4592,
+      "step": 5044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450641902112585e-05,
+      "loss": 0.9861,
+      "step": 5045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450363111260984e-05,
+      "loss": 1.2124,
+      "step": 5046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450084251685295e-05,
+      "loss": 1.273,
+      "step": 5047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449805323387542e-05,
+      "loss": 1.3639,
+      "step": 5048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449526326369752e-05,
+      "loss": 1.3969,
+      "step": 5049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944924726063396e-05,
+      "loss": 0.9014,
+      "step": 5050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448968126182195e-05,
+      "loss": 0.8352,
+      "step": 5051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944868892301648e-05,
+      "loss": 0.8284,
+      "step": 5052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9448409651138855e-05,
+      "loss": 0.8398,
+      "step": 5053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944813031055134e-05,
+      "loss": 1.1121,
+      "step": 5054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447850901255975e-05,
+      "loss": 1.1595,
+      "step": 5055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447571423254792e-05,
+      "loss": 1.2767,
+      "step": 5056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447291876549817e-05,
+      "loss": 0.8521,
+      "step": 5057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447012261143085e-05,
+      "loss": 0.9901,
+      "step": 5058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446732577036634e-05,
+      "loss": 1.0717,
+      "step": 5059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446452824232494e-05,
+      "loss": 1.0964,
+      "step": 5060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94461730027327e-05,
+      "loss": 0.9842,
+      "step": 5061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445893112539288e-05,
+      "loss": 1.0104,
+      "step": 5062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944561315365429e-05,
+      "loss": 1.1301,
+      "step": 5063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445333126079745e-05,
+      "loss": 0.9284,
+      "step": 5064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445053029817687e-05,
+      "loss": 0.7421,
+      "step": 5065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444772864870157e-05,
+      "loss": 1.2092,
+      "step": 5066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944449263123919e-05,
+      "loss": 1.2801,
+      "step": 5067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444212328926823e-05,
+      "loss": 1.1892,
+      "step": 5068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443931957935093e-05,
+      "loss": 1.1841,
+      "step": 5069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443651518266044e-05,
+      "loss": 0.6516,
+      "step": 5070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443371009921712e-05,
+      "loss": 0.9988,
+      "step": 5071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443090432904136e-05,
+      "loss": 1.2687,
+      "step": 5072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944280978721536e-05,
+      "loss": 0.9462,
+      "step": 5073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442529072857422e-05,
+      "loss": 1.0729,
+      "step": 5074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442248289832363e-05,
+      "loss": 1.0121,
+      "step": 5075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441967438142226e-05,
+      "loss": 1.0474,
+      "step": 5076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441686517789054e-05,
+      "loss": 1.2609,
+      "step": 5077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441405528774887e-05,
+      "loss": 0.9081,
+      "step": 5078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441124471101773e-05,
+      "loss": 1.1379,
+      "step": 5079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944084334477175e-05,
+      "loss": 1.2162,
+      "step": 5080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944056214978687e-05,
+      "loss": 1.5521,
+      "step": 5081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440280886149176e-05,
+      "loss": 0.9711,
+      "step": 5082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439999553860706e-05,
+      "loss": 0.8889,
+      "step": 5083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439718152923513e-05,
+      "loss": 0.8306,
+      "step": 5084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943943668333964e-05,
+      "loss": 1.0317,
+      "step": 5085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439155145111135e-05,
+      "loss": 0.9268,
+      "step": 5086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438873538240046e-05,
+      "loss": 0.8078,
+      "step": 5087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438591862728423e-05,
+      "loss": 1.0101,
+      "step": 5088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438310118578308e-05,
+      "loss": 1.3625,
+      "step": 5089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438028305791758e-05,
+      "loss": 0.8007,
+      "step": 5090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437746424370812e-05,
+      "loss": 1.0185,
+      "step": 5091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437464474317532e-05,
+      "loss": 1.074,
+      "step": 5092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437182455633962e-05,
+      "loss": 1.0792,
+      "step": 5093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943690036832215e-05,
+      "loss": 0.9365,
+      "step": 5094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943661821238415e-05,
+      "loss": 1.2017,
+      "step": 5095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436335987822015e-05,
+      "loss": 1.2233,
+      "step": 5096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436053694637798e-05,
+      "loss": 1.5062,
+      "step": 5097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943577133283355e-05,
+      "loss": 0.7595,
+      "step": 5098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943548890241133e-05,
+      "loss": 1.0797,
+      "step": 5099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943520640337318e-05,
+      "loss": 0.994,
+      "step": 5100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434923835721163e-05,
+      "loss": 0.6896,
+      "step": 5101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434641199457334e-05,
+      "loss": 0.8153,
+      "step": 5102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434358494583744e-05,
+      "loss": 1.0734,
+      "step": 5103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943407572110245e-05,
+      "loss": 1.098,
+      "step": 5104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433792879015517e-05,
+      "loss": 1.3584,
+      "step": 5105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943350996832499e-05,
+      "loss": 1.0897,
+      "step": 5106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943322698903293e-05,
+      "loss": 1.2601,
+      "step": 5107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432943941141397e-05,
+      "loss": 1.0307,
+      "step": 5108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432660824652448e-05,
+      "loss": 1.0091,
+      "step": 5109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432377639568142e-05,
+      "loss": 0.9346,
+      "step": 5110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432094385890542e-05,
+      "loss": 1.0899,
+      "step": 5111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94318110636217e-05,
+      "loss": 1.4505,
+      "step": 5112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943152767276368e-05,
+      "loss": 1.0626,
+      "step": 5113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9431244213318544e-05,
+      "loss": 0.9833,
+      "step": 5114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430960685288355e-05,
+      "loss": 0.895,
+      "step": 5115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943067708867517e-05,
+      "loss": 1.1715,
+      "step": 5116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430393423481054e-05,
+      "loss": 1.031,
+      "step": 5117
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943010968970807e-05,
+      "loss": 0.9788,
+      "step": 5118
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429825887358285e-05,
+      "loss": 0.9369,
+      "step": 5119
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429542016433753e-05,
+      "loss": 1.1721,
+      "step": 5120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429258076936547e-05,
+      "loss": 1.0141,
+      "step": 5121
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942897406886873e-05,
+      "loss": 0.9228,
+      "step": 5122
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428689992232364e-05,
+      "loss": 1.4541,
+      "step": 5123
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428405847029524e-05,
+      "loss": 1.2033,
+      "step": 5124
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428121633262264e-05,
+      "loss": 1.0765,
+      "step": 5125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9427837350932657e-05,
+      "loss": 0.9244,
+      "step": 5126
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942755300004277e-05,
+      "loss": 0.9585,
+      "step": 5127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942726858059467e-05,
+      "loss": 1.0823,
+      "step": 5128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426984092590427e-05,
+      "loss": 1.1196,
+      "step": 5129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942669953603211e-05,
+      "loss": 1.1275,
+      "step": 5130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426414910921785e-05,
+      "loss": 1.0755,
+      "step": 5131
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9426130217261524e-05,
+      "loss": 0.9121,
+      "step": 5132
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94258454550534e-05,
+      "loss": 0.7191,
+      "step": 5133
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942556062429948e-05,
+      "loss": 1.1717,
+      "step": 5134
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425275725001836e-05,
+      "loss": 1.054,
+      "step": 5135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424990757162542e-05,
+      "loss": 1.1301,
+      "step": 5136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424705720783663e-05,
+      "loss": 1.2703,
+      "step": 5137
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942442061586728e-05,
+      "loss": 1.135,
+      "step": 5138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424135442415468e-05,
+      "loss": 0.992,
+      "step": 5139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423850200430295e-05,
+      "loss": 1.0662,
+      "step": 5140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423564889913834e-05,
+      "loss": 0.9821,
+      "step": 5141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423279510868163e-05,
+      "loss": 1.0506,
+      "step": 5142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422994063295357e-05,
+      "loss": 1.0789,
+      "step": 5143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422708547197495e-05,
+      "loss": 1.0964,
+      "step": 5144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422422962576646e-05,
+      "loss": 1.1068,
+      "step": 5145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942213730943489e-05,
+      "loss": 1.2717,
+      "step": 5146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421851587774313e-05,
+      "loss": 1.1529,
+      "step": 5147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942156579759698e-05,
+      "loss": 0.6869,
+      "step": 5148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421279938904972e-05,
+      "loss": 0.6554,
+      "step": 5149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942099401170037e-05,
+      "loss": 1.2012,
+      "step": 5150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420708015985257e-05,
+      "loss": 1.2708,
+      "step": 5151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942042195176171e-05,
+      "loss": 1.061,
+      "step": 5152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420135819031803e-05,
+      "loss": 1.1398,
+      "step": 5153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941984961779763e-05,
+      "loss": 1.3036,
+      "step": 5154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419563348061256e-05,
+      "loss": 1.2152,
+      "step": 5155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941927700982478e-05,
+      "loss": 0.9597,
+      "step": 5156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418990603090265e-05,
+      "loss": 0.827,
+      "step": 5157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941870412785981e-05,
+      "loss": 1.0941,
+      "step": 5158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418417584135496e-05,
+      "loss": 0.8276,
+      "step": 5159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94181309719194e-05,
+      "loss": 0.8812,
+      "step": 5160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941784429121361e-05,
+      "loss": 1.2461,
+      "step": 5161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417557542020212e-05,
+      "loss": 1.2439,
+      "step": 5162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417270724341284e-05,
+      "loss": 1.2071,
+      "step": 5163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416983838178923e-05,
+      "loss": 1.0846,
+      "step": 5164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941669688353521e-05,
+      "loss": 1.2422,
+      "step": 5165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941640986041223e-05,
+      "loss": 1.1009,
+      "step": 5166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416122768812072e-05,
+      "loss": 1.2073,
+      "step": 5167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415835608736823e-05,
+      "loss": 0.8986,
+      "step": 5168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415548380188572e-05,
+      "loss": 1.2802,
+      "step": 5169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941526108316941e-05,
+      "loss": 1.0446,
+      "step": 5170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414973717681423e-05,
+      "loss": 0.9379,
+      "step": 5171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414686283726697e-05,
+      "loss": 0.9213,
+      "step": 5172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941439878130733e-05,
+      "loss": 1.1167,
+      "step": 5173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941411121042541e-05,
+      "loss": 1.5169,
+      "step": 5174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941382357108303e-05,
+      "loss": 1.0404,
+      "step": 5175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413535863282277e-05,
+      "loss": 1.0898,
+      "step": 5176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413248087025246e-05,
+      "loss": 1.3726,
+      "step": 5177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941296024231403e-05,
+      "loss": 1.175,
+      "step": 5178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941267232915072e-05,
+      "loss": 1.0671,
+      "step": 5179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412384347537415e-05,
+      "loss": 0.8733,
+      "step": 5180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412096297476205e-05,
+      "loss": 0.9273,
+      "step": 5181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411808178969186e-05,
+      "loss": 1.2588,
+      "step": 5182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941151999201845e-05,
+      "loss": 1.1438,
+      "step": 5183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94112317366261e-05,
+      "loss": 1.0593,
+      "step": 5184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410943412794224e-05,
+      "loss": 1.0451,
+      "step": 5185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941065502052492e-05,
+      "loss": 0.8965,
+      "step": 5186
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410366559820294e-05,
+      "loss": 0.8915,
+      "step": 5187
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410078030682432e-05,
+      "loss": 1.073,
+      "step": 5188
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940978943311344e-05,
+      "loss": 0.9596,
+      "step": 5189
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409500767115414e-05,
+      "loss": 1.0894,
+      "step": 5190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940921203269045e-05,
+      "loss": 1.1676,
+      "step": 5191
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408923229840658e-05,
+      "loss": 0.9965,
+      "step": 5192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408634358568127e-05,
+      "loss": 1.1382,
+      "step": 5193
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408345418874965e-05,
+      "loss": 0.8813,
+      "step": 5194
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940805641076327e-05,
+      "loss": 0.9484,
+      "step": 5195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407767334235142e-05,
+      "loss": 0.957,
+      "step": 5196
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407478189292687e-05,
+      "loss": 1.171,
+      "step": 5197
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407188975938002e-05,
+      "loss": 0.9281,
+      "step": 5198
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406899694173197e-05,
+      "loss": 1.1934,
+      "step": 5199
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406610344000374e-05,
+      "loss": 1.1416,
+      "step": 5200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406320925421634e-05,
+      "loss": 1.0036,
+      "step": 5201
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406031438439085e-05,
+      "loss": 1.3366,
+      "step": 5202
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940574188305483e-05,
+      "loss": 1.1477,
+      "step": 5203
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405452259270976e-05,
+      "loss": 1.1519,
+      "step": 5204
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405162567089627e-05,
+      "loss": 0.9518,
+      "step": 5205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404872806512895e-05,
+      "loss": 1.0482,
+      "step": 5206
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940458297754288e-05,
+      "loss": 0.9776,
+      "step": 5207
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404293080181698e-05,
+      "loss": 1.1454,
+      "step": 5208
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404003114431448e-05,
+      "loss": 1.1536,
+      "step": 5209
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403713080294245e-05,
+      "loss": 1.1497,
+      "step": 5210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9403422977772196e-05,
+      "loss": 1.0785,
+      "step": 5211
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940313280686741e-05,
+      "loss": 1.2168,
+      "step": 5212
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402842567581997e-05,
+      "loss": 1.2153,
+      "step": 5213
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402552259918072e-05,
+      "loss": 1.3836,
+      "step": 5214
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940226188387774e-05,
+      "loss": 1.1081,
+      "step": 5215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401971439463118e-05,
+      "loss": 1.0964,
+      "step": 5216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401680926676314e-05,
+      "loss": 0.8167,
+      "step": 5217
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940139034551944e-05,
+      "loss": 1.1534,
+      "step": 5218
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401099695994613e-05,
+      "loss": 1.2871,
+      "step": 5219
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400808978103948e-05,
+      "loss": 1.055,
+      "step": 5220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940051819184955e-05,
+      "loss": 1.0666,
+      "step": 5221
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400227337233545e-05,
+      "loss": 1.1654,
+      "step": 5222
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399936414258043e-05,
+      "loss": 1.1205,
+      "step": 5223
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939964542292516e-05,
+      "loss": 1.3118,
+      "step": 5224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939935436323701e-05,
+      "loss": 1.0742,
+      "step": 5225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939906323519571e-05,
+      "loss": 1.2017,
+      "step": 5226
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939877203880338e-05,
+      "loss": 1.066,
+      "step": 5227
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398480774062136e-05,
+      "loss": 1.1595,
+      "step": 5228
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398189440974097e-05,
+      "loss": 1.1386,
+      "step": 5229
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939789803954138e-05,
+      "loss": 0.9082,
+      "step": 5230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397606569766102e-05,
+      "loss": 1.0436,
+      "step": 5231
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397315031650388e-05,
+      "loss": 0.9075,
+      "step": 5232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9397023425196354e-05,
+      "loss": 1.0173,
+      "step": 5233
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396731750406127e-05,
+      "loss": 1.0241,
+      "step": 5234
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939644000728182e-05,
+      "loss": 1.1538,
+      "step": 5235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396148195825554e-05,
+      "loss": 1.0992,
+      "step": 5236
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939585631603946e-05,
+      "loss": 0.8087,
+      "step": 5237
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395564367925653e-05,
+      "loss": 1.1471,
+      "step": 5238
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939527235148626e-05,
+      "loss": 1.2873,
+      "step": 5239
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394980266723406e-05,
+      "loss": 0.8073,
+      "step": 5240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394688113639207e-05,
+      "loss": 1.0504,
+      "step": 5241
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394395892235797e-05,
+      "loss": 1.0953,
+      "step": 5242
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394103602515294e-05,
+      "loss": 0.8953,
+      "step": 5243
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393811244479825e-05,
+      "loss": 1.1247,
+      "step": 5244
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393518818131518e-05,
+      "loss": 0.8608,
+      "step": 5245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93932263234725e-05,
+      "loss": 1.2059,
+      "step": 5246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93929337605049e-05,
+      "loss": 1.1109,
+      "step": 5247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939264112923084e-05,
+      "loss": 0.7922,
+      "step": 5248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939234842965245e-05,
+      "loss": 1.1784,
+      "step": 5249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939205566177186e-05,
+      "loss": 1.015,
+      "step": 5250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391762825591198e-05,
+      "loss": 1.3004,
+      "step": 5251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391469921112596e-05,
+      "loss": 1.0306,
+      "step": 5252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939117694833818e-05,
+      "loss": 1.0202,
+      "step": 5253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390883907270083e-05,
+      "loss": 1.1938,
+      "step": 5254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939059079791043e-05,
+      "loss": 1.0124,
+      "step": 5255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390297620261368e-05,
+      "loss": 0.9655,
+      "step": 5256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939000437432501e-05,
+      "loss": 1.1836,
+      "step": 5257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93897110601035e-05,
+      "loss": 0.9094,
+      "step": 5258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389417677598972e-05,
+      "loss": 1.137,
+      "step": 5259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9389124226813552e-05,
+      "loss": 1.0265,
+      "step": 5260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388830707749382e-05,
+      "loss": 1.1532,
+      "step": 5261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388537120408587e-05,
+      "loss": 0.976,
+      "step": 5262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938824346479331e-05,
+      "loss": 0.9738,
+      "step": 5263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387949740905684e-05,
+      "loss": 0.9509,
+      "step": 5264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938765594874785e-05,
+      "loss": 1.1782,
+      "step": 5265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387362088321934e-05,
+      "loss": 0.8688,
+      "step": 5266
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938706815963008e-05,
+      "loss": 0.9863,
+      "step": 5267
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386774162674423e-05,
+      "loss": 1.2182,
+      "step": 5268
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93864800974571e-05,
+      "loss": 1.2445,
+      "step": 5269
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386185963980252e-05,
+      "loss": 1.1158,
+      "step": 5270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385891762246022e-05,
+      "loss": 1.5025,
+      "step": 5271
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938559749225654e-05,
+      "loss": 0.9854,
+      "step": 5272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385303154013955e-05,
+      "loss": 1.2856,
+      "step": 5273
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385008747520396e-05,
+      "loss": 1.2128,
+      "step": 5274
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938471427277802e-05,
+      "loss": 1.2684,
+      "step": 5275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384419729788956e-05,
+      "loss": 1.1667,
+      "step": 5276
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384125118555347e-05,
+      "loss": 0.9825,
+      "step": 5277
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938383043907934e-05,
+      "loss": 1.5295,
+      "step": 5278
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383535691363078e-05,
+      "loss": 1.1441,
+      "step": 5279
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93832408754087e-05,
+      "loss": 0.9959,
+      "step": 5280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382945991218355e-05,
+      "loss": 1.033,
+      "step": 5281
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938265103879418e-05,
+      "loss": 1.1133,
+      "step": 5282
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938235601813833e-05,
+      "loss": 0.9572,
+      "step": 5283
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938206092925294e-05,
+      "loss": 0.9936,
+      "step": 5284
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381765772140167e-05,
+      "loss": 0.898,
+      "step": 5285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381470546802145e-05,
+      "loss": 1.1755,
+      "step": 5286
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938117525324103e-05,
+      "loss": 1.1378,
+      "step": 5287
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938087989145897e-05,
+      "loss": 0.9103,
+      "step": 5288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380584461458107e-05,
+      "loss": 0.8301,
+      "step": 5289
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938028896324059e-05,
+      "loss": 0.9578,
+      "step": 5290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379993396808574e-05,
+      "loss": 0.8775,
+      "step": 5291
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379697762164205e-05,
+      "loss": 1.1756,
+      "step": 5292
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379402059309628e-05,
+      "loss": 1.447,
+      "step": 5293
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9379106288247e-05,
+      "loss": 1.2115,
+      "step": 5294
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937881044897847e-05,
+      "loss": 1.5414,
+      "step": 5295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378514541506187e-05,
+      "loss": 1.1356,
+      "step": 5296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378218565832306e-05,
+      "loss": 1.1815,
+      "step": 5297
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937792252195898e-05,
+      "loss": 0.9324,
+      "step": 5298
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377626409888357e-05,
+      "loss": 1.5794,
+      "step": 5299
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377330229622596e-05,
+      "loss": 1.3168,
+      "step": 5300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377033981163845e-05,
+      "loss": 1.3442,
+      "step": 5301
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376737664514262e-05,
+      "loss": 1.2416,
+      "step": 5302
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376441279676003e-05,
+      "loss": 1.0451,
+      "step": 5303
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376144826651223e-05,
+      "loss": 1.0555,
+      "step": 5304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375848305442076e-05,
+      "loss": 1.0211,
+      "step": 5305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375551716050717e-05,
+      "loss": 1.1661,
+      "step": 5306
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375255058479308e-05,
+      "loss": 1.2443,
+      "step": 5307
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374958332729997e-05,
+      "loss": 0.954,
+      "step": 5308
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374661538804955e-05,
+      "loss": 1.0639,
+      "step": 5309
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937436467670633e-05,
+      "loss": 0.9845,
+      "step": 5310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9374067746436283e-05,
+      "loss": 1.0059,
+      "step": 5311
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937377074799698e-05,
+      "loss": 1.1115,
+      "step": 5312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373473681390565e-05,
+      "loss": 1.5924,
+      "step": 5313
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373176546619218e-05,
+      "loss": 0.829,
+      "step": 5314
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372879343685087e-05,
+      "loss": 0.989,
+      "step": 5315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372582072590336e-05,
+      "loss": 0.9147,
+      "step": 5316
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372284733337128e-05,
+      "loss": 1.3769,
+      "step": 5317
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371987325927624e-05,
+      "loss": 1.1606,
+      "step": 5318
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371689850363985e-05,
+      "loss": 0.9496,
+      "step": 5319
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371392306648384e-05,
+      "loss": 1.5065,
+      "step": 5320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937109469478297e-05,
+      "loss": 1.3452,
+      "step": 5321
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937079701476992e-05,
+      "loss": 0.9534,
+      "step": 5322
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937049926661139e-05,
+      "loss": 1.2796,
+      "step": 5323
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9370201450309548e-05,
+      "loss": 1.0734,
+      "step": 5324
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369903565866565e-05,
+      "loss": 1.2633,
+      "step": 5325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93696056132846e-05,
+      "loss": 0.9192,
+      "step": 5326
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369307592565822e-05,
+      "loss": 0.9182,
+      "step": 5327
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93690095037124e-05,
+      "loss": 0.8234,
+      "step": 5328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93687113467265e-05,
+      "loss": 1.2495,
+      "step": 5329
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936841312161029e-05,
+      "loss": 1.1841,
+      "step": 5330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936811482836594e-05,
+      "loss": 0.968,
+      "step": 5331
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367816466995618e-05,
+      "loss": 1.2322,
+      "step": 5332
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367518037501496e-05,
+      "loss": 1.1989,
+      "step": 5333
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936721953988574e-05,
+      "loss": 0.6711,
+      "step": 5334
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366920974150526e-05,
+      "loss": 0.9362,
+      "step": 5335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366622340298023e-05,
+      "loss": 1.0424,
+      "step": 5336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93663236383304e-05,
+      "loss": 0.9297,
+      "step": 5337
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366024868249835e-05,
+      "loss": 0.8935,
+      "step": 5338
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936572603005849e-05,
+      "loss": 1.2151,
+      "step": 5339
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936542712375855e-05,
+      "loss": 1.1743,
+      "step": 5340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365128149352185e-05,
+      "loss": 0.9871,
+      "step": 5341
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364829106841565e-05,
+      "loss": 1.2913,
+      "step": 5342
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936452999622887e-05,
+      "loss": 1.1142,
+      "step": 5343
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364230817516274e-05,
+      "loss": 1.1082,
+      "step": 5344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936393157070595e-05,
+      "loss": 1.1815,
+      "step": 5345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363632255800074e-05,
+      "loss": 0.9923,
+      "step": 5346
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363332872800827e-05,
+      "loss": 1.0274,
+      "step": 5347
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363033421710384e-05,
+      "loss": 0.9045,
+      "step": 5348
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362733902530922e-05,
+      "loss": 1.1619,
+      "step": 5349
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362434315264615e-05,
+      "loss": 0.9037,
+      "step": 5350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936213465991365e-05,
+      "loss": 1.0331,
+      "step": 5351
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93618349364802e-05,
+      "loss": 1.1212,
+      "step": 5352
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936153514496645e-05,
+      "loss": 1.0427,
+      "step": 5353
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361235285374574e-05,
+      "loss": 0.9979,
+      "step": 5354
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9360935357706756e-05,
+      "loss": 1.0112,
+      "step": 5355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936063536196518e-05,
+      "loss": 1.1072,
+      "step": 5356
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936033529815202e-05,
+      "loss": 1.2011,
+      "step": 5357
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936003516626946e-05,
+      "loss": 1.0339,
+      "step": 5358
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359734966319693e-05,
+      "loss": 1.3053,
+      "step": 5359
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359434698304888e-05,
+      "loss": 0.9911,
+      "step": 5360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359134362227235e-05,
+      "loss": 1.5499,
+      "step": 5361
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935883395808892e-05,
+      "loss": 1.199,
+      "step": 5362
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935853348589212e-05,
+      "loss": 0.9348,
+      "step": 5363
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358232945639032e-05,
+      "loss": 1.1119,
+      "step": 5364
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935793233733183e-05,
+      "loss": 0.8765,
+      "step": 5365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357631660972708e-05,
+      "loss": 0.9144,
+      "step": 5366
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935733091656385e-05,
+      "loss": 1.0291,
+      "step": 5367
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357030104107437e-05,
+      "loss": 1.1434,
+      "step": 5368
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356729223605666e-05,
+      "loss": 1.0866,
+      "step": 5369
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356428275060722e-05,
+      "loss": 1.6098,
+      "step": 5370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935612725847479e-05,
+      "loss": 1.1115,
+      "step": 5371
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355826173850063e-05,
+      "loss": 0.8069,
+      "step": 5372
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935552502118873e-05,
+      "loss": 1.0634,
+      "step": 5373
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355223800492976e-05,
+      "loss": 1.0268,
+      "step": 5374
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354922511765e-05,
+      "loss": 1.038,
+      "step": 5375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935462115500699e-05,
+      "loss": 0.7564,
+      "step": 5376
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935431973022113e-05,
+      "loss": 1.0304,
+      "step": 5377
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354018237409624e-05,
+      "loss": 1.3329,
+      "step": 5378
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353716676574656e-05,
+      "loss": 1.035,
+      "step": 5379
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935341504771842e-05,
+      "loss": 1.0043,
+      "step": 5380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353113350843112e-05,
+      "loss": 1.022,
+      "step": 5381
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352811585950927e-05,
+      "loss": 1.2608,
+      "step": 5382
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352509753044057e-05,
+      "loss": 0.8325,
+      "step": 5383
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352207852124696e-05,
+      "loss": 0.8653,
+      "step": 5384
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351905883195044e-05,
+      "loss": 1.0311,
+      "step": 5385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935160384625729e-05,
+      "loss": 0.9529,
+      "step": 5386
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935130174131364e-05,
+      "loss": 1.3044,
+      "step": 5387
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935099956836628e-05,
+      "loss": 1.1477,
+      "step": 5388
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350697327417414e-05,
+      "loss": 1.2731,
+      "step": 5389
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350395018469238e-05,
+      "loss": 1.0054,
+      "step": 5390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935009264152395e-05,
+      "loss": 0.979,
+      "step": 5391
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934979019658375e-05,
+      "loss": 1.1997,
+      "step": 5392
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349487683650845e-05,
+      "loss": 1.2578,
+      "step": 5393
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934918510272742e-05,
+      "loss": 1.166,
+      "step": 5394
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348882453815684e-05,
+      "loss": 1.1342,
+      "step": 5395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934857973691784e-05,
+      "loss": 1.197,
+      "step": 5396
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934827695203608e-05,
+      "loss": 1.0615,
+      "step": 5397
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347974099172617e-05,
+      "loss": 1.0657,
+      "step": 5398
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347671178329646e-05,
+      "loss": 1.1822,
+      "step": 5399
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934736818950937e-05,
+      "loss": 1.2644,
+      "step": 5400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9347065132714e-05,
+      "loss": 1.1446,
+      "step": 5401
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346762007945733e-05,
+      "loss": 1.1342,
+      "step": 5402
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346458815206772e-05,
+      "loss": 1.2539,
+      "step": 5403
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346155554499322e-05,
+      "loss": 1.041,
+      "step": 5404
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93458522258256e-05,
+      "loss": 0.9849,
+      "step": 5405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345548829187794e-05,
+      "loss": 0.8878,
+      "step": 5406
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345245364588123e-05,
+      "loss": 1.2092,
+      "step": 5407
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344941832028793e-05,
+      "loss": 1.1209,
+      "step": 5408
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344638231512002e-05,
+      "loss": 0.923,
+      "step": 5409
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344334563039968e-05,
+      "loss": 0.976,
+      "step": 5410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9344030826614895e-05,
+      "loss": 0.9086,
+      "step": 5411
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934372702223899e-05,
+      "loss": 0.8667,
+      "step": 5412
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343423149914465e-05,
+      "loss": 0.9889,
+      "step": 5413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934311920964353e-05,
+      "loss": 0.9967,
+      "step": 5414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342815201428394e-05,
+      "loss": 1.2336,
+      "step": 5415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934251112527127e-05,
+      "loss": 1.1917,
+      "step": 5416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342206981174366e-05,
+      "loss": 1.2486,
+      "step": 5417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341902769139896e-05,
+      "loss": 1.2197,
+      "step": 5418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341598489170074e-05,
+      "loss": 1.1405,
+      "step": 5419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934129414126711e-05,
+      "loss": 0.6578,
+      "step": 5420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340989725433216e-05,
+      "loss": 1.3722,
+      "step": 5421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340685241670607e-05,
+      "loss": 1.4269,
+      "step": 5422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340380689981496e-05,
+      "loss": 1.3953,
+      "step": 5423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340076070368105e-05,
+      "loss": 1.4134,
+      "step": 5424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339771382832636e-05,
+      "loss": 1.0159,
+      "step": 5425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339466627377318e-05,
+      "loss": 0.9596,
+      "step": 5426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933916180400436e-05,
+      "loss": 0.9952,
+      "step": 5427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338856912715983e-05,
+      "loss": 1.3,
+      "step": 5428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93385519535144e-05,
+      "loss": 1.1015,
+      "step": 5429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9338246926401828e-05,
+      "loss": 1.09,
+      "step": 5430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337941831380485e-05,
+      "loss": 1.3136,
+      "step": 5431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337636668452596e-05,
+      "loss": 0.7208,
+      "step": 5432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337331437620375e-05,
+      "loss": 1.0408,
+      "step": 5433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337026138886043e-05,
+      "loss": 0.7588,
+      "step": 5434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336720772251822e-05,
+      "loss": 1.5527,
+      "step": 5435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933641533771993e-05,
+      "loss": 1.0593,
+      "step": 5436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933610983529259e-05,
+      "loss": 1.0236,
+      "step": 5437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335804264972018e-05,
+      "loss": 1.1587,
+      "step": 5438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335498626760446e-05,
+      "loss": 0.7847,
+      "step": 5439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933519292066009e-05,
+      "loss": 0.9233,
+      "step": 5440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334887146673172e-05,
+      "loss": 1.2102,
+      "step": 5441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334581304801917e-05,
+      "loss": 0.7919,
+      "step": 5442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933427539504855e-05,
+      "loss": 1.2628,
+      "step": 5443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333969417415302e-05,
+      "loss": 1.1841,
+      "step": 5444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333663371904388e-05,
+      "loss": 1.0435,
+      "step": 5445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333357258518036e-05,
+      "loss": 1.01,
+      "step": 5446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333051077258477e-05,
+      "loss": 0.6784,
+      "step": 5447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933274482812793e-05,
+      "loss": 1.0757,
+      "step": 5448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933243851112863e-05,
+      "loss": 1.0798,
+      "step": 5449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93321321262628e-05,
+      "loss": 1.1618,
+      "step": 5450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331825673532666e-05,
+      "loss": 0.9525,
+      "step": 5451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933151915294046e-05,
+      "loss": 1.319,
+      "step": 5452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331212564488413e-05,
+      "loss": 1.1645,
+      "step": 5453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330905908178753e-05,
+      "loss": 0.7926,
+      "step": 5454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330599184013708e-05,
+      "loss": 0.9456,
+      "step": 5455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933029239199551e-05,
+      "loss": 0.8062,
+      "step": 5456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329985532126384e-05,
+      "loss": 1.0487,
+      "step": 5457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932967860440857e-05,
+      "loss": 0.9029,
+      "step": 5458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329371608844303e-05,
+      "loss": 1.274,
+      "step": 5459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329064545435803e-05,
+      "loss": 1.0101,
+      "step": 5460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328757414185314e-05,
+      "loss": 1.1123,
+      "step": 5461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932845021509506e-05,
+      "loss": 1.1246,
+      "step": 5462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328142948167286e-05,
+      "loss": 0.8023,
+      "step": 5463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932783561340422e-05,
+      "loss": 1.0726,
+      "step": 5464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327528210808094e-05,
+      "loss": 0.9693,
+      "step": 5465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327220740381147e-05,
+      "loss": 0.8609,
+      "step": 5466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326913202125617e-05,
+      "loss": 0.8835,
+      "step": 5467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932660559604374e-05,
+      "loss": 0.9887,
+      "step": 5468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326297922137746e-05,
+      "loss": 1.1944,
+      "step": 5469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325990180409883e-05,
+      "loss": 0.874,
+      "step": 5470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325682370862383e-05,
+      "loss": 1.2079,
+      "step": 5471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932537449349748e-05,
+      "loss": 1.0403,
+      "step": 5472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325066548317423e-05,
+      "loss": 0.9602,
+      "step": 5473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324758535324447e-05,
+      "loss": 1.3196,
+      "step": 5474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932445045452079e-05,
+      "loss": 1.0272,
+      "step": 5475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324142305908695e-05,
+      "loss": 1.2235,
+      "step": 5476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323834089490402e-05,
+      "loss": 0.9085,
+      "step": 5477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323525805268152e-05,
+      "loss": 0.8952,
+      "step": 5478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323217453244184e-05,
+      "loss": 1.3197,
+      "step": 5479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322909033420748e-05,
+      "loss": 1.065,
+      "step": 5480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322600545800076e-05,
+      "loss": 1.2086,
+      "step": 5481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322291990384422e-05,
+      "loss": 1.201,
+      "step": 5482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321983367176025e-05,
+      "loss": 0.9046,
+      "step": 5483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932167467617713e-05,
+      "loss": 1.7283,
+      "step": 5484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932136591738998e-05,
+      "loss": 1.063,
+      "step": 5485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321057090816822e-05,
+      "loss": 0.936,
+      "step": 5486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320748196459904e-05,
+      "loss": 1.0758,
+      "step": 5487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932043923432147e-05,
+      "loss": 1.0443,
+      "step": 5488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320130204403765e-05,
+      "loss": 1.0317,
+      "step": 5489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931982110670904e-05,
+      "loss": 1.1213,
+      "step": 5490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931951194123954e-05,
+      "loss": 0.9159,
+      "step": 5491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319202707997512e-05,
+      "loss": 1.2048,
+      "step": 5492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931889340698521e-05,
+      "loss": 1.5483,
+      "step": 5493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931858403820488e-05,
+      "loss": 1.1224,
+      "step": 5494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318274601658768e-05,
+      "loss": 1.0861,
+      "step": 5495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317965097349135e-05,
+      "loss": 1.2339,
+      "step": 5496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931765552527822e-05,
+      "loss": 0.8082,
+      "step": 5497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317345885448282e-05,
+      "loss": 0.9128,
+      "step": 5498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317036177861567e-05,
+      "loss": 1.1213,
+      "step": 5499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316726402520335e-05,
+      "loss": 1.2416,
+      "step": 5500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931641655942683e-05,
+      "loss": 1.0071,
+      "step": 5501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316106648583312e-05,
+      "loss": 0.9986,
+      "step": 5502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315796669992028e-05,
+      "loss": 1.1858,
+      "step": 5503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931548662365524e-05,
+      "loss": 1.2408,
+      "step": 5504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315176509575196e-05,
+      "loss": 0.8498,
+      "step": 5505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314866327754158e-05,
+      "loss": 1.1362,
+      "step": 5506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314556078194374e-05,
+      "loss": 0.9121,
+      "step": 5507
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9314245760898104e-05,
+      "loss": 0.7438,
+      "step": 5508
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313935375867608e-05,
+      "loss": 0.9081,
+      "step": 5509
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313624923105135e-05,
+      "loss": 0.9577,
+      "step": 5510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313314402612953e-05,
+      "loss": 0.9444,
+      "step": 5511
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931300381439331e-05,
+      "loss": 1.1363,
+      "step": 5512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312693158448466e-05,
+      "loss": 1.3001,
+      "step": 5513
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931238243478069e-05,
+      "loss": 1.0141,
+      "step": 5514
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312071643392235e-05,
+      "loss": 0.9756,
+      "step": 5515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311760784285357e-05,
+      "loss": 1.5966,
+      "step": 5516
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311449857462322e-05,
+      "loss": 1.1339,
+      "step": 5517
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311138862925385e-05,
+      "loss": 1.0617,
+      "step": 5518
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310827800676818e-05,
+      "loss": 1.1596,
+      "step": 5519
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310516670718877e-05,
+      "loss": 0.8159,
+      "step": 5520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310205473053824e-05,
+      "loss": 0.9416,
+      "step": 5521
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930989420768392e-05,
+      "loss": 1.2119,
+      "step": 5522
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309582874611435e-05,
+      "loss": 0.7779,
+      "step": 5523
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930927147383863e-05,
+      "loss": 1.4635,
+      "step": 5524
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308960005367768e-05,
+      "loss": 1.3565,
+      "step": 5525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308648469201112e-05,
+      "loss": 1.1084,
+      "step": 5526
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308336865340935e-05,
+      "loss": 0.9929,
+      "step": 5527
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308025193789497e-05,
+      "loss": 0.9289,
+      "step": 5528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307713454549068e-05,
+      "loss": 1.0434,
+      "step": 5529
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930740164762191e-05,
+      "loss": 0.7675,
+      "step": 5530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930708977301029e-05,
+      "loss": 1.0025,
+      "step": 5531
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306777830716487e-05,
+      "loss": 1.265,
+      "step": 5532
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930646582074276e-05,
+      "loss": 1.0105,
+      "step": 5533
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930615374309138e-05,
+      "loss": 1.018,
+      "step": 5534
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305841597764615e-05,
+      "loss": 1.0416,
+      "step": 5535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930552938476474e-05,
+      "loss": 1.3285,
+      "step": 5536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930521710409402e-05,
+      "loss": 1.4224,
+      "step": 5537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930490475575473e-05,
+      "loss": 1.1283,
+      "step": 5538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304592339749133e-05,
+      "loss": 1.202,
+      "step": 5539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9304279856079513e-05,
+      "loss": 0.9966,
+      "step": 5540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303967304748136e-05,
+      "loss": 0.9891,
+      "step": 5541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303654685757275e-05,
+      "loss": 1.0793,
+      "step": 5542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303341999109203e-05,
+      "loss": 1.0025,
+      "step": 5543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303029244806194e-05,
+      "loss": 1.1171,
+      "step": 5544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302716422850524e-05,
+      "loss": 0.9962,
+      "step": 5545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302403533244467e-05,
+      "loss": 1.1256,
+      "step": 5546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.93020905759903e-05,
+      "loss": 1.1781,
+      "step": 5547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301777551090297e-05,
+      "loss": 1.3535,
+      "step": 5548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301464458546734e-05,
+      "loss": 1.2865,
+      "step": 5549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301151298361887e-05,
+      "loss": 0.8282,
+      "step": 5550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300838070538038e-05,
+      "loss": 1.2699,
+      "step": 5551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300524775077463e-05,
+      "loss": 0.5835,
+      "step": 5552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300211411982435e-05,
+      "loss": 1.0217,
+      "step": 5553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299897981255238e-05,
+      "loss": 1.1971,
+      "step": 5554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929958448289815e-05,
+      "loss": 1.2157,
+      "step": 5555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929927091691345e-05,
+      "loss": 1.111,
+      "step": 5556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298957283303422e-05,
+      "loss": 0.8444,
+      "step": 5557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929864358207034e-05,
+      "loss": 1.074,
+      "step": 5558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298329813216496e-05,
+      "loss": 1.025,
+      "step": 5559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298015976744163e-05,
+      "loss": 1.2958,
+      "step": 5560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297702072655622e-05,
+      "loss": 1.3932,
+      "step": 5561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297388100953157e-05,
+      "loss": 1.2059,
+      "step": 5562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929707406163906e-05,
+      "loss": 0.9735,
+      "step": 5563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296759954715605e-05,
+      "loss": 1.1172,
+      "step": 5564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296445780185077e-05,
+      "loss": 0.9371,
+      "step": 5565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296131538049766e-05,
+      "loss": 1.2203,
+      "step": 5566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295817228311953e-05,
+      "loss": 0.8849,
+      "step": 5567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295502850973925e-05,
+      "loss": 1.0378,
+      "step": 5568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295188406037967e-05,
+      "loss": 0.7568,
+      "step": 5569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929487389350637e-05,
+      "loss": 1.0908,
+      "step": 5570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294559313381412e-05,
+      "loss": 1.322,
+      "step": 5571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294244665665388e-05,
+      "loss": 1.0367,
+      "step": 5572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929392995036059e-05,
+      "loss": 0.914,
+      "step": 5573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92936151674693e-05,
+      "loss": 1.1196,
+      "step": 5574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9293300316993804e-05,
+      "loss": 1.2448,
+      "step": 5575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92929853989364e-05,
+      "loss": 1.4046,
+      "step": 5576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292670413299373e-05,
+      "loss": 1.0999,
+      "step": 5577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929235536008501e-05,
+      "loss": 1.04,
+      "step": 5578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292040239295614e-05,
+      "loss": 0.9106,
+      "step": 5579
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929172505093347e-05,
+      "loss": 1.0134,
+      "step": 5580
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291409795000863e-05,
+      "loss": 0.8245,
+      "step": 5581
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291094471500097e-05,
+      "loss": 1.1442,
+      "step": 5582
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929077908043346e-05,
+      "loss": 0.8982,
+      "step": 5583
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290463621803244e-05,
+      "loss": 1.1014,
+      "step": 5584
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290148095611746e-05,
+      "loss": 0.9693,
+      "step": 5585
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928983250186126e-05,
+      "loss": 1.4319,
+      "step": 5586
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928951684055408e-05,
+      "loss": 0.7557,
+      "step": 5587
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92892011116925e-05,
+      "loss": 1.3419,
+      "step": 5588
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928888531527882e-05,
+      "loss": 0.9878,
+      "step": 5589
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288569451315335e-05,
+      "loss": 1.0723,
+      "step": 5590
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288253519804338e-05,
+      "loss": 0.9984,
+      "step": 5591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287937520748134e-05,
+      "loss": 0.8324,
+      "step": 5592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287621454149016e-05,
+      "loss": 1.0002,
+      "step": 5593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287305320009282e-05,
+      "loss": 1.2545,
+      "step": 5594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928698911833123e-05,
+      "loss": 1.0186,
+      "step": 5595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928667284911717e-05,
+      "loss": 0.9865,
+      "step": 5596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286356512369388e-05,
+      "loss": 1.033,
+      "step": 5597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928604010809019e-05,
+      "loss": 1.1113,
+      "step": 5598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285723636281877e-05,
+      "loss": 0.6794,
+      "step": 5599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285407096946752e-05,
+      "loss": 0.8518,
+      "step": 5600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285090490087116e-05,
+      "loss": 0.8172,
+      "step": 5601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928477381570527e-05,
+      "loss": 0.9807,
+      "step": 5602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284457073803516e-05,
+      "loss": 1.1291,
+      "step": 5603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928414026438416e-05,
+      "loss": 0.9072,
+      "step": 5604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928382338744951e-05,
+      "loss": 1.3833,
+      "step": 5605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928350644300186e-05,
+      "loss": 1.0304,
+      "step": 5606
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283189431043522e-05,
+      "loss": 0.7421,
+      "step": 5607
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282872351576798e-05,
+      "loss": 1.1651,
+      "step": 5608
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282555204603998e-05,
+      "loss": 1.1208,
+      "step": 5609
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282237990127425e-05,
+      "loss": 1.0165,
+      "step": 5610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281920708149387e-05,
+      "loss": 0.8726,
+      "step": 5611
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281603358672188e-05,
+      "loss": 1.0257,
+      "step": 5612
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928128594169814e-05,
+      "loss": 1.0252,
+      "step": 5613
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928096845722955e-05,
+      "loss": 1.1866,
+      "step": 5614
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928065090526873e-05,
+      "loss": 1.0506,
+      "step": 5615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928033328581798e-05,
+      "loss": 0.9887,
+      "step": 5616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928001559887962e-05,
+      "loss": 1.0638,
+      "step": 5617
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279697844455954e-05,
+      "loss": 1.1085,
+      "step": 5618
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92793800225493e-05,
+      "loss": 0.999,
+      "step": 5619
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279062133161957e-05,
+      "loss": 1.5144,
+      "step": 5620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278744176296246e-05,
+      "loss": 1.2707,
+      "step": 5621
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278426151954478e-05,
+      "loss": 0.8921,
+      "step": 5622
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278108060138966e-05,
+      "loss": 1.3064,
+      "step": 5623
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277789900852018e-05,
+      "loss": 1.0075,
+      "step": 5624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277471674095952e-05,
+      "loss": 0.7222,
+      "step": 5625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277153379873085e-05,
+      "loss": 1.2271,
+      "step": 5626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276835018185727e-05,
+      "loss": 1.0691,
+      "step": 5627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276516589036196e-05,
+      "loss": 1.0308,
+      "step": 5628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276198092426807e-05,
+      "loss": 1.1288,
+      "step": 5629
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927587952835987e-05,
+      "loss": 0.8998,
+      "step": 5630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927556089683771e-05,
+      "loss": 1.0305,
+      "step": 5631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927524219786264e-05,
+      "loss": 0.9838,
+      "step": 5632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927492343143698e-05,
+      "loss": 1.0014,
+      "step": 5633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927460459756305e-05,
+      "loss": 1.0096,
+      "step": 5634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927428569624316e-05,
+      "loss": 1.2216,
+      "step": 5635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273966727479636e-05,
+      "loss": 0.75,
+      "step": 5636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92736476912748e-05,
+      "loss": 1.5823,
+      "step": 5637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273328587630964e-05,
+      "loss": 1.3019,
+      "step": 5638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273009416550456e-05,
+      "loss": 1.0945,
+      "step": 5639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927269017803559e-05,
+      "loss": 1.0982,
+      "step": 5640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9272370872088693e-05,
+      "loss": 1.0892,
+      "step": 5641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927205149871208e-05,
+      "loss": 1.1292,
+      "step": 5642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271732057908086e-05,
+      "loss": 1.1264,
+      "step": 5643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927141254967903e-05,
+      "loss": 0.8416,
+      "step": 5644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271092974027222e-05,
+      "loss": 0.7007,
+      "step": 5645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270773330955005e-05,
+      "loss": 1.2602,
+      "step": 5646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927045362046469e-05,
+      "loss": 1.0466,
+      "step": 5647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270133842558608e-05,
+      "loss": 1.1086,
+      "step": 5648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269813997239084e-05,
+      "loss": 1.0336,
+      "step": 5649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269494084508446e-05,
+      "loss": 1.2042,
+      "step": 5650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269174104369012e-05,
+      "loss": 1.1282,
+      "step": 5651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926885405682312e-05,
+      "loss": 0.959,
+      "step": 5652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268533941873086e-05,
+      "loss": 1.223,
+      "step": 5653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926821375952125e-05,
+      "loss": 1.0254,
+      "step": 5654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267893509769927e-05,
+      "loss": 1.6335,
+      "step": 5655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926757319262146e-05,
+      "loss": 1.211,
+      "step": 5656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926725280807817e-05,
+      "loss": 1.0754,
+      "step": 5657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266932356142386e-05,
+      "loss": 0.9031,
+      "step": 5658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266611836816445e-05,
+      "loss": 0.9733,
+      "step": 5659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926629125010267e-05,
+      "loss": 1.0889,
+      "step": 5660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265970596003396e-05,
+      "loss": 1.2105,
+      "step": 5661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265649874520957e-05,
+      "loss": 1.1921,
+      "step": 5662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265329085657683e-05,
+      "loss": 1.0432,
+      "step": 5663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265008229415904e-05,
+      "loss": 1.3538,
+      "step": 5664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926468730579796e-05,
+      "loss": 0.9418,
+      "step": 5665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926436631480618e-05,
+      "loss": 0.8897,
+      "step": 5666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264045256442898e-05,
+      "loss": 1.1808,
+      "step": 5667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263724130710454e-05,
+      "loss": 1.0006,
+      "step": 5668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9263402937611176e-05,
+      "loss": 0.9468,
+      "step": 5669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92630816771474e-05,
+      "loss": 1.0209,
+      "step": 5670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926276034932147e-05,
+      "loss": 1.2141,
+      "step": 5671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9262438954135718e-05,
+      "loss": 1.0779,
+      "step": 5672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9262117491592484e-05,
+      "loss": 0.8966,
+      "step": 5673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92617959616941e-05,
+      "loss": 1.4026,
+      "step": 5674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261474364442906e-05,
+      "loss": 1.0377,
+      "step": 5675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261152699841243e-05,
+      "loss": 1.1954,
+      "step": 5676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926083096789145e-05,
+      "loss": 1.148,
+      "step": 5677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926050916859586e-05,
+      "loss": 1.0756,
+      "step": 5678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260187301956827e-05,
+      "loss": 0.9493,
+      "step": 5679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259865367976682e-05,
+      "loss": 0.6636,
+      "step": 5680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259543366657766e-05,
+      "loss": 1.0034,
+      "step": 5681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925922129800242e-05,
+      "loss": 0.8125,
+      "step": 5682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925889916201299e-05,
+      "loss": 1.0655,
+      "step": 5683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258576958691822e-05,
+      "loss": 0.9003,
+      "step": 5684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925825468804125e-05,
+      "loss": 1.5828,
+      "step": 5685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925793235006362e-05,
+      "loss": 1.1154,
+      "step": 5686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257609944761282e-05,
+      "loss": 1.0979,
+      "step": 5687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257287472136575e-05,
+      "loss": 1.1041,
+      "step": 5688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256964932191848e-05,
+      "loss": 1.3098,
+      "step": 5689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925664232492944e-05,
+      "loss": 1.1637,
+      "step": 5690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256319650351703e-05,
+      "loss": 0.9815,
+      "step": 5691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925599690846098e-05,
+      "loss": 1.3074,
+      "step": 5692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925567409925962e-05,
+      "loss": 1.0396,
+      "step": 5693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255351222749968e-05,
+      "loss": 1.1512,
+      "step": 5694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255028278934378e-05,
+      "loss": 0.8639,
+      "step": 5695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9254705267815193e-05,
+      "loss": 1.1352,
+      "step": 5696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925438218939477e-05,
+      "loss": 0.9484,
+      "step": 5697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925405904367544e-05,
+      "loss": 0.976,
+      "step": 5698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925373583065957e-05,
+      "loss": 1.2295,
+      "step": 5699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253412550349507e-05,
+      "loss": 1.3484,
+      "step": 5700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253089202747602e-05,
+      "loss": 1.1397,
+      "step": 5701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252765787856204e-05,
+      "loss": 1.0903,
+      "step": 5702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925244230567766e-05,
+      "loss": 1.0826,
+      "step": 5703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252118756214336e-05,
+      "loss": 1.0692,
+      "step": 5704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251795139468573e-05,
+      "loss": 1.3126,
+      "step": 5705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925147145544273e-05,
+      "loss": 0.8116,
+      "step": 5706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251147704139158e-05,
+      "loss": 0.9746,
+      "step": 5707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925082388556021e-05,
+      "loss": 1.1882,
+      "step": 5708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250499999708246e-05,
+      "loss": 1.1575,
+      "step": 5709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925017604658562e-05,
+      "loss": 1.2051,
+      "step": 5710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9249852026194686e-05,
+      "loss": 1.1298,
+      "step": 5711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92495279385378e-05,
+      "loss": 1.0605,
+      "step": 5712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924920378361732e-05,
+      "loss": 0.7778,
+      "step": 5713
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92488795614356e-05,
+      "loss": 1.0201,
+      "step": 5714
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248555271995006e-05,
+      "loss": 1.3783,
+      "step": 5715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924823091529789e-05,
+      "loss": 0.8987,
+      "step": 5716
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924790649134661e-05,
+      "loss": 0.8693,
+      "step": 5717
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247582000143524e-05,
+      "loss": 1.1541,
+      "step": 5718
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247257441691002e-05,
+      "loss": 0.8351,
+      "step": 5719
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9246932815991393e-05,
+      "loss": 0.9026,
+      "step": 5720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924660812304706e-05,
+      "loss": 1.1885,
+      "step": 5721
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924628336286037e-05,
+      "loss": 1.0678,
+      "step": 5722
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245958535433677e-05,
+      "loss": 0.9393,
+      "step": 5723
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924563364076935e-05,
+      "loss": 1.1675,
+      "step": 5724
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245308678869747e-05,
+      "loss": 1.0802,
+      "step": 5725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244983649737232e-05,
+      "loss": 1.1812,
+      "step": 5726
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924465855337417e-05,
+      "loss": 1.0425,
+      "step": 5727
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244333389782924e-05,
+      "loss": 1.2024,
+      "step": 5728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244008158965857e-05,
+      "loss": 1.0263,
+      "step": 5729
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924368286092534e-05,
+      "loss": 0.9574,
+      "step": 5730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9243357495663733e-05,
+      "loss": 0.9114,
+      "step": 5731
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9243032063183406e-05,
+      "loss": 0.9258,
+      "step": 5732
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924270656348672e-05,
+      "loss": 1.0894,
+      "step": 5733
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242380996576044e-05,
+      "loss": 0.7743,
+      "step": 5734
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924205536245375e-05,
+      "loss": 1.0005,
+      "step": 5735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241729661122204e-05,
+      "loss": 1.1803,
+      "step": 5736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241403892583774e-05,
+      "loss": 0.9902,
+      "step": 5737
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241078056840825e-05,
+      "loss": 1.1436,
+      "step": 5738
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924075215389573e-05,
+      "loss": 1.1854,
+      "step": 5739
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240426183750863e-05,
+      "loss": 0.9589,
+      "step": 5740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924010014640859e-05,
+      "loss": 1.1838,
+      "step": 5741
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239774041871282e-05,
+      "loss": 0.8652,
+      "step": 5742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923944787014131e-05,
+      "loss": 1.0243,
+      "step": 5743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923912163122105e-05,
+      "loss": 1.0867,
+      "step": 5744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238795325112867e-05,
+      "loss": 1.2866,
+      "step": 5745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238468951819144e-05,
+      "loss": 0.7345,
+      "step": 5746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238142511342243e-05,
+      "loss": 1.3183,
+      "step": 5747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237816003684547e-05,
+      "loss": 0.9694,
+      "step": 5748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923748942884843e-05,
+      "loss": 1.155,
+      "step": 5749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923716278683626e-05,
+      "loss": 1.1932,
+      "step": 5750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236836077650414e-05,
+      "loss": 1.0008,
+      "step": 5751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236509301293278e-05,
+      "loss": 1.0882,
+      "step": 5752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236182457767218e-05,
+      "loss": 1.1214,
+      "step": 5753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923585554707461e-05,
+      "loss": 0.9357,
+      "step": 5754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235528569217835e-05,
+      "loss": 1.1216,
+      "step": 5755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235201524199273e-05,
+      "loss": 0.8025,
+      "step": 5756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234874412021298e-05,
+      "loss": 1.1369,
+      "step": 5757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234547232686292e-05,
+      "loss": 1.0181,
+      "step": 5758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234219986196633e-05,
+      "loss": 0.8278,
+      "step": 5759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92338926725547e-05,
+      "loss": 1.3015,
+      "step": 5760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923356529176288e-05,
+      "loss": 1.0539,
+      "step": 5761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923323784382354e-05,
+      "loss": 0.7648,
+      "step": 5762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923291032873907e-05,
+      "loss": 1.2428,
+      "step": 5763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232582746511858e-05,
+      "loss": 1.1053,
+      "step": 5764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232255097144273e-05,
+      "loss": 1.1345,
+      "step": 5765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231927380638705e-05,
+      "loss": 1.1068,
+      "step": 5766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231599596997533e-05,
+      "loss": 1.0744,
+      "step": 5767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231271746223146e-05,
+      "loss": 1.3107,
+      "step": 5768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230943828317924e-05,
+      "loss": 0.8751,
+      "step": 5769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923061584328426e-05,
+      "loss": 1.2575,
+      "step": 5770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230287791124523e-05,
+      "loss": 0.8798,
+      "step": 5771
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229959671841114e-05,
+      "loss": 0.9452,
+      "step": 5772
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229631485436413e-05,
+      "loss": 1.1947,
+      "step": 5773
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229303231912807e-05,
+      "loss": 1.1529,
+      "step": 5774
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9228974911272682e-05,
+      "loss": 1.1115,
+      "step": 5775
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9228646523518427e-05,
+      "loss": 0.7002,
+      "step": 5776
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922831806865243e-05,
+      "loss": 1.0626,
+      "step": 5777
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922798954667708e-05,
+      "loss": 0.623,
+      "step": 5778
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922766095759476e-05,
+      "loss": 0.6805,
+      "step": 5779
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922733230140787e-05,
+      "loss": 0.8945,
+      "step": 5780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9227003578118793e-05,
+      "loss": 1.4192,
+      "step": 5781
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226674787729923e-05,
+      "loss": 0.9667,
+      "step": 5782
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226345930243647e-05,
+      "loss": 0.9909,
+      "step": 5783
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922601700566236e-05,
+      "loss": 1.2436,
+      "step": 5784
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9225688013988453e-05,
+      "loss": 1.0212,
+      "step": 5785
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922535895522432e-05,
+      "loss": 1.0346,
+      "step": 5786
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922502982937235e-05,
+      "loss": 1.0268,
+      "step": 5787
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224700636434938e-05,
+      "loss": 0.7101,
+      "step": 5788
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922437137641448e-05,
+      "loss": 0.7991,
+      "step": 5789
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922404204931337e-05,
+      "loss": 1.1014,
+      "step": 5790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223712655134e-05,
+      "loss": 1.1907,
+      "step": 5791
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223383193878767e-05,
+      "loss": 1.0858,
+      "step": 5792
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223053665550067e-05,
+      "loss": 0.9774,
+      "step": 5793
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222724070150297e-05,
+      "loss": 1.1189,
+      "step": 5794
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222394407681854e-05,
+      "loss": 0.8558,
+      "step": 5795
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222064678147136e-05,
+      "loss": 1.1184,
+      "step": 5796
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922173488154854e-05,
+      "loss": 1.222,
+      "step": 5797
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922140501788846e-05,
+      "loss": 0.9541,
+      "step": 5798
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221075087169302e-05,
+      "loss": 0.971,
+      "step": 5799
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220745089393462e-05,
+      "loss": 0.9146,
+      "step": 5800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220415024563336e-05,
+      "loss": 1.0218,
+      "step": 5801
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220084892681334e-05,
+      "loss": 0.976,
+      "step": 5802
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219754693749846e-05,
+      "loss": 1.0959,
+      "step": 5803
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921942442777128e-05,
+      "loss": 1.0364,
+      "step": 5804
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219094094748035e-05,
+      "loss": 0.9764,
+      "step": 5805
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218763694682513e-05,
+      "loss": 1.0233,
+      "step": 5806
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921843322757712e-05,
+      "loss": 1.0502,
+      "step": 5807
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218102693434256e-05,
+      "loss": 0.8753,
+      "step": 5808
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217772092256326e-05,
+      "loss": 1.2056,
+      "step": 5809
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217441424045733e-05,
+      "loss": 1.1207,
+      "step": 5810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217110688804882e-05,
+      "loss": 1.1742,
+      "step": 5811
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921677988653618e-05,
+      "loss": 1.0535,
+      "step": 5812
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921644901724203e-05,
+      "loss": 1.3734,
+      "step": 5813
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921611808092484e-05,
+      "loss": 1.5492,
+      "step": 5814
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215787077587013e-05,
+      "loss": 1.306,
+      "step": 5815
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215456007230963e-05,
+      "loss": 1.0622,
+      "step": 5816
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215124869859093e-05,
+      "loss": 0.9844,
+      "step": 5817
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921479366547381e-05,
+      "loss": 1.2581,
+      "step": 5818
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214462394077525e-05,
+      "loss": 1.0681,
+      "step": 5819
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214131055672648e-05,
+      "loss": 1.0524,
+      "step": 5820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213799650261582e-05,
+      "loss": 1.2296,
+      "step": 5821
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213468177846745e-05,
+      "loss": 1.3503,
+      "step": 5822
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213136638430544e-05,
+      "loss": 0.7847,
+      "step": 5823
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212805032015393e-05,
+      "loss": 0.6742,
+      "step": 5824
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212473358603696e-05,
+      "loss": 1.3619,
+      "step": 5825
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921214161819787e-05,
+      "loss": 1.2673,
+      "step": 5826
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921180981080033e-05,
+      "loss": 0.9236,
+      "step": 5827
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211477936413484e-05,
+      "loss": 1.1301,
+      "step": 5828
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211145995039747e-05,
+      "loss": 1.2521,
+      "step": 5829
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9210813986681536e-05,
+      "loss": 0.839,
+      "step": 5830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9210481911341264e-05,
+      "loss": 0.9585,
+      "step": 5831
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921014976902134e-05,
+      "loss": 0.8757,
+      "step": 5832
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209817559724186e-05,
+      "loss": 0.7314,
+      "step": 5833
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209485283452215e-05,
+      "loss": 1.1753,
+      "step": 5834
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209152940207846e-05,
+      "loss": 1.0816,
+      "step": 5835
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92088205299935e-05,
+      "loss": 1.0805,
+      "step": 5836
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208488052811577e-05,
+      "loss": 1.0747,
+      "step": 5837
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208155508664515e-05,
+      "loss": 1.1226,
+      "step": 5838
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9207822897554718e-05,
+      "loss": 0.8868,
+      "step": 5839
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9207490219484612e-05,
+      "loss": 1.1321,
+      "step": 5840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9207157474456615e-05,
+      "loss": 1.0801,
+      "step": 5841
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206824662473148e-05,
+      "loss": 1.1287,
+      "step": 5842
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920649178353663e-05,
+      "loss": 0.6535,
+      "step": 5843
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920615883764948e-05,
+      "loss": 1.1165,
+      "step": 5844
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920582582481412e-05,
+      "loss": 1.0999,
+      "step": 5845
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205492745032975e-05,
+      "loss": 1.0665,
+      "step": 5846
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920515959830846e-05,
+      "loss": 0.9938,
+      "step": 5847
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204826384643005e-05,
+      "loss": 1.1453,
+      "step": 5848
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920449310403903e-05,
+      "loss": 0.8146,
+      "step": 5849
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920415975649896e-05,
+      "loss": 1.3513,
+      "step": 5850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203826342025214e-05,
+      "loss": 0.87,
+      "step": 5851
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203492860620225e-05,
+      "loss": 1.0373,
+      "step": 5852
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203159312286415e-05,
+      "loss": 0.9847,
+      "step": 5853
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202825697026208e-05,
+      "loss": 0.6987,
+      "step": 5854
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202492014842026e-05,
+      "loss": 1.0957,
+      "step": 5855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9202158265736305e-05,
+      "loss": 1.0749,
+      "step": 5856
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201824449711464e-05,
+      "loss": 1.2016,
+      "step": 5857
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201490566769933e-05,
+      "loss": 0.9753,
+      "step": 5858
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201156616914144e-05,
+      "loss": 0.9223,
+      "step": 5859
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920082260014652e-05,
+      "loss": 1.1754,
+      "step": 5860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200488516469494e-05,
+      "loss": 0.9343,
+      "step": 5861
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200154365885494e-05,
+      "loss": 1.1268,
+      "step": 5862
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199820148396947e-05,
+      "loss": 1.0498,
+      "step": 5863
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919948586400629e-05,
+      "loss": 1.0763,
+      "step": 5864
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199151512715948e-05,
+      "loss": 1.1306,
+      "step": 5865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9198817094528355e-05,
+      "loss": 1.0466,
+      "step": 5866
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919848260944594e-05,
+      "loss": 0.874,
+      "step": 5867
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9198148057471144e-05,
+      "loss": 1.1314,
+      "step": 5868
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197813438606393e-05,
+      "loss": 0.9032,
+      "step": 5869
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197478752854115e-05,
+      "loss": 1.4597,
+      "step": 5870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197144000216756e-05,
+      "loss": 1.1473,
+      "step": 5871
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196809180696743e-05,
+      "loss": 1.1153,
+      "step": 5872
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919647429429651e-05,
+      "loss": 1.0946,
+      "step": 5873
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196139341018498e-05,
+      "loss": 0.8633,
+      "step": 5874
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195804320865138e-05,
+      "loss": 1.0008,
+      "step": 5875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919546923383887e-05,
+      "loss": 1.4289,
+      "step": 5876
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195134079942128e-05,
+      "loss": 1.1431,
+      "step": 5877
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194798859177342e-05,
+      "loss": 1.1915,
+      "step": 5878
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919446357154697e-05,
+      "loss": 1.0196,
+      "step": 5879
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919412821705343e-05,
+      "loss": 1.151,
+      "step": 5880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919379279569917e-05,
+      "loss": 1.0123,
+      "step": 5881
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919345730748663e-05,
+      "loss": 0.9997,
+      "step": 5882
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9193121752418244e-05,
+      "loss": 0.6006,
+      "step": 5883
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192786130496456e-05,
+      "loss": 1.0546,
+      "step": 5884
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919245044172371e-05,
+      "loss": 1.0956,
+      "step": 5885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919211468610244e-05,
+      "loss": 0.9577,
+      "step": 5886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191778863635092e-05,
+      "loss": 1.3082,
+      "step": 5887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191442974324112e-05,
+      "loss": 1.1236,
+      "step": 5888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191107018171932e-05,
+      "loss": 0.7742,
+      "step": 5889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190770995181006e-05,
+      "loss": 1.0484,
+      "step": 5890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190434905353766e-05,
+      "loss": 1.248,
+      "step": 5891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.919009874869267e-05,
+      "loss": 1.2113,
+      "step": 5892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918976252520015e-05,
+      "loss": 0.9743,
+      "step": 5893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189426234878663e-05,
+      "loss": 1.1752,
+      "step": 5894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918908987773064e-05,
+      "loss": 1.0669,
+      "step": 5895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188753453758538e-05,
+      "loss": 1.0258,
+      "step": 5896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188416962964803e-05,
+      "loss": 0.9804,
+      "step": 5897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188080405351878e-05,
+      "loss": 0.9812,
+      "step": 5898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187743780922212e-05,
+      "loss": 0.9928,
+      "step": 5899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187407089678252e-05,
+      "loss": 1.0524,
+      "step": 5900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187070331622448e-05,
+      "loss": 0.8131,
+      "step": 5901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186733506757248e-05,
+      "loss": 1.0923,
+      "step": 5902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186396615085103e-05,
+      "loss": 1.0812,
+      "step": 5903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918605965660846e-05,
+      "loss": 0.9425,
+      "step": 5904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918572263132977e-05,
+      "loss": 1.1113,
+      "step": 5905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185385539251488e-05,
+      "loss": 0.9718,
+      "step": 5906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185048380376064e-05,
+      "loss": 0.745,
+      "step": 5907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9184711154705943e-05,
+      "loss": 1.1887,
+      "step": 5908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9184373862243587e-05,
+      "loss": 0.881,
+      "step": 5909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918403650299144e-05,
+      "loss": 0.8744,
+      "step": 5910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183699076951962e-05,
+      "loss": 1.1938,
+      "step": 5911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183361584127605e-05,
+      "loss": 0.9744,
+      "step": 5912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183024024520823e-05,
+      "loss": 1.2906,
+      "step": 5913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918268639813407e-05,
+      "loss": 0.9366,
+      "step": 5914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182348704969803e-05,
+      "loss": 1.0286,
+      "step": 5915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182010945030476e-05,
+      "loss": 1.1171,
+      "step": 5916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181673118318542e-05,
+      "loss": 0.9673,
+      "step": 5917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181335224836467e-05,
+      "loss": 1.2774,
+      "step": 5918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91809972645867e-05,
+      "loss": 0.9993,
+      "step": 5919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91806592375717e-05,
+      "loss": 1.068,
+      "step": 5920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918032114379393e-05,
+      "loss": 0.9743,
+      "step": 5921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179982983255842e-05,
+      "loss": 1.1736,
+      "step": 5922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91796447559599e-05,
+      "loss": 1.2833,
+      "step": 5923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917930646190856e-05,
+      "loss": 1.1427,
+      "step": 5924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178968101104285e-05,
+      "loss": 0.9235,
+      "step": 5925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178629673549536e-05,
+      "loss": 1.5605,
+      "step": 5926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178291179246772e-05,
+      "loss": 0.8774,
+      "step": 5927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177952618198455e-05,
+      "loss": 0.8781,
+      "step": 5928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917761399040705e-05,
+      "loss": 1.0508,
+      "step": 5929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177275295875014e-05,
+      "loss": 1.5899,
+      "step": 5930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9176936534604812e-05,
+      "loss": 0.9939,
+      "step": 5931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917659770659891e-05,
+      "loss": 1.0876,
+      "step": 5932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9176258811859772e-05,
+      "loss": 1.2258,
+      "step": 5933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917591985038986e-05,
+      "loss": 1.0289,
+      "step": 5934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175580822191638e-05,
+      "loss": 0.8446,
+      "step": 5935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175241727267572e-05,
+      "loss": 1.1172,
+      "step": 5936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174902565620136e-05,
+      "loss": 1.0037,
+      "step": 5937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174563337251785e-05,
+      "loss": 1.218,
+      "step": 5938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917422404216499e-05,
+      "loss": 1.4955,
+      "step": 5939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917388468036222e-05,
+      "loss": 1.3742,
+      "step": 5940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173545251845943e-05,
+      "loss": 1.1938,
+      "step": 5941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173205756618626e-05,
+      "loss": 0.9235,
+      "step": 5942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172866194682735e-05,
+      "loss": 0.8312,
+      "step": 5943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172526566040746e-05,
+      "loss": 1.0143,
+      "step": 5944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172186870695122e-05,
+      "loss": 1.0639,
+      "step": 5945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917184710864834e-05,
+      "loss": 0.8821,
+      "step": 5946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171507279902864e-05,
+      "loss": 0.8114,
+      "step": 5947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171167384461174e-05,
+      "loss": 1.1214,
+      "step": 5948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917082742232573e-05,
+      "loss": 0.9062,
+      "step": 5949
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917048739349901e-05,
+      "loss": 0.7331,
+      "step": 5950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9170147297983494e-05,
+      "loss": 1.3503,
+      "step": 5951
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169807135781645e-05,
+      "loss": 1.1656,
+      "step": 5952
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916946690689594e-05,
+      "loss": 0.8411,
+      "step": 5953
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169126611328856e-05,
+      "loss": 1.1361,
+      "step": 5954
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168786249082862e-05,
+      "loss": 0.8576,
+      "step": 5955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168445820160438e-05,
+      "loss": 1.146,
+      "step": 5956
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168105324564056e-05,
+      "loss": 1.047,
+      "step": 5957
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167764762296196e-05,
+      "loss": 0.8562,
+      "step": 5958
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167424133359328e-05,
+      "loss": 0.8977,
+      "step": 5959
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916708343775594e-05,
+      "loss": 1.0269,
+      "step": 5960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91667426754885e-05,
+      "loss": 1.1767,
+      "step": 5961
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916640184655949e-05,
+      "loss": 0.8766,
+      "step": 5962
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166060950971388e-05,
+      "loss": 1.4955,
+      "step": 5963
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165719988726672e-05,
+      "loss": 0.7766,
+      "step": 5964
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165378959827823e-05,
+      "loss": 0.7632,
+      "step": 5965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916503786427732e-05,
+      "loss": 0.9041,
+      "step": 5966
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164696702077646e-05,
+      "loss": 2.1784,
+      "step": 5967
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164355473231276e-05,
+      "loss": 0.9217,
+      "step": 5968
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91640141777407e-05,
+      "loss": 1.2776,
+      "step": 5969
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163672815608392e-05,
+      "loss": 1.1631,
+      "step": 5970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163331386836835e-05,
+      "loss": 0.9603,
+      "step": 5971
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916298989142852e-05,
+      "loss": 1.1479,
+      "step": 5972
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162648329385926e-05,
+      "loss": 1.0687,
+      "step": 5973
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916230670071153e-05,
+      "loss": 1.0278,
+      "step": 5974
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161965005407826e-05,
+      "loss": 0.9649,
+      "step": 5975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161623243477296e-05,
+      "loss": 1.1099,
+      "step": 5976
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916128141492242e-05,
+      "loss": 0.8761,
+      "step": 5977
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160939519745693e-05,
+      "loss": 1.1999,
+      "step": 5978
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160597557949594e-05,
+      "loss": 1.0305,
+      "step": 5979
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916025552953661e-05,
+      "loss": 0.6544,
+      "step": 5980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159913434509235e-05,
+      "loss": 1.1689,
+      "step": 5981
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915957127286995e-05,
+      "loss": 0.8337,
+      "step": 5982
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159229044621246e-05,
+      "loss": 0.9907,
+      "step": 5983
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915888674976561e-05,
+      "loss": 0.8108,
+      "step": 5984
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158544388305534e-05,
+      "loss": 1.1564,
+      "step": 5985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158201960243507e-05,
+      "loss": 1.2395,
+      "step": 5986
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915785946558202e-05,
+      "loss": 1.0652,
+      "step": 5987
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157516904323557e-05,
+      "loss": 0.9524,
+      "step": 5988
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157174276470614e-05,
+      "loss": 0.8318,
+      "step": 5989
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156831582025687e-05,
+      "loss": 1.1802,
+      "step": 5990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156488820991264e-05,
+      "loss": 1.2322,
+      "step": 5991
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156145993369833e-05,
+      "loss": 1.4229,
+      "step": 5992
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155803099163894e-05,
+      "loss": 0.9844,
+      "step": 5993
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155460138375938e-05,
+      "loss": 0.9194,
+      "step": 5994
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915511711100846e-05,
+      "loss": 1.2631,
+      "step": 5995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154774017063954e-05,
+      "loss": 0.9632,
+      "step": 5996
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154430856544913e-05,
+      "loss": 1.2701,
+      "step": 5997
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154087629453837e-05,
+      "loss": 1.1931,
+      "step": 5998
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153744335793218e-05,
+      "loss": 1.0288,
+      "step": 5999
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915340097556555e-05,
+      "loss": 0.9048,
+      "step": 6000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153057548773338e-05,
+      "loss": 1.4246,
+      "step": 6001
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152714055419074e-05,
+      "loss": 0.9688,
+      "step": 6002
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152370495505256e-05,
+      "loss": 0.847,
+      "step": 6003
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152026869034384e-05,
+      "loss": 0.8705,
+      "step": 6004
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915168317600896e-05,
+      "loss": 1.2323,
+      "step": 6005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151339416431477e-05,
+      "loss": 0.9752,
+      "step": 6006
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150995590304437e-05,
+      "loss": 0.7991,
+      "step": 6007
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150651697630337e-05,
+      "loss": 0.8495,
+      "step": 6008
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150307738411686e-05,
+      "loss": 1.0129,
+      "step": 6009
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914996371265098e-05,
+      "loss": 0.9972,
+      "step": 6010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149619620350724e-05,
+      "loss": 0.8872,
+      "step": 6011
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149275461513414e-05,
+      "loss": 1.0711,
+      "step": 6012
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914893123614156e-05,
+      "loss": 0.9716,
+      "step": 6013
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148586944237662e-05,
+      "loss": 0.7744,
+      "step": 6014
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914824258580422e-05,
+      "loss": 0.9839,
+      "step": 6015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914789816084375e-05,
+      "loss": 1.0061,
+      "step": 6016
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914755366935874e-05,
+      "loss": 0.7978,
+      "step": 6017
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914720911135171e-05,
+      "loss": 1.0251,
+      "step": 6018
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146864486825156e-05,
+      "loss": 1.0193,
+      "step": 6019
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146519795781588e-05,
+      "loss": 0.9852,
+      "step": 6020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146175038223515e-05,
+      "loss": 0.8928,
+      "step": 6021
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9145830214153437e-05,
+      "loss": 0.8355,
+      "step": 6022
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914548532357387e-05,
+      "loss": 1.0568,
+      "step": 6023
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9145140366487318e-05,
+      "loss": 1.2378,
+      "step": 6024
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144795342896287e-05,
+      "loss": 1.2621,
+      "step": 6025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914445025280329e-05,
+      "loss": 1.1245,
+      "step": 6026
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914410509621084e-05,
+      "loss": 1.4329,
+      "step": 6027
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143759873121436e-05,
+      "loss": 1.2386,
+      "step": 6028
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143414583537598e-05,
+      "loss": 1.1939,
+      "step": 6029
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914306922746183e-05,
+      "loss": 1.1779,
+      "step": 6030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914272380489665e-05,
+      "loss": 1.025,
+      "step": 6031
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142378315844567e-05,
+      "loss": 1.0588,
+      "step": 6032
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142032760308095e-05,
+      "loss": 1.08,
+      "step": 6033
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141687138289744e-05,
+      "loss": 0.8656,
+      "step": 6034
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141341449792027e-05,
+      "loss": 1.031,
+      "step": 6035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140995694817466e-05,
+      "loss": 0.9216,
+      "step": 6036
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140649873368562e-05,
+      "loss": 1.3644,
+      "step": 6037
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140303985447843e-05,
+      "loss": 1.1395,
+      "step": 6038
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139958031057816e-05,
+      "loss": 0.7474,
+      "step": 6039
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139612010201e-05,
+      "loss": 1.722,
+      "step": 6040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913926592287991e-05,
+      "loss": 1.1849,
+      "step": 6041
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138919769097062e-05,
+      "loss": 1.2105,
+      "step": 6042
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138573548854974e-05,
+      "loss": 0.8191,
+      "step": 6043
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138227262156165e-05,
+      "loss": 1.3849,
+      "step": 6044
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137880909003155e-05,
+      "loss": 0.8679,
+      "step": 6045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137534489398456e-05,
+      "loss": 0.774,
+      "step": 6046
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137188003344595e-05,
+      "loss": 1.1117,
+      "step": 6047
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913684145084409e-05,
+      "loss": 1.0741,
+      "step": 6048
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136494831899456e-05,
+      "loss": 0.9433,
+      "step": 6049
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136148146513216e-05,
+      "loss": 1.2472,
+      "step": 6050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135801394687897e-05,
+      "loss": 1.24,
+      "step": 6051
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913545457642601e-05,
+      "loss": 1.0215,
+      "step": 6052
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135107691730085e-05,
+      "loss": 0.9903,
+      "step": 6053
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134760740602643e-05,
+      "loss": 1.2146,
+      "step": 6054
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913441372304621e-05,
+      "loss": 1.4138,
+      "step": 6055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91340666390633e-05,
+      "loss": 1.1509,
+      "step": 6056
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9133719488656446e-05,
+      "loss": 1.0432,
+      "step": 6057
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913337227182817e-05,
+      "loss": 1.0481,
+      "step": 6058
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9133024988580993e-05,
+      "loss": 1.0481,
+      "step": 6059
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913267763891745e-05,
+      "loss": 1.0574,
+      "step": 6060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132330222840055e-05,
+      "loss": 1.2807,
+      "step": 6061
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131982740351346e-05,
+      "loss": 0.9889,
+      "step": 6062
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913163519145384e-05,
+      "loss": 0.7921,
+      "step": 6063
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913128757615007e-05,
+      "loss": 0.9397,
+      "step": 6064
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130939894442564e-05,
+      "loss": 1.0307,
+      "step": 6065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913059214633385e-05,
+      "loss": 0.8429,
+      "step": 6066
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130244331826454e-05,
+      "loss": 1.0181,
+      "step": 6067
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129896450922906e-05,
+      "loss": 0.9889,
+      "step": 6068
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129548503625738e-05,
+      "loss": 0.9597,
+      "step": 6069
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129200489937483e-05,
+      "loss": 0.8752,
+      "step": 6070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128852409860664e-05,
+      "loss": 0.9036,
+      "step": 6071
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912850426339782e-05,
+      "loss": 1.0683,
+      "step": 6072
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128156050551476e-05,
+      "loss": 1.0045,
+      "step": 6073
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912780777132417e-05,
+      "loss": 1.3819,
+      "step": 6074
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912745942571843e-05,
+      "loss": 0.984,
+      "step": 6075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127111013736793e-05,
+      "loss": 1.1933,
+      "step": 6076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126762535381795e-05,
+      "loss": 1.189,
+      "step": 6077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912641399065596e-05,
+      "loss": 1.3514,
+      "step": 6078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912606537956183e-05,
+      "loss": 1.0818,
+      "step": 6079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125716702101943e-05,
+      "loss": 1.0389,
+      "step": 6080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125367958278827e-05,
+      "loss": 1.1149,
+      "step": 6081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125019148095023e-05,
+      "loss": 1.2187,
+      "step": 6082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124670271553065e-05,
+      "loss": 0.8443,
+      "step": 6083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124321328655495e-05,
+      "loss": 1.3595,
+      "step": 6084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123972319404847e-05,
+      "loss": 0.8275,
+      "step": 6085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123623243803655e-05,
+      "loss": 1.0829,
+      "step": 6086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123274101854464e-05,
+      "loss": 1.052,
+      "step": 6087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912292489355981e-05,
+      "loss": 0.893,
+      "step": 6088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9122575618922235e-05,
+      "loss": 0.9596,
+      "step": 6089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9122226277944276e-05,
+      "loss": 1.0612,
+      "step": 6090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912187687062847e-05,
+      "loss": 0.8841,
+      "step": 6091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912152739697737e-05,
+      "loss": 1.0501,
+      "step": 6092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121177856993505e-05,
+      "loss": 1.0815,
+      "step": 6093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120828250679423e-05,
+      "loss": 1.1324,
+      "step": 6094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120478578037666e-05,
+      "loss": 1.0127,
+      "step": 6095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120128839070777e-05,
+      "loss": 1.1082,
+      "step": 6096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119779033781296e-05,
+      "loss": 0.99,
+      "step": 6097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119429162171767e-05,
+      "loss": 1.1056,
+      "step": 6098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911907922424474e-05,
+      "loss": 0.8859,
+      "step": 6099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9118729220002757e-05,
+      "loss": 1.1833,
+      "step": 6100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911837914944836e-05,
+      "loss": 1.1838,
+      "step": 6101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91180290125841e-05,
+      "loss": 0.842,
+      "step": 6102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117678809412516e-05,
+      "loss": 0.9515,
+      "step": 6103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117328539936162e-05,
+      "loss": 1.1005,
+      "step": 6104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116978204157583e-05,
+      "loss": 0.8432,
+      "step": 6105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116627802079323e-05,
+      "loss": 1.2941,
+      "step": 6106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9116277333703936e-05,
+      "loss": 0.8564,
+      "step": 6107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911592679903397e-05,
+      "loss": 1.2505,
+      "step": 6108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911557619807197e-05,
+      "loss": 1.0664,
+      "step": 6109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115225530820486e-05,
+      "loss": 0.9992,
+      "step": 6110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114874797282072e-05,
+      "loss": 1.4896,
+      "step": 6111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114523997459274e-05,
+      "loss": 1.2808,
+      "step": 6112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114173131354644e-05,
+      "loss": 0.9026,
+      "step": 6113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911382219897074e-05,
+      "loss": 1.2585,
+      "step": 6114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113471200310105e-05,
+      "loss": 1.0063,
+      "step": 6115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113120135375296e-05,
+      "loss": 1.2841,
+      "step": 6116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911276900416887e-05,
+      "loss": 1.0702,
+      "step": 6117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911241780669337e-05,
+      "loss": 1.067,
+      "step": 6118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9112066542951358e-05,
+      "loss": 1.324,
+      "step": 6119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111715212945384e-05,
+      "loss": 1.017,
+      "step": 6120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111363816678007e-05,
+      "loss": 1.1191,
+      "step": 6121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911101235415178e-05,
+      "loss": 1.0071,
+      "step": 6122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110660825369264e-05,
+      "loss": 0.8508,
+      "step": 6123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110309230333006e-05,
+      "loss": 1.0647,
+      "step": 6124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109957569045568e-05,
+      "loss": 0.9501,
+      "step": 6125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109605841509508e-05,
+      "loss": 1.3857,
+      "step": 6126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910925404772738e-05,
+      "loss": 1.136,
+      "step": 6127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108902187701747e-05,
+      "loss": 1.1173,
+      "step": 6128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108550261435166e-05,
+      "loss": 0.8889,
+      "step": 6129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108198268930195e-05,
+      "loss": 1.183,
+      "step": 6130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9107846210189396e-05,
+      "loss": 0.9899,
+      "step": 6131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910749408521533e-05,
+      "loss": 1.0646,
+      "step": 6132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910714189401055e-05,
+      "loss": 0.8739,
+      "step": 6133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106789636577625e-05,
+      "loss": 0.8704,
+      "step": 6134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106437312919116e-05,
+      "loss": 0.9977,
+      "step": 6135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106084923037586e-05,
+      "loss": 1.0397,
+      "step": 6136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910573246693559e-05,
+      "loss": 0.9989,
+      "step": 6137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105379944615697e-05,
+      "loss": 1.2552,
+      "step": 6138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105027356080472e-05,
+      "loss": 1.0789,
+      "step": 6139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104674701332478e-05,
+      "loss": 0.9721,
+      "step": 6140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104321980374277e-05,
+      "loss": 1.0521,
+      "step": 6141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103969193208436e-05,
+      "loss": 0.8231,
+      "step": 6142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103616339837518e-05,
+      "loss": 1.1185,
+      "step": 6143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103263420264097e-05,
+      "loss": 1.3072,
+      "step": 6144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102910434490724e-05,
+      "loss": 0.9024,
+      "step": 6145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102557382519982e-05,
+      "loss": 0.8218,
+      "step": 6146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910220426435443e-05,
+      "loss": 0.989,
+      "step": 6147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101851079996638e-05,
+      "loss": 0.7113,
+      "step": 6148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101497829449175e-05,
+      "loss": 0.6967,
+      "step": 6149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910114451271461e-05,
+      "loss": 1.0392,
+      "step": 6150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910079112979551e-05,
+      "loss": 1.0833,
+      "step": 6151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100437680694445e-05,
+      "loss": 1.0642,
+      "step": 6152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100084165413988e-05,
+      "loss": 1.1784,
+      "step": 6153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909973058395671e-05,
+      "loss": 1.4185,
+      "step": 6154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909937693632518e-05,
+      "loss": 1.0268,
+      "step": 6155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909902322252197e-05,
+      "loss": 1.5019,
+      "step": 6156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098669442549652e-05,
+      "loss": 0.8719,
+      "step": 6157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90983155964108e-05,
+      "loss": 1.082,
+      "step": 6158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097961684107985e-05,
+      "loss": 1.2642,
+      "step": 6159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097607705643783e-05,
+      "loss": 1.0982,
+      "step": 6160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097253661020767e-05,
+      "loss": 1.0158,
+      "step": 6161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096899550241514e-05,
+      "loss": 0.9384,
+      "step": 6162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096545373308598e-05,
+      "loss": 1.3767,
+      "step": 6163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096191130224594e-05,
+      "loss": 0.7909,
+      "step": 6164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095836820992074e-05,
+      "loss": 1.1378,
+      "step": 6165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909548244561362e-05,
+      "loss": 1.1778,
+      "step": 6166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909512800409181e-05,
+      "loss": 1.1928,
+      "step": 6167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094773496429217e-05,
+      "loss": 0.9415,
+      "step": 6168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094418922628423e-05,
+      "loss": 1.2071,
+      "step": 6169
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094064282692004e-05,
+      "loss": 1.0573,
+      "step": 6170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909370957662254e-05,
+      "loss": 0.9761,
+      "step": 6171
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909335480442261e-05,
+      "loss": 1.0282,
+      "step": 6172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909299996609479e-05,
+      "loss": 1.0465,
+      "step": 6173
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909264506164167e-05,
+      "loss": 0.9411,
+      "step": 6174
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092290091065823e-05,
+      "loss": 0.9319,
+      "step": 6175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091935054369834e-05,
+      "loss": 1.1654,
+      "step": 6176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091579951556282e-05,
+      "loss": 1.1152,
+      "step": 6177
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091224782627754e-05,
+      "loss": 1.1842,
+      "step": 6178
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090869547586826e-05,
+      "loss": 0.9197,
+      "step": 6179
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090514246436085e-05,
+      "loss": 0.881,
+      "step": 6180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909015887917812e-05,
+      "loss": 0.7828,
+      "step": 6181
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089803445815507e-05,
+      "loss": 0.9231,
+      "step": 6182
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089447946350832e-05,
+      "loss": 0.9121,
+      "step": 6183
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089092380786685e-05,
+      "loss": 1.6257,
+      "step": 6184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908873674912565e-05,
+      "loss": 0.924,
+      "step": 6185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908838105137031e-05,
+      "loss": 1.0343,
+      "step": 6186
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9088025287523253e-05,
+      "loss": 1.05,
+      "step": 6187
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087669457587065e-05,
+      "loss": 1.0386,
+      "step": 6188
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908731356156434e-05,
+      "loss": 1.2276,
+      "step": 6189
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908695759945766e-05,
+      "loss": 1.3376,
+      "step": 6190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086601571269615e-05,
+      "loss": 1.2997,
+      "step": 6191
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086245477002793e-05,
+      "loss": 1.0792,
+      "step": 6192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085889316659787e-05,
+      "loss": 0.7678,
+      "step": 6193
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085533090243184e-05,
+      "loss": 0.9288,
+      "step": 6194
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9085176797755575e-05,
+      "loss": 1.0823,
+      "step": 6195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084820439199555e-05,
+      "loss": 1.2645,
+      "step": 6196
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908446401457771e-05,
+      "loss": 1.1463,
+      "step": 6197
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908410752389263e-05,
+      "loss": 1.0056,
+      "step": 6198
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083750967146913e-05,
+      "loss": 1.486,
+      "step": 6199
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083394344343153e-05,
+      "loss": 1.2606,
+      "step": 6200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908303765548394e-05,
+      "loss": 0.8201,
+      "step": 6201
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908268090057187e-05,
+      "loss": 1.141,
+      "step": 6202
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908232407960953e-05,
+      "loss": 1.1743,
+      "step": 6203
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081967192599525e-05,
+      "loss": 1.1232,
+      "step": 6204
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081610239544446e-05,
+      "loss": 1.1973,
+      "step": 6205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908125322044689e-05,
+      "loss": 1.2393,
+      "step": 6206
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908089613530945e-05,
+      "loss": 1.0167,
+      "step": 6207
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080538984134725e-05,
+      "loss": 1.2403,
+      "step": 6208
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080181766925313e-05,
+      "loss": 1.077,
+      "step": 6209
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079824483683808e-05,
+      "loss": 1.1859,
+      "step": 6210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907946713441281e-05,
+      "loss": 0.8108,
+      "step": 6211
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079109719114926e-05,
+      "loss": 0.9822,
+      "step": 6212
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9078752237792743e-05,
+      "loss": 0.6829,
+      "step": 6213
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907839469044887e-05,
+      "loss": 1.119,
+      "step": 6214
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9078037077085897e-05,
+      "loss": 1.0966,
+      "step": 6215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077679397706432e-05,
+      "loss": 0.9397,
+      "step": 6216
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9077321652313075e-05,
+      "loss": 1.011,
+      "step": 6217
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076963840908427e-05,
+      "loss": 0.9497,
+      "step": 6218
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907660596349509e-05,
+      "loss": 0.987,
+      "step": 6219
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076248020075663e-05,
+      "loss": 1.202,
+      "step": 6220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075890010652758e-05,
+      "loss": 0.8849,
+      "step": 6221
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907553193522897e-05,
+      "loss": 1.2446,
+      "step": 6222
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075173793806907e-05,
+      "loss": 0.9783,
+      "step": 6223
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074815586389173e-05,
+      "loss": 1.2518,
+      "step": 6224
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074457312978373e-05,
+      "loss": 1.1256,
+      "step": 6225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907409897357711e-05,
+      "loss": 0.8126,
+      "step": 6226
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907374056818799e-05,
+      "loss": 1.105,
+      "step": 6227
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073382096813623e-05,
+      "loss": 1.1541,
+      "step": 6228
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9073023559456613e-05,
+      "loss": 1.0648,
+      "step": 6229
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907266495611957e-05,
+      "loss": 1.1047,
+      "step": 6230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90723062868051e-05,
+      "loss": 1.138,
+      "step": 6231
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071947551515805e-05,
+      "loss": 0.9492,
+      "step": 6232
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907158875025431e-05,
+      "loss": 1.0375,
+      "step": 6233
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907122988302321e-05,
+      "loss": 0.9741,
+      "step": 6234
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070870949825117e-05,
+      "loss": 1.1207,
+      "step": 6235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907051195066264e-05,
+      "loss": 0.9247,
+      "step": 6236
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90701528855384e-05,
+      "loss": 0.9894,
+      "step": 6237
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069793754454996e-05,
+      "loss": 1.1701,
+      "step": 6238
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069434557415046e-05,
+      "loss": 0.8853,
+      "step": 6239
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069075294421163e-05,
+      "loss": 1.1287,
+      "step": 6240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068715965475957e-05,
+      "loss": 0.7693,
+      "step": 6241
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906835657058204e-05,
+      "loss": 1.0234,
+      "step": 6242
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9067997109742025e-05,
+      "loss": 1.0782,
+      "step": 6243
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906763758295853e-05,
+      "loss": 0.9223,
+      "step": 6244
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906727799023417e-05,
+      "loss": 0.8299,
+      "step": 6245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066918331571553e-05,
+      "loss": 1.1741,
+      "step": 6246
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066558606973303e-05,
+      "loss": 0.9132,
+      "step": 6247
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066198816442032e-05,
+      "loss": 1.1219,
+      "step": 6248
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065838959980353e-05,
+      "loss": 1.1122,
+      "step": 6249
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065479037590887e-05,
+      "loss": 1.1229,
+      "step": 6250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065119049276254e-05,
+      "loss": 1.3018,
+      "step": 6251
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064758995039066e-05,
+      "loss": 1.3672,
+      "step": 6252
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064398874881945e-05,
+      "loss": 1.1042,
+      "step": 6253
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064038688807508e-05,
+      "loss": 1.1159,
+      "step": 6254
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063678436818372e-05,
+      "loss": 0.8701,
+      "step": 6255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063318118917168e-05,
+      "loss": 0.7805,
+      "step": 6256
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90629577351065e-05,
+      "loss": 0.8089,
+      "step": 6257
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062597285389003e-05,
+      "loss": 1.9353,
+      "step": 6258
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906223676976729e-05,
+      "loss": 1.2358,
+      "step": 6259
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061876188243984e-05,
+      "loss": 1.1145,
+      "step": 6260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906151554082171e-05,
+      "loss": 1.1536,
+      "step": 6261
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061154827503085e-05,
+      "loss": 1.0763,
+      "step": 6262
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906079404829074e-05,
+      "loss": 1.0704,
+      "step": 6263
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060433203187288e-05,
+      "loss": 0.955,
+      "step": 6264
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060072292195366e-05,
+      "loss": 1.1111,
+      "step": 6265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9059711315317593e-05,
+      "loss": 1.2051,
+      "step": 6266
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905935027255659e-05,
+      "loss": 0.8825,
+      "step": 6267
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058989163914985e-05,
+      "loss": 1.2421,
+      "step": 6268
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905862798939541e-05,
+      "loss": 0.9868,
+      "step": 6269
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905826674900048e-05,
+      "loss": 1.4795,
+      "step": 6270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057905442732832e-05,
+      "loss": 1.0171,
+      "step": 6271
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905754407059509e-05,
+      "loss": 0.9257,
+      "step": 6272
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905718263258988e-05,
+      "loss": 1.0509,
+      "step": 6273
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905682112871983e-05,
+      "loss": 1.8928,
+      "step": 6274
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9056459558987573e-05,
+      "loss": 0.8317,
+      "step": 6275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9056097923395736e-05,
+      "loss": 1.1866,
+      "step": 6276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9055736221946952e-05,
+      "loss": 0.8484,
+      "step": 6277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9055374454643846e-05,
+      "loss": 1.3489,
+      "step": 6278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9055012621489055e-05,
+      "loss": 0.8814,
+      "step": 6279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90546507224852e-05,
+      "loss": 1.5045,
+      "step": 6280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9054288757634927e-05,
+      "loss": 1.0263,
+      "step": 6281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053926726940855e-05,
+      "loss": 0.956,
+      "step": 6282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053564630405624e-05,
+      "loss": 1.2151,
+      "step": 6283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053202468031865e-05,
+      "loss": 0.9854,
+      "step": 6284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052840239822218e-05,
+      "loss": 1.0104,
+      "step": 6285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9052477945779304e-05,
+      "loss": 0.8893,
+      "step": 6286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905211558590577e-05,
+      "loss": 1.0739,
+      "step": 6287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905175316020424e-05,
+      "loss": 1.0946,
+      "step": 6288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051390668677364e-05,
+      "loss": 1.0477,
+      "step": 6289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051028111327764e-05,
+      "loss": 1.0583,
+      "step": 6290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050665488158086e-05,
+      "loss": 0.9367,
+      "step": 6291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050302799170963e-05,
+      "loss": 1.2493,
+      "step": 6292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904994004436903e-05,
+      "loss": 0.9532,
+      "step": 6293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904957722375493e-05,
+      "loss": 1.1872,
+      "step": 6294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90492143373313e-05,
+      "loss": 1.2466,
+      "step": 6295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904885138510078e-05,
+      "loss": 1.3482,
+      "step": 6296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9048488367066006e-05,
+      "loss": 0.9237,
+      "step": 6297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9048125283229616e-05,
+      "loss": 1.2577,
+      "step": 6298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904776213359426e-05,
+      "loss": 0.9354,
+      "step": 6299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904739891816257e-05,
+      "loss": 1.5633,
+      "step": 6300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047035636937194e-05,
+      "loss": 1.1466,
+      "step": 6301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046672289920765e-05,
+      "loss": 0.8171,
+      "step": 6302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904630887711593e-05,
+      "loss": 1.1687,
+      "step": 6303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904594539852534e-05,
+      "loss": 1.0517,
+      "step": 6304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045581854151623e-05,
+      "loss": 1.0509,
+      "step": 6305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045218243997433e-05,
+      "loss": 1.4002,
+      "step": 6306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044854568065413e-05,
+      "loss": 0.942,
+      "step": 6307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044490826358204e-05,
+      "loss": 1.2223,
+      "step": 6308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044127018878457e-05,
+      "loss": 0.7691,
+      "step": 6309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043763145628807e-05,
+      "loss": 0.9673,
+      "step": 6310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904339920661191e-05,
+      "loss": 1.4357,
+      "step": 6311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043035201830414e-05,
+      "loss": 0.9033,
+      "step": 6312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904267113128696e-05,
+      "loss": 1.2459,
+      "step": 6313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042306994984194e-05,
+      "loss": 1.0809,
+      "step": 6314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041942792924767e-05,
+      "loss": 1.2548,
+      "step": 6315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041578525111334e-05,
+      "loss": 1.1741,
+      "step": 6316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041214191546535e-05,
+      "loss": 0.9256,
+      "step": 6317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904084979223302e-05,
+      "loss": 1.163,
+      "step": 6318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040485327173443e-05,
+      "loss": 1.2076,
+      "step": 6319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040120796370452e-05,
+      "loss": 1.5075,
+      "step": 6320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90397561998267e-05,
+      "loss": 0.9723,
+      "step": 6321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039391537544836e-05,
+      "loss": 1.1894,
+      "step": 6322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039026809527514e-05,
+      "loss": 0.9281,
+      "step": 6323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038662015777385e-05,
+      "loss": 1.0831,
+      "step": 6324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038297156297103e-05,
+      "loss": 1.2603,
+      "step": 6325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037932231089318e-05,
+      "loss": 1.2229,
+      "step": 6326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903756724015669e-05,
+      "loss": 1.0113,
+      "step": 6327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037202183501865e-05,
+      "loss": 0.9596,
+      "step": 6328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9036837061127503e-05,
+      "loss": 1.2493,
+      "step": 6329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903647187303626e-05,
+      "loss": 1.4055,
+      "step": 6330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903610661923079e-05,
+      "loss": 1.0562,
+      "step": 6331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903574129971375e-05,
+      "loss": 0.9821,
+      "step": 6332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035375914487794e-05,
+      "loss": 0.9116,
+      "step": 6333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035010463555583e-05,
+      "loss": 1.1394,
+      "step": 6334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034644946919768e-05,
+      "loss": 0.9649,
+      "step": 6335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034279364583015e-05,
+      "loss": 1.1487,
+      "step": 6336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903391371654798e-05,
+      "loss": 0.7628,
+      "step": 6337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033548002817318e-05,
+      "loss": 1.3194,
+      "step": 6338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033182223393698e-05,
+      "loss": 0.8876,
+      "step": 6339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032816378279767e-05,
+      "loss": 0.7445,
+      "step": 6340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032450467478196e-05,
+      "loss": 0.9526,
+      "step": 6341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032084490991643e-05,
+      "loss": 1.14,
+      "step": 6342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031718448822767e-05,
+      "loss": 1.0305,
+      "step": 6343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031352340974232e-05,
+      "loss": 0.8332,
+      "step": 6344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030986167448696e-05,
+      "loss": 0.9205,
+      "step": 6345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903061992824883e-05,
+      "loss": 0.9031,
+      "step": 6346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030253623377294e-05,
+      "loss": 0.9837,
+      "step": 6347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029887252836746e-05,
+      "loss": 1.1388,
+      "step": 6348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029520816629858e-05,
+      "loss": 0.9925,
+      "step": 6349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029154314759295e-05,
+      "loss": 1.0793,
+      "step": 6350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9028787747227713e-05,
+      "loss": 1.0751,
+      "step": 6351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902842111403779e-05,
+      "loss": 1.2136,
+      "step": 6352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9028054415192182e-05,
+      "loss": 1.0443,
+      "step": 6353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902768765069356e-05,
+      "loss": 0.9962,
+      "step": 6354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9027320820544592e-05,
+      "loss": 1.2412,
+      "step": 6355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026953924747943e-05,
+      "loss": 0.9194,
+      "step": 6356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026586963306285e-05,
+      "loss": 1.0432,
+      "step": 6357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026219936222284e-05,
+      "loss": 0.9988,
+      "step": 6358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902585284349861e-05,
+      "loss": 1.3354,
+      "step": 6359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902548568513793e-05,
+      "loss": 2.1555,
+      "step": 6360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902511846114291e-05,
+      "loss": 0.9509,
+      "step": 6361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024751171516237e-05,
+      "loss": 1.2522,
+      "step": 6362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024383816260566e-05,
+      "loss": 1.2852,
+      "step": 6363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024016395378574e-05,
+      "loss": 1.0276,
+      "step": 6364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9023648908872932e-05,
+      "loss": 1.109,
+      "step": 6365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9023281356746316e-05,
+      "loss": 0.9991,
+      "step": 6366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022913739001395e-05,
+      "loss": 1.1223,
+      "step": 6367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902254605564084e-05,
+      "loss": 1.4254,
+      "step": 6368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022178306667332e-05,
+      "loss": 1.3191,
+      "step": 6369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021810492083542e-05,
+      "loss": 0.9753,
+      "step": 6370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902144261189214e-05,
+      "loss": 0.9789,
+      "step": 6371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902107466609581e-05,
+      "loss": 0.9874,
+      "step": 6372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902070665469722e-05,
+      "loss": 1.3143,
+      "step": 6373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902033857769905e-05,
+      "loss": 1.5424,
+      "step": 6374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019970435103978e-05,
+      "loss": 0.9459,
+      "step": 6375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019602226914677e-05,
+      "loss": 1.5329,
+      "step": 6376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901923395313383e-05,
+      "loss": 1.1728,
+      "step": 6377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901886561376411e-05,
+      "loss": 1.363,
+      "step": 6378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018497208808197e-05,
+      "loss": 0.936,
+      "step": 6379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018128738268774e-05,
+      "loss": 0.936,
+      "step": 6380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9017760202148513e-05,
+      "loss": 1.3573,
+      "step": 6381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90173916004501e-05,
+      "loss": 1.0369,
+      "step": 6382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9017022933176216e-05,
+      "loss": 0.8587,
+      "step": 6383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016654200329538e-05,
+      "loss": 1.1939,
+      "step": 6384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901628540191275e-05,
+      "loss": 1.235,
+      "step": 6385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901591653792853e-05,
+      "loss": 1.0864,
+      "step": 6386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901554760837957e-05,
+      "loss": 1.341,
+      "step": 6387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015178613268544e-05,
+      "loss": 0.6707,
+      "step": 6388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9014809552598136e-05,
+      "loss": 1.1581,
+      "step": 6389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9014440426371034e-05,
+      "loss": 1.1072,
+      "step": 6390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901407123458992e-05,
+      "loss": 0.9131,
+      "step": 6391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901370197725748e-05,
+      "loss": 1.1245,
+      "step": 6392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9013332654376395e-05,
+      "loss": 1.5527,
+      "step": 6393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012963265949356e-05,
+      "loss": 0.9159,
+      "step": 6394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012593811979047e-05,
+      "loss": 1.062,
+      "step": 6395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012224292468157e-05,
+      "loss": 1.1995,
+      "step": 6396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011854707419367e-05,
+      "loss": 1.2717,
+      "step": 6397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011485056835374e-05,
+      "loss": 1.4128,
+      "step": 6398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011115340718857e-05,
+      "loss": 0.9319,
+      "step": 6399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901074555907251e-05,
+      "loss": 1.0932,
+      "step": 6400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010375711899017e-05,
+      "loss": 0.8975,
+      "step": 6401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010005799201078e-05,
+      "loss": 0.9034,
+      "step": 6402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9009635820981372e-05,
+      "loss": 1.2448,
+      "step": 6403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9009265777242595e-05,
+      "loss": 0.8017,
+      "step": 6404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9008895667987434e-05,
+      "loss": 1.4379,
+      "step": 6405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900852549321859e-05,
+      "loss": 1.0524,
+      "step": 6406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900815525293874e-05,
+      "loss": 1.0356,
+      "step": 6407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007784947150592e-05,
+      "loss": 1.314,
+      "step": 6408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900741457585683e-05,
+      "loss": 0.738,
+      "step": 6409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007044139060143e-05,
+      "loss": 1.5122,
+      "step": 6410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006673636763237e-05,
+      "loss": 1.1787,
+      "step": 6411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006303068968797e-05,
+      "loss": 1.0904,
+      "step": 6412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9005932435679522e-05,
+      "loss": 1.2518,
+      "step": 6413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9005561736898105e-05,
+      "loss": 1.0875,
+      "step": 6414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9005190972627247e-05,
+      "loss": 1.1598,
+      "step": 6415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004820142869636e-05,
+      "loss": 1.103,
+      "step": 6416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004449247627976e-05,
+      "loss": 1.1863,
+      "step": 6417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900407828690496e-05,
+      "loss": 1.1097,
+      "step": 6418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003707260703285e-05,
+      "loss": 0.9185,
+      "step": 6419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003336169025655e-05,
+      "loss": 0.9561,
+      "step": 6420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900296501187476e-05,
+      "loss": 0.7621,
+      "step": 6421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002593789253304e-05,
+      "loss": 1.0367,
+      "step": 6422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900222250116399e-05,
+      "loss": 1.0643,
+      "step": 6423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001851147609512e-05,
+      "loss": 1.2435,
+      "step": 6424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001479728592574e-05,
+      "loss": 1.24,
+      "step": 6425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001108244115877e-05,
+      "loss": 1.0766,
+      "step": 6426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900073669418212e-05,
+      "loss": 1.2789,
+      "step": 6427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9000365078794003e-05,
+      "loss": 1.0219,
+      "step": 6428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999993397954234e-05,
+      "loss": 1.1335,
+      "step": 6429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999621651665514e-05,
+      "loss": 1.24,
+      "step": 6430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999249839930546e-05,
+      "loss": 1.0276,
+      "step": 6431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998877962752033e-05,
+      "loss": 1.1326,
+      "step": 6432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899850602013268e-05,
+      "loss": 0.6421,
+      "step": 6433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998134012075193e-05,
+      "loss": 0.9037,
+      "step": 6434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997761938582277e-05,
+      "loss": 0.907,
+      "step": 6435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997389799656635e-05,
+      "loss": 0.9905,
+      "step": 6436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997017595300977e-05,
+      "loss": 1.1065,
+      "step": 6437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996645325518008e-05,
+      "loss": 0.8131,
+      "step": 6438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996272990310434e-05,
+      "loss": 1.0491,
+      "step": 6439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8995900589680962e-05,
+      "loss": 0.8787,
+      "step": 6440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8995528123632307e-05,
+      "loss": 1.1113,
+      "step": 6441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899515559216717e-05,
+      "loss": 1.0831,
+      "step": 6442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899478299528826e-05,
+      "loss": 0.8872,
+      "step": 6443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899441033299829e-05,
+      "loss": 0.8931,
+      "step": 6444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994037605299972e-05,
+      "loss": 1.0713,
+      "step": 6445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993664812196013e-05,
+      "loss": 0.7156,
+      "step": 6446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993291953689126e-05,
+      "loss": 1.1509,
+      "step": 6447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992919029782018e-05,
+      "loss": 1.0508,
+      "step": 6448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992546040477405e-05,
+      "loss": 1.0389,
+      "step": 6449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992172985778002e-05,
+      "loss": 1.0547,
+      "step": 6450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8991799865686514e-05,
+      "loss": 0.9919,
+      "step": 6451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899142668020566e-05,
+      "loss": 1.2115,
+      "step": 6452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8991053429338155e-05,
+      "loss": 0.8542,
+      "step": 6453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990680113086708e-05,
+      "loss": 0.7934,
+      "step": 6454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990306731454038e-05,
+      "loss": 0.938,
+      "step": 6455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989933284442862e-05,
+      "loss": 1.2126,
+      "step": 6456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989559772055888e-05,
+      "loss": 1.0879,
+      "step": 6457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898918619429584e-05,
+      "loss": 0.9076,
+      "step": 6458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898881255116543e-05,
+      "loss": 1.002,
+      "step": 6459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988438842667377e-05,
+      "loss": 1.0764,
+      "step": 6460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89880650688044e-05,
+      "loss": 1.3769,
+      "step": 6461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987691229579214e-05,
+      "loss": 1.0639,
+      "step": 6462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987317324994545e-05,
+      "loss": 1.375,
+      "step": 6463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986943355053098e-05,
+      "loss": 1.1158,
+      "step": 6464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986569319757605e-05,
+      "loss": 0.9102,
+      "step": 6465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986195219110782e-05,
+      "loss": 0.9589,
+      "step": 6466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898582105311535e-05,
+      "loss": 1.2091,
+      "step": 6467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898544682177403e-05,
+      "loss": 1.0594,
+      "step": 6468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898507252508954e-05,
+      "loss": 1.1552,
+      "step": 6469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898469816306461e-05,
+      "loss": 0.9194,
+      "step": 6470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984323735701952e-05,
+      "loss": 1.0498,
+      "step": 6471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983949243004293e-05,
+      "loss": 0.8154,
+      "step": 6472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983574684974363e-05,
+      "loss": 0.9769,
+      "step": 6473
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983200061614876e-05,
+      "loss": 0.8481,
+      "step": 6474
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898282537292856e-05,
+      "loss": 0.9929,
+      "step": 6475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8982450618918145e-05,
+      "loss": 0.9883,
+      "step": 6476
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8982075799586347e-05,
+      "loss": 1.3787,
+      "step": 6477
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89817009149359e-05,
+      "loss": 1.0901,
+      "step": 6478
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8981325964969525e-05,
+      "loss": 1.1998,
+      "step": 6479
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980950949689952e-05,
+      "loss": 0.9521,
+      "step": 6480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89805758690999e-05,
+      "loss": 1.0411,
+      "step": 6481
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980200723202112e-05,
+      "loss": 0.7944,
+      "step": 6482
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89798255119993e-05,
+      "loss": 0.9771,
+      "step": 6483
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979450235494204e-05,
+      "loss": 1.0518,
+      "step": 6484
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897907489368955e-05,
+      "loss": 0.9216,
+      "step": 6485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897869948658806e-05,
+      "loss": 1.1163,
+      "step": 6486
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8978324014192474e-05,
+      "loss": 0.8655,
+      "step": 6487
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977948476505518e-05,
+      "loss": 0.9025,
+      "step": 6488
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977572873529926e-05,
+      "loss": 1.1579,
+      "step": 6489
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977197205268427e-05,
+      "loss": 1.0408,
+      "step": 6490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897682147172375e-05,
+      "loss": 1.0269,
+      "step": 6491
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897644567289863e-05,
+      "loss": 0.8426,
+      "step": 6492
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976069808795804e-05,
+      "loss": 1.0991,
+      "step": 6493
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8975693879417998e-05,
+      "loss": 1.1876,
+      "step": 6494
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897531788476795e-05,
+      "loss": 1.1311,
+      "step": 6495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974941824848398e-05,
+      "loss": 0.9343,
+      "step": 6496
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897456569966207e-05,
+      "loss": 1.2199,
+      "step": 6497
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89741895092117e-05,
+      "loss": 0.9805,
+      "step": 6498
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973813253500027e-05,
+      "loss": 0.7471,
+      "step": 6499
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973436932529795e-05,
+      "loss": 0.9488,
+      "step": 6500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897306054630373e-05,
+      "loss": 1.1489,
+      "step": 6501
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897268409482457e-05,
+      "loss": 0.941,
+      "step": 6502
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972307578095054e-05,
+      "loss": 1.0053,
+      "step": 6503
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897193099611792e-05,
+      "loss": 0.9469,
+      "step": 6504
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971554348895914e-05,
+      "loss": 1.2363,
+      "step": 6505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971177636431764e-05,
+      "loss": 0.9822,
+      "step": 6506
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970800858728214e-05,
+      "loss": 0.8431,
+      "step": 6507
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970424015788006e-05,
+      "loss": 0.8469,
+      "step": 6508
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970047107613877e-05,
+      "loss": 1.1385,
+      "step": 6509
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896967013420857e-05,
+      "loss": 1.0106,
+      "step": 6510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969293095574826e-05,
+      "loss": 1.2189,
+      "step": 6511
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8968915991715385e-05,
+      "loss": 0.9988,
+      "step": 6512
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896853882263299e-05,
+      "loss": 0.9423,
+      "step": 6513
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896816158833039e-05,
+      "loss": 1.0325,
+      "step": 6514
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8967784288810315e-05,
+      "loss": 1.0637,
+      "step": 6515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896740692407552e-05,
+      "loss": 1.2785,
+      "step": 6516
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896702949412875e-05,
+      "loss": 0.8823,
+      "step": 6517
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8966651998972743e-05,
+      "loss": 1.1595,
+      "step": 6518
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8966274438610246e-05,
+      "loss": 0.9624,
+      "step": 6519
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965896813044005e-05,
+      "loss": 0.9643,
+      "step": 6520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965519122276774e-05,
+      "loss": 1.2486,
+      "step": 6521
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965141366311285e-05,
+      "loss": 0.7887,
+      "step": 6522
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964763545150294e-05,
+      "loss": 0.9872,
+      "step": 6523
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964385658796548e-05,
+      "loss": 1.0621,
+      "step": 6524
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896400770725279e-05,
+      "loss": 1.07,
+      "step": 6525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963629690521776e-05,
+      "loss": 1.1963,
+      "step": 6526
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963251608606252e-05,
+      "loss": 1.4236,
+      "step": 6527
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962873461508965e-05,
+      "loss": 1.0554,
+      "step": 6528
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962495249232668e-05,
+      "loss": 1.0382,
+      "step": 6529
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896211697178011e-05,
+      "loss": 0.8764,
+      "step": 6530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961738629154043e-05,
+      "loss": 1.1391,
+      "step": 6531
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961360221357216e-05,
+      "loss": 0.8944,
+      "step": 6532
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960981748392388e-05,
+      "loss": 0.9322,
+      "step": 6533
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89606032102623e-05,
+      "loss": 1.1637,
+      "step": 6534
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896022460696971e-05,
+      "loss": 1.2555,
+      "step": 6535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895984593851737e-05,
+      "loss": 0.9097,
+      "step": 6536
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895946720490804e-05,
+      "loss": 0.893,
+      "step": 6537
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895908840614447e-05,
+      "loss": 1.1871,
+      "step": 6538
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895870954222941e-05,
+      "loss": 0.8434,
+      "step": 6539
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8958330613165622e-05,
+      "loss": 0.8246,
+      "step": 6540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8957951618955858e-05,
+      "loss": 1.1756,
+      "step": 6541
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8957572559602877e-05,
+      "loss": 1.0002,
+      "step": 6542
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895719343510943e-05,
+      "loss": 0.9574,
+      "step": 6543
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956814245478282e-05,
+      "loss": 1.0442,
+      "step": 6544
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956434990712185e-05,
+      "loss": 1.1726,
+      "step": 6545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89560556708139e-05,
+      "loss": 1.0591,
+      "step": 6546
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895567628578618e-05,
+      "loss": 1.0563,
+      "step": 6547
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8955296835631786e-05,
+      "loss": 1.2239,
+      "step": 6548
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8954917320353482e-05,
+      "loss": 1.1214,
+      "step": 6549
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8954537739954024e-05,
+      "loss": 1.4501,
+      "step": 6550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895415809443617e-05,
+      "loss": 0.6598,
+      "step": 6551
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953778383802688e-05,
+      "loss": 1.079,
+      "step": 6552
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953398608056335e-05,
+      "loss": 0.8693,
+      "step": 6553
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953018767199874e-05,
+      "loss": 1.2497,
+      "step": 6554
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952638861236066e-05,
+      "loss": 0.7678,
+      "step": 6555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952258890167673e-05,
+      "loss": 1.1536,
+      "step": 6556
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895187885399746e-05,
+      "loss": 1.1754,
+      "step": 6557
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895149875272819e-05,
+      "loss": 1.0849,
+      "step": 6558
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895111858636262e-05,
+      "loss": 1.1168,
+      "step": 6559
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895073835490353e-05,
+      "loss": 0.9378,
+      "step": 6560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8950358058353674e-05,
+      "loss": 1.0739,
+      "step": 6561
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894997769671582e-05,
+      "loss": 1.0124,
+      "step": 6562
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894959726999274e-05,
+      "loss": 0.8114,
+      "step": 6563
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894921677818719e-05,
+      "loss": 0.9804,
+      "step": 6564
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894883622130194e-05,
+      "loss": 1.0899,
+      "step": 6565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894845559933976e-05,
+      "loss": 0.7385,
+      "step": 6566
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894807491230342e-05,
+      "loss": 0.9996,
+      "step": 6567
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947694160195684e-05,
+      "loss": 0.7533,
+      "step": 6568
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947313343019324e-05,
+      "loss": 1.101,
+      "step": 6569
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8946932460777105e-05,
+      "loss": 1.4362,
+      "step": 6570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89465515134718e-05,
+      "loss": 1.0975,
+      "step": 6571
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894617050110618e-05,
+      "loss": 1.0698,
+      "step": 6572
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945789423683015e-05,
+      "loss": 1.0253,
+      "step": 6573
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894540828120507e-05,
+      "loss": 1.3845,
+      "step": 6574
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945027073675132e-05,
+      "loss": 0.8604,
+      "step": 6575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944645801095956e-05,
+      "loss": 1.0553,
+      "step": 6576
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894426446347033e-05,
+      "loss": 0.952,
+      "step": 6577
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8943883060801014e-05,
+      "loss": 0.8992,
+      "step": 6578
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8943501593090788e-05,
+      "loss": 1.1558,
+      "step": 6579
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8943120060342427e-05,
+      "loss": 1.0283,
+      "step": 6580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89427384625587e-05,
+      "loss": 1.4718,
+      "step": 6581
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894235679974239e-05,
+      "loss": 1.2991,
+      "step": 6582
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894197507189627e-05,
+      "loss": 1.1272,
+      "step": 6583
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941593279023108e-05,
+      "loss": 1.1532,
+      "step": 6584
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894121142112569e-05,
+      "loss": 1.1995,
+      "step": 6585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894082949820679e-05,
+      "loss": 1.0561,
+      "step": 6586
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940447510269186e-05,
+      "loss": 1.0336,
+      "step": 6587
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940065457315656e-05,
+      "loss": 0.8261,
+      "step": 6588
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939683339348972e-05,
+      "loss": 1.0938,
+      "step": 6589
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893930115637192e-05,
+      "loss": 0.9796,
+      "step": 6590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893891890838728e-05,
+      "loss": 0.8316,
+      "step": 6591
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938536595397826e-05,
+      "loss": 0.9727,
+      "step": 6592
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938154217406343e-05,
+      "loss": 0.6567,
+      "step": 6593
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893777177441561e-05,
+      "loss": 0.8423,
+      "step": 6594
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893738926642841e-05,
+      "loss": 1.3998,
+      "step": 6595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893700669344752e-05,
+      "loss": 0.99,
+      "step": 6596
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893662405547572e-05,
+      "loss": 0.9737,
+      "step": 6597
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936241352515806e-05,
+      "loss": 1.0951,
+      "step": 6598
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935858584570547e-05,
+      "loss": 0.8865,
+      "step": 6599
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935475751642736e-05,
+      "loss": 1.1157,
+      "step": 6600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935092853735152e-05,
+      "loss": 1.1102,
+      "step": 6601
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893470989085058e-05,
+      "loss": 0.8858,
+      "step": 6602
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934326862991804e-05,
+      "loss": 0.8246,
+      "step": 6603
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933943770161613e-05,
+      "loss": 1.1112,
+      "step": 6604
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933560612362793e-05,
+      "loss": 1.1441,
+      "step": 6605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933177389598124e-05,
+      "loss": 1.3191,
+      "step": 6606
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89327941018704e-05,
+      "loss": 0.8082,
+      "step": 6607
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932410749182403e-05,
+      "loss": 1.0883,
+      "step": 6608
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932027331536923e-05,
+      "loss": 1.1214,
+      "step": 6609
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893164384893675e-05,
+      "loss": 1.3891,
+      "step": 6610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893126030138467e-05,
+      "loss": 0.9553,
+      "step": 6611
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893087668888347e-05,
+      "loss": 1.1206,
+      "step": 6612
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893049301143595e-05,
+      "loss": 1.1104,
+      "step": 6613
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8930109269044886e-05,
+      "loss": 0.934,
+      "step": 6614
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929725461713083e-05,
+      "loss": 0.8799,
+      "step": 6615
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892934158944332e-05,
+      "loss": 0.8946,
+      "step": 6616
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928957652238392e-05,
+      "loss": 1.132,
+      "step": 6617
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928573650101097e-05,
+      "loss": 1.1086,
+      "step": 6618
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928189583034224e-05,
+      "loss": 0.9502,
+      "step": 6619
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892780545104056e-05,
+      "loss": 0.9586,
+      "step": 6620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8927421254122907e-05,
+      "loss": 1.1332,
+      "step": 6621
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892703699228405e-05,
+      "loss": 1.158,
+      "step": 6622
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8926652665526792e-05,
+      "loss": 1.3553,
+      "step": 6623
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892626827385393e-05,
+      "loss": 1.0443,
+      "step": 6624
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925883817268245e-05,
+      "loss": 0.7016,
+      "step": 6625
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925499295772546e-05,
+      "loss": 1.149,
+      "step": 6626
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925114709369623e-05,
+      "loss": 1.1237,
+      "step": 6627
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924730058062276e-05,
+      "loss": 0.8049,
+      "step": 6628
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89243453418533e-05,
+      "loss": 1.3336,
+      "step": 6629
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923960560745495e-05,
+      "loss": 1.0843,
+      "step": 6630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892357571474166e-05,
+      "loss": 0.9106,
+      "step": 6631
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923190803844585e-05,
+      "loss": 0.9787,
+      "step": 6632
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8922805828057083e-05,
+      "loss": 0.9374,
+      "step": 6633
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8922420787381938e-05,
+      "loss": 1.0803,
+      "step": 6634
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8922035681821965e-05,
+      "loss": 0.8114,
+      "step": 6635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921650511379955e-05,
+      "loss": 1.2871,
+      "step": 6636
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921265276058712e-05,
+      "loss": 1.0182,
+      "step": 6637
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920879975861037e-05,
+      "loss": 1.0429,
+      "step": 6638
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920494610789735e-05,
+      "loss": 1.2167,
+      "step": 6639
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920109180847602e-05,
+      "loss": 0.9621,
+      "step": 6640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919723686037447e-05,
+      "loss": 0.9084,
+      "step": 6641
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919338126362067e-05,
+      "loss": 1.0917,
+      "step": 6642
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918952501824275e-05,
+      "loss": 1.1936,
+      "step": 6643
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8918566812426866e-05,
+      "loss": 1.2174,
+      "step": 6644
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891818105817265e-05,
+      "loss": 1.0228,
+      "step": 6645
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891779523906443e-05,
+      "loss": 1.1924,
+      "step": 6646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917409355105015e-05,
+      "loss": 0.9969,
+      "step": 6647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917023406297207e-05,
+      "loss": 1.2946,
+      "step": 6648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8916637392643816e-05,
+      "loss": 1.3213,
+      "step": 6649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8916251314147647e-05,
+      "loss": 0.9167,
+      "step": 6650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915865170811506e-05,
+      "loss": 1.1883,
+      "step": 6651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915478962638208e-05,
+      "loss": 1.2934,
+      "step": 6652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915092689630555e-05,
+      "loss": 0.9864,
+      "step": 6653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914706351791357e-05,
+      "loss": 1.2237,
+      "step": 6654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914319949123424e-05,
+      "loss": 0.9649,
+      "step": 6655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8913933481629567e-05,
+      "loss": 0.865,
+      "step": 6656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89135469493126e-05,
+      "loss": 1.2823,
+      "step": 6657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8913160352175325e-05,
+      "loss": 1.0328,
+      "step": 6658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8912773690220558e-05,
+      "loss": 1.0229,
+      "step": 6659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891238696345111e-05,
+      "loss": 0.9294,
+      "step": 6660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89120001718698e-05,
+      "loss": 0.8552,
+      "step": 6661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891161331547943e-05,
+      "loss": 0.8027,
+      "step": 6662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8911226394282823e-05,
+      "loss": 1.1026,
+      "step": 6663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891083940828278e-05,
+      "loss": 1.0607,
+      "step": 6664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891045235748213e-05,
+      "loss": 1.1175,
+      "step": 6665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891006524188368e-05,
+      "loss": 1.3366,
+      "step": 6666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909678061490244e-05,
+      "loss": 0.9037,
+      "step": 6667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909290816304643e-05,
+      "loss": 0.9402,
+      "step": 6668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908903506329688e-05,
+      "loss": 0.93,
+      "step": 6669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89085161315682e-05,
+      "loss": 0.9354,
+      "step": 6670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908128692022987e-05,
+      "loss": 1.2399,
+      "step": 6671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8907741187696878e-05,
+      "loss": 0.8333,
+      "step": 6672
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8907353618592684e-05,
+      "loss": 0.9091,
+      "step": 6673
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906965984713224e-05,
+      "loss": 1.0173,
+      "step": 6674
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906578286061325e-05,
+      "loss": 0.8858,
+      "step": 6675
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890619052263979e-05,
+      "loss": 1.0523,
+      "step": 6676
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905802694451458e-05,
+      "loss": 0.9171,
+      "step": 6677
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905414801499135e-05,
+      "loss": 1.2977,
+      "step": 6678
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905026843785647e-05,
+      "loss": 0.9712,
+      "step": 6679
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904638821313814e-05,
+      "loss": 1.1746,
+      "step": 6680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890425073408646e-05,
+      "loss": 1.1462,
+      "step": 6681
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903862582106405e-05,
+      "loss": 0.9341,
+      "step": 6682
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903474365376473e-05,
+      "loss": 1.1003,
+      "step": 6683
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903086083899485e-05,
+      "loss": 0.9051,
+      "step": 6684
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890269773767827e-05,
+      "loss": 1.144,
+      "step": 6685
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902309326715646e-05,
+      "loss": 0.7098,
+      "step": 6686
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890192085101444e-05,
+      "loss": 0.7785,
+      "step": 6687
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8901532310577474e-05,
+      "loss": 1.0382,
+      "step": 6688
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8901143705407584e-05,
+      "loss": 0.7524,
+      "step": 6689
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890075503550758e-05,
+      "loss": 1.4182,
+      "step": 6690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900366300880302e-05,
+      "loss": 1.1308,
+      "step": 6691
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8899977501528573e-05,
+      "loss": 0.8201,
+      "step": 6692
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889958863745522e-05,
+      "loss": 1.141,
+      "step": 6693
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8899199708663065e-05,
+      "loss": 1.0554,
+      "step": 6694
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898810715154945e-05,
+      "loss": 1.1373,
+      "step": 6695
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898421656933684e-05,
+      "loss": 1.0528,
+      "step": 6696
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889803253400211e-05,
+      "loss": 1.0418,
+      "step": 6697
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8897643346363062e-05,
+      "loss": 1.1003,
+      "step": 6698
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889725409401936e-05,
+      "loss": 0.731,
+      "step": 6699
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889686477697384e-05,
+      "loss": 1.3992,
+      "step": 6700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896475395229327e-05,
+      "loss": 1.2121,
+      "step": 6701
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889608594878866e-05,
+      "loss": 0.9354,
+      "step": 6702
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8895696437654668e-05,
+      "loss": 1.1265,
+      "step": 6703
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8895306861830184e-05,
+      "loss": 0.7585,
+      "step": 6704
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894917221318038e-05,
+      "loss": 1.0046,
+      "step": 6705
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889452751612107e-05,
+      "loss": 1.1259,
+      "step": 6706
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889413774624211e-05,
+      "loss": 1.0411,
+      "step": 6707
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893747911683987e-05,
+      "loss": 1.1197,
+      "step": 6708
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893358012449547e-05,
+      "loss": 1.2222,
+      "step": 6709
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889296804854162e-05,
+      "loss": 0.8984,
+      "step": 6710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892578019963038e-05,
+      "loss": 0.7213,
+      "step": 6711
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892187926716646e-05,
+      "loss": 1.0484,
+      "step": 6712
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889179776880527e-05,
+      "loss": 1.2987,
+      "step": 6713
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889140754623176e-05,
+      "loss": 1.0183,
+      "step": 6714
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8891017258998944e-05,
+      "loss": 1.1835,
+      "step": 6715
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890626907109663e-05,
+      "loss": 0.9589,
+      "step": 6716
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890236490566754e-05,
+      "loss": 1.0664,
+      "step": 6717
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888984600937306e-05,
+      "loss": 0.7922,
+      "step": 6718
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8889455463531418e-05,
+      "loss": 0.8892,
+      "step": 6719
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888906485304467e-05,
+      "loss": 1.0969,
+      "step": 6720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888674177915652e-05,
+      "loss": 1.1599,
+      "step": 6721
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888828343814721e-05,
+      "loss": 0.9826,
+      "step": 6722
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887892633742188e-05,
+      "loss": 1.2683,
+      "step": 6723
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887501764703416e-05,
+      "loss": 0.9491,
+      "step": 6724
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888711083103375e-05,
+      "loss": 1.2216,
+      "step": 6725
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886719832736026e-05,
+      "loss": 1.1988,
+      "step": 6726
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886328769813086e-05,
+      "loss": 0.7626,
+      "step": 6727
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888593764226778e-05,
+      "loss": 0.9185,
+      "step": 6728
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885546450102947e-05,
+      "loss": 1.0707,
+      "step": 6729
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885155193321433e-05,
+      "loss": 1.0486,
+      "step": 6730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8884763871926085e-05,
+      "loss": 1.1578,
+      "step": 6731
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8884372485919747e-05,
+      "loss": 1.1468,
+      "step": 6732
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883981035305267e-05,
+      "loss": 1.1557,
+      "step": 6733
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883589520085487e-05,
+      "loss": 1.2581,
+      "step": 6734
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888319794026326e-05,
+      "loss": 0.8723,
+      "step": 6735
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888280629584143e-05,
+      "loss": 1.0049,
+      "step": 6736
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882414586822846e-05,
+      "loss": 1.1864,
+      "step": 6737
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8882022813210363e-05,
+      "loss": 1.0191,
+      "step": 6738
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881630975006816e-05,
+      "loss": 1.0781,
+      "step": 6739
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8881239072215063e-05,
+      "loss": 1.1703,
+      "step": 6740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880847104837955e-05,
+      "loss": 0.9455,
+      "step": 6741
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880455072878342e-05,
+      "loss": 0.9649,
+      "step": 6742
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888006297633907e-05,
+      "loss": 0.9774,
+      "step": 6743
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8879670815222997e-05,
+      "loss": 1.7486,
+      "step": 6744
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887927858953297e-05,
+      "loss": 1.5444,
+      "step": 6745
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8878886299271846e-05,
+      "loss": 1.1953,
+      "step": 6746
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887849394444247e-05,
+      "loss": 1.0915,
+      "step": 6747
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88781015250477e-05,
+      "loss": 0.8284,
+      "step": 6748
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8877709041090397e-05,
+      "loss": 1.1652,
+      "step": 6749
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88773164925734e-05,
+      "loss": 1.0114,
+      "step": 6750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887692387949958e-05,
+      "loss": 1.0773,
+      "step": 6751
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8876531201871776e-05,
+      "loss": 1.1805,
+      "step": 6752
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8876138459692855e-05,
+      "loss": 1.0666,
+      "step": 6753
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887574565296567e-05,
+      "loss": 1.1866,
+      "step": 6754
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875352781693074e-05,
+      "loss": 0.8309,
+      "step": 6755
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874959845877932e-05,
+      "loss": 0.8352,
+      "step": 6756
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887456684552309e-05,
+      "loss": 1.303,
+      "step": 6757
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874173780631418e-05,
+      "loss": 0.7674,
+      "step": 6758
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873780651205768e-05,
+      "loss": 1.0371,
+      "step": 6759
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8873387457248994e-05,
+      "loss": 1.0737,
+      "step": 6760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872994198763966e-05,
+      "loss": 1.0559,
+      "step": 6761
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887260087575354e-05,
+      "loss": 0.8797,
+      "step": 6762
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872207488220573e-05,
+      "loss": 0.8519,
+      "step": 6763
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871814036167927e-05,
+      "loss": 0.7703,
+      "step": 6764
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887142051959847e-05,
+      "loss": 0.8458,
+      "step": 6765
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887102693851505e-05,
+      "loss": 0.7927,
+      "step": 6766
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887063329292054e-05,
+      "loss": 1.1882,
+      "step": 6767
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88702395828178e-05,
+      "loss": 1.1169,
+      "step": 6768
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8869845808209693e-05,
+      "loss": 1.0587,
+      "step": 6769
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886945196909908e-05,
+      "loss": 1.2667,
+      "step": 6770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886905806548883e-05,
+      "loss": 1.2233,
+      "step": 6771
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868664097381805e-05,
+      "loss": 1.0386,
+      "step": 6772
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868270064780868e-05,
+      "loss": 1.2975,
+      "step": 6773
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867875967688885e-05,
+      "loss": 0.851,
+      "step": 6774
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867481806108726e-05,
+      "loss": 1.2693,
+      "step": 6775
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886708758004326e-05,
+      "loss": 0.9234,
+      "step": 6776
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886669328949534e-05,
+      "loss": 1.1727,
+      "step": 6777
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8866298934467842e-05,
+      "loss": 0.9569,
+      "step": 6778
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865904514963634e-05,
+      "loss": 1.0289,
+      "step": 6779
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865510030985588e-05,
+      "loss": 1.0696,
+      "step": 6780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8865115482536565e-05,
+      "loss": 1.4646,
+      "step": 6781
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886472086961944e-05,
+      "loss": 0.9934,
+      "step": 6782
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886432619223708e-05,
+      "loss": 1.148,
+      "step": 6783
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863931450392354e-05,
+      "loss": 0.9731,
+      "step": 6784
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863536644088137e-05,
+      "loss": 0.9749,
+      "step": 6785
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863141773327292e-05,
+      "loss": 1.1449,
+      "step": 6786
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88627468381127e-05,
+      "loss": 0.9893,
+      "step": 6787
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886235183844723e-05,
+      "loss": 0.9584,
+      "step": 6788
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861956774333752e-05,
+      "loss": 0.9715,
+      "step": 6789
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861561645775136e-05,
+      "loss": 1.2062,
+      "step": 6790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861166452774268e-05,
+      "loss": 0.9454,
+      "step": 6791
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860771195334005e-05,
+      "loss": 0.991,
+      "step": 6792
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8860375873457233e-05,
+      "loss": 0.9359,
+      "step": 6793
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859980487146826e-05,
+      "loss": 1.1415,
+      "step": 6794
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859585036405653e-05,
+      "loss": 1.2263,
+      "step": 6795
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88591895212366e-05,
+      "loss": 1.6031,
+      "step": 6796
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885879394164253e-05,
+      "loss": 0.8725,
+      "step": 6797
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8858398297626334e-05,
+      "loss": 0.9834,
+      "step": 6798
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8858002589190873e-05,
+      "loss": 0.7837,
+      "step": 6799
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857606816339042e-05,
+      "loss": 0.9582,
+      "step": 6800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857210979073705e-05,
+      "loss": 1.1448,
+      "step": 6801
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856815077397747e-05,
+      "loss": 1.2013,
+      "step": 6802
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856419111314048e-05,
+      "loss": 0.8418,
+      "step": 6803
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856023080825486e-05,
+      "loss": 0.8791,
+      "step": 6804
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885562698593494e-05,
+      "loss": 0.7312,
+      "step": 6805
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885523082664529e-05,
+      "loss": 0.85,
+      "step": 6806
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885483460295942e-05,
+      "loss": 1.0184,
+      "step": 6807
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854438314880206e-05,
+      "loss": 0.7728,
+      "step": 6808
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854041962410535e-05,
+      "loss": 1.0957,
+      "step": 6809
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885364554555329e-05,
+      "loss": 0.9736,
+      "step": 6810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885324906431135e-05,
+      "loss": 0.8753,
+      "step": 6811
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88528525186876e-05,
+      "loss": 1.1369,
+      "step": 6812
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8852455908684924e-05,
+      "loss": 1.1462,
+      "step": 6813
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8852059234306203e-05,
+      "loss": 0.8697,
+      "step": 6814
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851662495554326e-05,
+      "loss": 0.8514,
+      "step": 6815
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851265692432176e-05,
+      "loss": 1.1265,
+      "step": 6816
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850868824942638e-05,
+      "loss": 1.087,
+      "step": 6817
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88504718930886e-05,
+      "loss": 0.9834,
+      "step": 6818
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885007489687295e-05,
+      "loss": 0.8882,
+      "step": 6819
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884967783629857e-05,
+      "loss": 1.0248,
+      "step": 6820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884928071136835e-05,
+      "loss": 0.8573,
+      "step": 6821
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848883522085178e-05,
+      "loss": 0.6825,
+      "step": 6822
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848486268451944e-05,
+      "loss": 1.2978,
+      "step": 6823
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8848088950471534e-05,
+      "loss": 0.9788,
+      "step": 6824
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884769156814684e-05,
+      "loss": 1.1149,
+      "step": 6825
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847294121480746e-05,
+      "loss": 1.0571,
+      "step": 6826
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846896610476147e-05,
+      "loss": 0.9892,
+      "step": 6827
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884649903513594e-05,
+      "loss": 0.9921,
+      "step": 6828
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8846101395463005e-05,
+      "loss": 1.3728,
+      "step": 6829
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884570369146024e-05,
+      "loss": 0.7284,
+      "step": 6830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845305923130534e-05,
+      "loss": 0.9611,
+      "step": 6831
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884490809047678e-05,
+      "loss": 0.7729,
+      "step": 6832
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844510193501874e-05,
+      "loss": 1.1636,
+      "step": 6833
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884411223220871e-05,
+      "loss": 0.8281,
+      "step": 6834
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843714206600174e-05,
+      "loss": 0.9845,
+      "step": 6835
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843316116679168e-05,
+      "loss": 1.1119,
+      "step": 6836
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8842917962448588e-05,
+      "loss": 1.0927,
+      "step": 6837
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8842519743911328e-05,
+      "loss": 1.3802,
+      "step": 6838
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8842121461070277e-05,
+      "loss": 1.0772,
+      "step": 6839
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884172311392834e-05,
+      "loss": 0.7607,
+      "step": 6840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884132470248841e-05,
+      "loss": 1.0991,
+      "step": 6841
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840926226753385e-05,
+      "loss": 1.2144,
+      "step": 6842
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840527686726166e-05,
+      "loss": 0.8464,
+      "step": 6843
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884012908240964e-05,
+      "loss": 1.1361,
+      "step": 6844
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839730413806722e-05,
+      "loss": 0.6211,
+      "step": 6845
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839331680920295e-05,
+      "loss": 0.9585,
+      "step": 6846
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838932883753273e-05,
+      "loss": 1.2026,
+      "step": 6847
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838534022308546e-05,
+      "loss": 0.9967,
+      "step": 6848
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883813509658902e-05,
+      "loss": 0.9709,
+      "step": 6849
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837736106597595e-05,
+      "loss": 1.3764,
+      "step": 6850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837337052337168e-05,
+      "loss": 0.9075,
+      "step": 6851
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836937933810646e-05,
+      "loss": 1.2645,
+      "step": 6852
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883653875102093e-05,
+      "loss": 0.9889,
+      "step": 6853
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836139503970927e-05,
+      "loss": 1.2581,
+      "step": 6854
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883574019266353e-05,
+      "loss": 1.0684,
+      "step": 6855
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8835340817101654e-05,
+      "loss": 1.3997,
+      "step": 6856
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834941377288197e-05,
+      "loss": 1.0826,
+      "step": 6857
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834541873226065e-05,
+      "loss": 0.7835,
+      "step": 6858
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883414230491816e-05,
+      "loss": 1.2846,
+      "step": 6859
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833742672367395e-05,
+      "loss": 1.0887,
+      "step": 6860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883334297557667e-05,
+      "loss": 0.9587,
+      "step": 6861
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832943214548897e-05,
+      "loss": 1.091,
+      "step": 6862
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883254338928698e-05,
+      "loss": 1.0429,
+      "step": 6863
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832143499793823e-05,
+      "loss": 1.1803,
+      "step": 6864
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883174354607234e-05,
+      "loss": 1.0587,
+      "step": 6865
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831343528125437e-05,
+      "loss": 0.9954,
+      "step": 6866
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830943445956024e-05,
+      "loss": 0.8163,
+      "step": 6867
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830543299567008e-05,
+      "loss": 0.8819,
+      "step": 6868
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830143088961304e-05,
+      "loss": 0.8969,
+      "step": 6869
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829742814141813e-05,
+      "loss": 0.9829,
+      "step": 6870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829342475111457e-05,
+      "loss": 0.9698,
+      "step": 6871
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882894207187314e-05,
+      "loss": 1.0525,
+      "step": 6872
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828541604429775e-05,
+      "loss": 1.0549,
+      "step": 6873
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828141072784276e-05,
+      "loss": 0.7611,
+      "step": 6874
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827740476939555e-05,
+      "loss": 1.1,
+      "step": 6875
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827339816898526e-05,
+      "loss": 1.2883,
+      "step": 6876
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88269390926641e-05,
+      "loss": 1.317,
+      "step": 6877
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826538304239192e-05,
+      "loss": 0.8235,
+      "step": 6878
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882613745162672e-05,
+      "loss": 1.0142,
+      "step": 6879
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825736534829593e-05,
+      "loss": 0.6891,
+      "step": 6880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882533555385073e-05,
+      "loss": 0.9935,
+      "step": 6881
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882493450869305e-05,
+      "loss": 0.9314,
+      "step": 6882
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824533399359466e-05,
+      "loss": 0.9223,
+      "step": 6883
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882413222585289e-05,
+      "loss": 0.9502,
+      "step": 6884
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882373098817625e-05,
+      "loss": 1.5221,
+      "step": 6885
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823329686332456e-05,
+      "loss": 1.1858,
+      "step": 6886
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882292832032443e-05,
+      "loss": 1.0942,
+      "step": 6887
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822526890155092e-05,
+      "loss": 0.8773,
+      "step": 6888
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822125395827352e-05,
+      "loss": 0.8616,
+      "step": 6889
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882172383734414e-05,
+      "loss": 1.3309,
+      "step": 6890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821322214708374e-05,
+      "loss": 1.0668,
+      "step": 6891
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882092052792297e-05,
+      "loss": 1.3046,
+      "step": 6892
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820518776990857e-05,
+      "loss": 0.9969,
+      "step": 6893
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882011696191495e-05,
+      "loss": 0.8315,
+      "step": 6894
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881971508269817e-05,
+      "loss": 0.7767,
+      "step": 6895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819313139343445e-05,
+      "loss": 1.2817,
+      "step": 6896
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818911131853696e-05,
+      "loss": 1.0166,
+      "step": 6897
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818509060231843e-05,
+      "loss": 1.1102,
+      "step": 6898
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818106924480817e-05,
+      "loss": 0.9624,
+      "step": 6899
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817704724603536e-05,
+      "loss": 0.9451,
+      "step": 6900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881730246060293e-05,
+      "loss": 0.812,
+      "step": 6901
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816900132481915e-05,
+      "loss": 1.011,
+      "step": 6902
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816497740243428e-05,
+      "loss": 0.8561,
+      "step": 6903
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881609528389039e-05,
+      "loss": 1.1749,
+      "step": 6904
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815692763425725e-05,
+      "loss": 0.95,
+      "step": 6905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815290178852364e-05,
+      "loss": 1.2041,
+      "step": 6906
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881488753017324e-05,
+      "loss": 1.1426,
+      "step": 6907
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814484817391267e-05,
+      "loss": 1.1994,
+      "step": 6908
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881408204050939e-05,
+      "loss": 1.1861,
+      "step": 6909
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813679199530522e-05,
+      "loss": 1.1258,
+      "step": 6910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813276294457603e-05,
+      "loss": 1.3081,
+      "step": 6911
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812873325293558e-05,
+      "loss": 1.1507,
+      "step": 6912
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812470292041322e-05,
+      "loss": 1.1194,
+      "step": 6913
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812067194703823e-05,
+      "loss": 1.002,
+      "step": 6914
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811664033283993e-05,
+      "loss": 0.9728,
+      "step": 6915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811260807784762e-05,
+      "loss": 1.4241,
+      "step": 6916
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810857518209067e-05,
+      "loss": 1.1271,
+      "step": 6917
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810454164559836e-05,
+      "loss": 1.0004,
+      "step": 6918
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810050746840006e-05,
+      "loss": 0.9911,
+      "step": 6919
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809647265052505e-05,
+      "loss": 0.992,
+      "step": 6920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809243719200273e-05,
+      "loss": 0.9627,
+      "step": 6921
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880884010928624e-05,
+      "loss": 1.1818,
+      "step": 6922
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808436435313345e-05,
+      "loss": 1.1193,
+      "step": 6923
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808032697284523e-05,
+      "loss": 1.0737,
+      "step": 6924
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807628895202707e-05,
+      "loss": 1.161,
+      "step": 6925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880722502907084e-05,
+      "loss": 1.0452,
+      "step": 6926
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880682109889185e-05,
+      "loss": 0.7877,
+      "step": 6927
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880641710466868e-05,
+      "loss": 0.9514,
+      "step": 6928
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806013046404268e-05,
+      "loss": 0.9429,
+      "step": 6929
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880560892410155e-05,
+      "loss": 1.3276,
+      "step": 6930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880520473776347e-05,
+      "loss": 0.8703,
+      "step": 6931
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880480048739296e-05,
+      "loss": 0.934,
+      "step": 6932
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804396172992962e-05,
+      "loss": 1.063,
+      "step": 6933
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880399179456642e-05,
+      "loss": 0.8649,
+      "step": 6934
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880358735211627e-05,
+      "loss": 1.3112,
+      "step": 6935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803182845645454e-05,
+      "loss": 0.5819,
+      "step": 6936
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802778275156917e-05,
+      "loss": 1.1713,
+      "step": 6937
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88023736406536e-05,
+      "loss": 1.0931,
+      "step": 6938
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801968942138443e-05,
+      "loss": 1.3445,
+      "step": 6939
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801564179614388e-05,
+      "loss": 0.7234,
+      "step": 6940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801159353084383e-05,
+      "loss": 1.0317,
+      "step": 6941
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800754462551373e-05,
+      "loss": 0.8426,
+      "step": 6942
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800349508018294e-05,
+      "loss": 1.2039,
+      "step": 6943
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87999444894881e-05,
+      "loss": 0.6661,
+      "step": 6944
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879953940696373e-05,
+      "loss": 0.9587,
+      "step": 6945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799134260448134e-05,
+      "loss": 1.063,
+      "step": 6946
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798729049944257e-05,
+      "loss": 1.0952,
+      "step": 6947
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798323775455043e-05,
+      "loss": 1.4257,
+      "step": 6948
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879791843698344e-05,
+      "loss": 1.7997,
+      "step": 6949
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87975130345324e-05,
+      "loss": 1.1512,
+      "step": 6950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797107568104868e-05,
+      "loss": 1.1967,
+      "step": 6951
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796702037703793e-05,
+      "loss": 1.1234,
+      "step": 6952
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796296443332122e-05,
+      "loss": 1.1944,
+      "step": 6953
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879589078499281e-05,
+      "loss": 1.1651,
+      "step": 6954
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795485062688802e-05,
+      "loss": 1.1166,
+      "step": 6955
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795079276423048e-05,
+      "loss": 1.029,
+      "step": 6956
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794673426198503e-05,
+      "loss": 1.1881,
+      "step": 6957
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794267512018115e-05,
+      "loss": 0.8926,
+      "step": 6958
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793861533884836e-05,
+      "loss": 0.8876,
+      "step": 6959
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793455491801623e-05,
+      "loss": 0.9889,
+      "step": 6960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793049385771424e-05,
+      "loss": 1.1853,
+      "step": 6961
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879264321579719e-05,
+      "loss": 0.8732,
+      "step": 6962
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879223698188188e-05,
+      "loss": 1.0104,
+      "step": 6963
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879183068402845e-05,
+      "loss": 1.3101,
+      "step": 6964
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791424322239848e-05,
+      "loss": 0.9157,
+      "step": 6965
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879101789651903e-05,
+      "loss": 1.2715,
+      "step": 6966
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879061140686896e-05,
+      "loss": 1.2189,
+      "step": 6967
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879020485329258e-05,
+      "loss": 1.3459,
+      "step": 6968
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789798235792856e-05,
+      "loss": 0.898,
+      "step": 6969
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789391554372746e-05,
+      "loss": 1.2567,
+      "step": 6970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788984809035204e-05,
+      "loss": 0.89,
+      "step": 6971
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788577999783185e-05,
+      "loss": 0.8137,
+      "step": 6972
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788171126619653e-05,
+      "loss": 0.8306,
+      "step": 6973
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787764189547567e-05,
+      "loss": 1.0499,
+      "step": 6974
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878735718856988e-05,
+      "loss": 0.7781,
+      "step": 6975
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878695012368956e-05,
+      "loss": 1.1588,
+      "step": 6976
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786542994909564e-05,
+      "loss": 1.2612,
+      "step": 6977
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786135802232845e-05,
+      "loss": 1.2196,
+      "step": 6978
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785728545662377e-05,
+      "loss": 1.0835,
+      "step": 6979
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878532122520111e-05,
+      "loss": 1.5275,
+      "step": 6980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784913840852015e-05,
+      "loss": 1.2304,
+      "step": 6981
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878450639261805e-05,
+      "loss": 1.1194,
+      "step": 6982
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878409888050218e-05,
+      "loss": 1.2495,
+      "step": 6983
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783691304507365e-05,
+      "loss": 1.5793,
+      "step": 6984
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783283664636573e-05,
+      "loss": 0.9978,
+      "step": 6985
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782875960892767e-05,
+      "loss": 0.9825,
+      "step": 6986
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878246819327891e-05,
+      "loss": 0.7943,
+      "step": 6987
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782060361797972e-05,
+      "loss": 0.8851,
+      "step": 6988
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781652466452915e-05,
+      "loss": 0.949,
+      "step": 6989
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781244507246706e-05,
+      "loss": 0.9745,
+      "step": 6990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780836484182308e-05,
+      "loss": 0.9075,
+      "step": 6991
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780428397262698e-05,
+      "loss": 1.0355,
+      "step": 6992
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780020246490834e-05,
+      "loss": 0.8339,
+      "step": 6993
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779612031869688e-05,
+      "loss": 1.0352,
+      "step": 6994
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877920375340223e-05,
+      "loss": 0.9871,
+      "step": 6995
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778795411091425e-05,
+      "loss": 1.0348,
+      "step": 6996
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877838700494025e-05,
+      "loss": 1.2479,
+      "step": 6997
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777978534951663e-05,
+      "loss": 1.1258,
+      "step": 6998
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777570001128643e-05,
+      "loss": 1.2857,
+      "step": 6999
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777161403474163e-05,
+      "loss": 0.874,
+      "step": 7000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776752741991187e-05,
+      "loss": 1.1899,
+      "step": 7001
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776344016682692e-05,
+      "loss": 0.8119,
+      "step": 7002
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877593522755165e-05,
+      "loss": 1.034,
+      "step": 7003
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877552637460103e-05,
+      "loss": 0.8905,
+      "step": 7004
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877511745783381e-05,
+      "loss": 0.9095,
+      "step": 7005
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877470847725296e-05,
+      "loss": 1.2605,
+      "step": 7006
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774299432861458e-05,
+      "loss": 0.9259,
+      "step": 7007
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773890324662273e-05,
+      "loss": 1.7017,
+      "step": 7008
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877348115265838e-05,
+      "loss": 0.8235,
+      "step": 7009
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773071916852764e-05,
+      "loss": 0.8852,
+      "step": 7010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772662617248392e-05,
+      "loss": 1.2111,
+      "step": 7011
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772253253848244e-05,
+      "loss": 1.1758,
+      "step": 7012
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771843826655296e-05,
+      "loss": 0.8304,
+      "step": 7013
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771434335672527e-05,
+      "loss": 0.9183,
+      "step": 7014
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771024780902913e-05,
+      "loss": 1.1187,
+      "step": 7015
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770615162349427e-05,
+      "loss": 1.1735,
+      "step": 7016
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877020548001506e-05,
+      "loss": 0.8459,
+      "step": 7017
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769795733902784e-05,
+      "loss": 0.7292,
+      "step": 7018
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769385924015577e-05,
+      "loss": 1.1895,
+      "step": 7019
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768976050356428e-05,
+      "loss": 1.0442,
+      "step": 7020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768566112928305e-05,
+      "loss": 0.9106,
+      "step": 7021
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768156111734197e-05,
+      "loss": 1.0027,
+      "step": 7022
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767746046777085e-05,
+      "loss": 0.966,
+      "step": 7023
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767335918059948e-05,
+      "loss": 0.9774,
+      "step": 7024
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766925725585773e-05,
+      "loss": 1.2164,
+      "step": 7025
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876651546935754e-05,
+      "loss": 0.9971,
+      "step": 7026
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8766105149378235e-05,
+      "loss": 1.0528,
+      "step": 7027
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765694765650837e-05,
+      "loss": 0.8825,
+      "step": 7028
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765284318178332e-05,
+      "loss": 0.8916,
+      "step": 7029
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876487380696371e-05,
+      "loss": 1.0807,
+      "step": 7030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876446323200995e-05,
+      "loss": 0.9287,
+      "step": 7031
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876405259332004e-05,
+      "loss": 0.7931,
+      "step": 7032
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763641890896965e-05,
+      "loss": 1.1524,
+      "step": 7033
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876323112474372e-05,
+      "loss": 0.8913,
+      "step": 7034
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876282029486328e-05,
+      "loss": 1.0229,
+      "step": 7035
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762409401258636e-05,
+      "loss": 0.9067,
+      "step": 7036
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876199844393278e-05,
+      "loss": 1.3007,
+      "step": 7037
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761587422888698e-05,
+      "loss": 1.163,
+      "step": 7038
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876117633812938e-05,
+      "loss": 1.2016,
+      "step": 7039
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876076518965781e-05,
+      "loss": 0.6884,
+      "step": 7040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876035397747699e-05,
+      "loss": 0.8899,
+      "step": 7041
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87599427015899e-05,
+      "loss": 0.9663,
+      "step": 7042
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759531361999534e-05,
+      "loss": 1.0387,
+      "step": 7043
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875911995870888e-05,
+      "loss": 0.7485,
+      "step": 7044
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758708491720938e-05,
+      "loss": 1.4651,
+      "step": 7045
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758296961038695e-05,
+      "loss": 1.1996,
+      "step": 7046
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757885366665138e-05,
+      "loss": 1.0161,
+      "step": 7047
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875747370860327e-05,
+      "loss": 0.7132,
+      "step": 7048
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875706198685608e-05,
+      "loss": 1.2163,
+      "step": 7049
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756650201426565e-05,
+      "loss": 1.014,
+      "step": 7050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756238352317716e-05,
+      "loss": 0.9615,
+      "step": 7051
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875582643953253e-05,
+      "loss": 1.1517,
+      "step": 7052
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755414463073998e-05,
+      "loss": 1.0497,
+      "step": 7053
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755002422945124e-05,
+      "loss": 0.9219,
+      "step": 7054
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87545903191489e-05,
+      "loss": 0.9875,
+      "step": 7055
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875417815168832e-05,
+      "loss": 0.9391,
+      "step": 7056
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753765920566385e-05,
+      "loss": 0.847,
+      "step": 7057
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753353625786095e-05,
+      "loss": 1.0259,
+      "step": 7058
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752941267350446e-05,
+      "loss": 0.9243,
+      "step": 7059
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752528845262432e-05,
+      "loss": 1.2268,
+      "step": 7060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752116359525062e-05,
+      "loss": 0.9358,
+      "step": 7061
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751703810141325e-05,
+      "loss": 1.0224,
+      "step": 7062
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751291197114226e-05,
+      "loss": 1.108,
+      "step": 7063
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750878520446767e-05,
+      "loss": 0.9277,
+      "step": 7064
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750465780141946e-05,
+      "loss": 1.6701,
+      "step": 7065
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875005297620277e-05,
+      "loss": 1.0334,
+      "step": 7066
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749640108632233e-05,
+      "loss": 1.1555,
+      "step": 7067
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749227177433345e-05,
+      "loss": 1.4085,
+      "step": 7068
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748814182609103e-05,
+      "loss": 0.8212,
+      "step": 7069
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748401124162515e-05,
+      "loss": 1.1381,
+      "step": 7070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747988002096577e-05,
+      "loss": 1.138,
+      "step": 7071
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87475748164143e-05,
+      "loss": 1.1748,
+      "step": 7072
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747161567118695e-05,
+      "loss": 1.1436,
+      "step": 7073
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874674825421275e-05,
+      "loss": 0.9964,
+      "step": 7074
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874633487769949e-05,
+      "loss": 0.6536,
+      "step": 7075
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745921437581907e-05,
+      "loss": 1.2396,
+      "step": 7076
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874550793386301e-05,
+      "loss": 1.3202,
+      "step": 7077
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874509436654581e-05,
+      "loss": 1.3778,
+      "step": 7078
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744680735633314e-05,
+      "loss": 0.8244,
+      "step": 7079
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744267041128528e-05,
+      "loss": 1.225,
+      "step": 7080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874385328303446e-05,
+      "loss": 0.9783,
+      "step": 7081
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874343946135412e-05,
+      "loss": 1.2401,
+      "step": 7082
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743025576090518e-05,
+      "loss": 1.221,
+      "step": 7083
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874261162724666e-05,
+      "loss": 1.2348,
+      "step": 7084
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8742197614825563e-05,
+      "loss": 1.4025,
+      "step": 7085
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741783538830236e-05,
+      "loss": 0.9687,
+      "step": 7086
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741369399263686e-05,
+      "loss": 0.9087,
+      "step": 7087
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740955196128926e-05,
+      "loss": 0.6445,
+      "step": 7088
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874054092942897e-05,
+      "loss": 1.1949,
+      "step": 7089
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874012659916683e-05,
+      "loss": 0.9345,
+      "step": 7090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873971220534552e-05,
+      "loss": 1.2133,
+      "step": 7091
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739297747968047e-05,
+      "loss": 0.9828,
+      "step": 7092
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738883227037435e-05,
+      "loss": 1.0733,
+      "step": 7093
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738468642556693e-05,
+      "loss": 1.1136,
+      "step": 7094
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738053994528835e-05,
+      "loss": 0.8942,
+      "step": 7095
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873763928295688e-05,
+      "loss": 1.1815,
+      "step": 7096
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873722450784384e-05,
+      "loss": 1.1161,
+      "step": 7097
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736809669192733e-05,
+      "loss": 0.7861,
+      "step": 7098
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873639476700658e-05,
+      "loss": 0.8305,
+      "step": 7099
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873597980128839e-05,
+      "loss": 1.1251,
+      "step": 7100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873556477204119e-05,
+      "loss": 0.8356,
+      "step": 7101
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735149679267986e-05,
+      "loss": 0.9034,
+      "step": 7102
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734734522971804e-05,
+      "loss": 1.0209,
+      "step": 7103
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734319303155667e-05,
+      "loss": 1.4984,
+      "step": 7104
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733904019822588e-05,
+      "loss": 1.2408,
+      "step": 7105
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873348867297559e-05,
+      "loss": 1.1855,
+      "step": 7106
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873307326261769e-05,
+      "loss": 1.103,
+      "step": 7107
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732657788751917e-05,
+      "loss": 1.0721,
+      "step": 7108
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873224225138128e-05,
+      "loss": 1.094,
+      "step": 7109
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731826650508812e-05,
+      "loss": 1.1721,
+      "step": 7110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873141098613753e-05,
+      "loss": 0.8845,
+      "step": 7111
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730995258270456e-05,
+      "loss": 1.1533,
+      "step": 7112
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730579466910618e-05,
+      "loss": 0.6647,
+      "step": 7113
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730163612061035e-05,
+      "loss": 1.3532,
+      "step": 7114
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872974769372473e-05,
+      "loss": 0.8807,
+      "step": 7115
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729331711904736e-05,
+      "loss": 0.9436,
+      "step": 7116
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728915666604068e-05,
+      "loss": 1.3898,
+      "step": 7117
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728499557825757e-05,
+      "loss": 0.9703,
+      "step": 7118
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728083385572828e-05,
+      "loss": 0.9936,
+      "step": 7119
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872766714984831e-05,
+      "loss": 0.7127,
+      "step": 7120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727250850655222e-05,
+      "loss": 1.3171,
+      "step": 7121
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87268344879966e-05,
+      "loss": 0.9382,
+      "step": 7122
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726418061875472e-05,
+      "loss": 0.766,
+      "step": 7123
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872600157229486e-05,
+      "loss": 1.0673,
+      "step": 7124
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725585019257794e-05,
+      "loss": 1.1358,
+      "step": 7125
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725168402767305e-05,
+      "loss": 1.025,
+      "step": 7126
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872475172282642e-05,
+      "loss": 0.7592,
+      "step": 7127
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724334979438176e-05,
+      "loss": 1.1622,
+      "step": 7128
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723918172605597e-05,
+      "loss": 1.1587,
+      "step": 7129
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723501302331718e-05,
+      "loss": 1.2266,
+      "step": 7130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723084368619567e-05,
+      "loss": 1.1645,
+      "step": 7131
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872266737147218e-05,
+      "loss": 1.1242,
+      "step": 7132
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722250310892586e-05,
+      "loss": 1.1628,
+      "step": 7133
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721833186883818e-05,
+      "loss": 1.1311,
+      "step": 7134
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721415999448914e-05,
+      "loss": 1.1196,
+      "step": 7135
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87209987485909e-05,
+      "loss": 1.2032,
+      "step": 7136
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720581434312816e-05,
+      "loss": 1.2331,
+      "step": 7137
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720164056617697e-05,
+      "loss": 1.1111,
+      "step": 7138
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719746615508575e-05,
+      "loss": 1.2838,
+      "step": 7139
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719329110988487e-05,
+      "loss": 1.2543,
+      "step": 7140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871891154306047e-05,
+      "loss": 0.9068,
+      "step": 7141
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718493911727563e-05,
+      "loss": 1.1248,
+      "step": 7142
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718076216992796e-05,
+      "loss": 1.0862,
+      "step": 7143
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717658458859215e-05,
+      "loss": 1.0686,
+      "step": 7144
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717240637329847e-05,
+      "loss": 1.3672,
+      "step": 7145
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716822752407743e-05,
+      "loss": 1.1,
+      "step": 7146
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716404804095933e-05,
+      "loss": 1.1687,
+      "step": 7147
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871598679239746e-05,
+      "loss": 1.3246,
+      "step": 7148
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715568717315364e-05,
+      "loss": 0.9711,
+      "step": 7149
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715150578852684e-05,
+      "loss": 1.0679,
+      "step": 7150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871473237701246e-05,
+      "loss": 0.9172,
+      "step": 7151
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714314111797734e-05,
+      "loss": 0.9275,
+      "step": 7152
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713895783211553e-05,
+      "loss": 0.9095,
+      "step": 7153
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713477391256946e-05,
+      "loss": 1.0317,
+      "step": 7154
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871305893593697e-05,
+      "loss": 1.2454,
+      "step": 7155
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871264041725466e-05,
+      "loss": 0.9067,
+      "step": 7156
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871222183521306e-05,
+      "loss": 1.0205,
+      "step": 7157
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711803189815222e-05,
+      "loss": 1.2657,
+      "step": 7158
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711384481064178e-05,
+      "loss": 1.056,
+      "step": 7159
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710965708962977e-05,
+      "loss": 1.172,
+      "step": 7160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871054687351467e-05,
+      "loss": 1.1344,
+      "step": 7161
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710127974722296e-05,
+      "loss": 1.1969,
+      "step": 7162
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709709012588905e-05,
+      "loss": 1.0268,
+      "step": 7163
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709289987117544e-05,
+      "loss": 0.7367,
+      "step": 7164
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870887089831126e-05,
+      "loss": 0.9566,
+      "step": 7165
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708451746173097e-05,
+      "loss": 1.2933,
+      "step": 7166
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708032530706106e-05,
+      "loss": 1.0652,
+      "step": 7167
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707613251913337e-05,
+      "loss": 1.0527,
+      "step": 7168
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707193909797835e-05,
+      "loss": 1.2993,
+      "step": 7169
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706774504362655e-05,
+      "loss": 0.8075,
+      "step": 7170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706355035610845e-05,
+      "loss": 1.0055,
+      "step": 7171
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870593550354545e-05,
+      "loss": 1.0772,
+      "step": 7172
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705515908169526e-05,
+      "loss": 0.9454,
+      "step": 7173
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870509624948613e-05,
+      "loss": 1.1653,
+      "step": 7174
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87046765274983e-05,
+      "loss": 1.1458,
+      "step": 7175
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704256742209096e-05,
+      "loss": 1.0574,
+      "step": 7176
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703836893621573e-05,
+      "loss": 0.9081,
+      "step": 7177
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703416981738784e-05,
+      "loss": 0.9567,
+      "step": 7178
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702997006563776e-05,
+      "loss": 0.8788,
+      "step": 7179
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702576968099607e-05,
+      "loss": 1.095,
+      "step": 7180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702156866349335e-05,
+      "loss": 1.0254,
+      "step": 7181
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870173670131601e-05,
+      "loss": 1.1182,
+      "step": 7182
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870131647300269e-05,
+      "loss": 0.695,
+      "step": 7183
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870089618141243e-05,
+      "loss": 1.0421,
+      "step": 7184
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700475826548285e-05,
+      "loss": 0.9592,
+      "step": 7185
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700055408413318e-05,
+      "loss": 0.9231,
+      "step": 7186
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699634927010577e-05,
+      "loss": 1.0603,
+      "step": 7187
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699214382343127e-05,
+      "loss": 0.865,
+      "step": 7188
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698793774414026e-05,
+      "loss": 1.1705,
+      "step": 7189
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869837310322633e-05,
+      "loss": 1.3589,
+      "step": 7190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697952368783093e-05,
+      "loss": 1.0499,
+      "step": 7191
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697531571087387e-05,
+      "loss": 1.1295,
+      "step": 7192
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697110710142267e-05,
+      "loss": 0.8352,
+      "step": 7193
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869668978595079e-05,
+      "loss": 0.965,
+      "step": 7194
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696268798516016e-05,
+      "loss": 1.2861,
+      "step": 7195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695847747841012e-05,
+      "loss": 1.2117,
+      "step": 7196
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695426633928834e-05,
+      "loss": 1.1537,
+      "step": 7197
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695005456782554e-05,
+      "loss": 1.3636,
+      "step": 7198
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694584216405224e-05,
+      "loss": 1.3109,
+      "step": 7199
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694162912799917e-05,
+      "loss": 1.4072,
+      "step": 7200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693741545969688e-05,
+      "loss": 1.5191,
+      "step": 7201
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693320115917605e-05,
+      "loss": 1.4157,
+      "step": 7202
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692898622646733e-05,
+      "loss": 0.9628,
+      "step": 7203
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869247706616014e-05,
+      "loss": 1.0917,
+      "step": 7204
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692055446460886e-05,
+      "loss": 1.3386,
+      "step": 7205
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869163376355204e-05,
+      "loss": 0.7167,
+      "step": 7206
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8691212017436666e-05,
+      "loss": 1.0482,
+      "step": 7207
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869079020811783e-05,
+      "loss": 0.8087,
+      "step": 7208
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869036833559861e-05,
+      "loss": 1.0762,
+      "step": 7209
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689946399882064e-05,
+      "loss": 1.0071,
+      "step": 7210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689524400971262e-05,
+      "loss": 0.952,
+      "step": 7211
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689102338869275e-05,
+      "loss": 0.8085,
+      "step": 7212
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868868021357917e-05,
+      "loss": 1.1401,
+      "step": 7213
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8688258025104017e-05,
+      "loss": 1.1482,
+      "step": 7214
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868783577344689e-05,
+      "loss": 1.0043,
+      "step": 7215
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687413458610853e-05,
+      "loss": 1.0682,
+      "step": 7216
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868699108059898e-05,
+      "loss": 1.0917,
+      "step": 7217
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686568639414344e-05,
+      "loss": 1.0822,
+      "step": 7218
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686146135060016e-05,
+      "loss": 1.0277,
+      "step": 7219
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868572356753907e-05,
+      "loss": 0.7921,
+      "step": 7220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685300936854576e-05,
+      "loss": 1.3907,
+      "step": 7221
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684878243009613e-05,
+      "loss": 0.8751,
+      "step": 7222
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684455486007248e-05,
+      "loss": 0.8164,
+      "step": 7223
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684032665850557e-05,
+      "loss": 1.2881,
+      "step": 7224
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683609782542616e-05,
+      "loss": 1.1224,
+      "step": 7225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683186836086506e-05,
+      "loss": 1.3963,
+      "step": 7226
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682763826485292e-05,
+      "loss": 0.9593,
+      "step": 7227
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682340753742058e-05,
+      "loss": 0.8452,
+      "step": 7228
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681917617859877e-05,
+      "loss": 0.8528,
+      "step": 7229
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681494418841825e-05,
+      "loss": 1.3608,
+      "step": 7230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681071156690982e-05,
+      "loss": 1.2321,
+      "step": 7231
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680647831410428e-05,
+      "loss": 0.9731,
+      "step": 7232
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680224443003236e-05,
+      "loss": 1.1686,
+      "step": 7233
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867980099147249e-05,
+      "loss": 1.1024,
+      "step": 7234
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679377476821266e-05,
+      "loss": 0.8637,
+      "step": 7235
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867895389905265e-05,
+      "loss": 1.0568,
+      "step": 7236
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678530258169713e-05,
+      "loss": 0.9278,
+      "step": 7237
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678106554175537e-05,
+      "loss": 0.8606,
+      "step": 7238
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677682787073213e-05,
+      "loss": 0.9197,
+      "step": 7239
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677258956865812e-05,
+      "loss": 0.9803,
+      "step": 7240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867683506355642e-05,
+      "loss": 1.0744,
+      "step": 7241
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676411107148123e-05,
+      "loss": 1.1084,
+      "step": 7242
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675987087644e-05,
+      "loss": 1.3783,
+      "step": 7243
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675563005047134e-05,
+      "loss": 0.8122,
+      "step": 7244
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867513885936061e-05,
+      "loss": 0.9972,
+      "step": 7245
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674714650587512e-05,
+      "loss": 1.1993,
+      "step": 7246
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867429037873093e-05,
+      "loss": 1.0108,
+      "step": 7247
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673866043793946e-05,
+      "loss": 1.1185,
+      "step": 7248
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867344164577964e-05,
+      "loss": 1.0173,
+      "step": 7249
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673017184691105e-05,
+      "loss": 1.076,
+      "step": 7250
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672592660531427e-05,
+      "loss": 1.4049,
+      "step": 7251
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672168073303688e-05,
+      "loss": 1.0012,
+      "step": 7252
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671743423010985e-05,
+      "loss": 0.9811,
+      "step": 7253
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671318709656398e-05,
+      "loss": 1.2633,
+      "step": 7254
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670893933243017e-05,
+      "loss": 0.9694,
+      "step": 7255
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867046909377394e-05,
+      "loss": 1.0384,
+      "step": 7256
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867004419125224e-05,
+      "loss": 0.9522,
+      "step": 7257
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866961922568102e-05,
+      "loss": 0.9312,
+      "step": 7258
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866919419706336e-05,
+      "loss": 1.0933,
+      "step": 7259
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668769105402366e-05,
+      "loss": 1.1008,
+      "step": 7260
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668343950701116e-05,
+      "loss": 1.1537,
+      "step": 7261
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866791873296271e-05,
+      "loss": 0.933,
+      "step": 7262
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667493452190234e-05,
+      "loss": 1.2618,
+      "step": 7263
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667068108386782e-05,
+      "loss": 0.8734,
+      "step": 7264
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866664270155545e-05,
+      "loss": 1.0323,
+      "step": 7265
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666217231699328e-05,
+      "loss": 1.0745,
+      "step": 7266
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665791698821513e-05,
+      "loss": 1.1209,
+      "step": 7267
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665366102925097e-05,
+      "loss": 1.3799,
+      "step": 7268
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664940444013178e-05,
+      "loss": 1.0751,
+      "step": 7269
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866451472208885e-05,
+      "loss": 1.1216,
+      "step": 7270
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664088937155207e-05,
+      "loss": 1.0046,
+      "step": 7271
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866366308921535e-05,
+      "loss": 1.0567,
+      "step": 7272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663237178272374e-05,
+      "loss": 1.2826,
+      "step": 7273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866281120432937e-05,
+      "loss": 0.9143,
+      "step": 7274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662385167389443e-05,
+      "loss": 1.0103,
+      "step": 7275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866195906745569e-05,
+      "loss": 0.787,
+      "step": 7276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866153290453121e-05,
+      "loss": 1.1963,
+      "step": 7277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661106678619097e-05,
+      "loss": 1.321,
+      "step": 7278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866068038972246e-05,
+      "loss": 1.4597,
+      "step": 7279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866025403784439e-05,
+      "loss": 1.2628,
+      "step": 7280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865982762298799e-05,
+      "loss": 1.0416,
+      "step": 7281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659401145156364e-05,
+      "loss": 1.0921,
+      "step": 7282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658974604352608e-05,
+      "loss": 0.9769,
+      "step": 7283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865854800057983e-05,
+      "loss": 0.904,
+      "step": 7284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658121333841127e-05,
+      "loss": 1.2063,
+      "step": 7285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865769460413961e-05,
+      "loss": 1.2216,
+      "step": 7286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865726781147837e-05,
+      "loss": 0.7954,
+      "step": 7287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865684095586052e-05,
+      "loss": 0.9589,
+      "step": 7288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865641403728916e-05,
+      "loss": 1.1492,
+      "step": 7289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655987055767396e-05,
+      "loss": 0.9011,
+      "step": 7290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655560011298334e-05,
+      "loss": 0.7827,
+      "step": 7291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865513290388508e-05,
+      "loss": 0.9755,
+      "step": 7292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654705733530735e-05,
+      "loss": 0.9814,
+      "step": 7293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654278500238412e-05,
+      "loss": 1.1775,
+      "step": 7294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865385120401121e-05,
+      "loss": 0.9206,
+      "step": 7295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653423844852247e-05,
+      "loss": 1.042,
+      "step": 7296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865299642276462e-05,
+      "loss": 1.0166,
+      "step": 7297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652568937751444e-05,
+      "loss": 1.112,
+      "step": 7298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652141389815827e-05,
+      "loss": 0.8591,
+      "step": 7299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8651713778960875e-05,
+      "loss": 0.9509,
+      "step": 7300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8651286105189702e-05,
+      "loss": 0.8631,
+      "step": 7301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650858368505414e-05,
+      "loss": 1.0157,
+      "step": 7302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650430568911125e-05,
+      "loss": 0.9937,
+      "step": 7303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650002706409938e-05,
+      "loss": 0.9311,
+      "step": 7304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864957478100498e-05,
+      "loss": 1.6071,
+      "step": 7305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649146792699348e-05,
+      "loss": 1.2154,
+      "step": 7306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864871874149616e-05,
+      "loss": 1.0963,
+      "step": 7307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648290627398532e-05,
+      "loss": 1.139,
+      "step": 7308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864786245040957e-05,
+      "loss": 1.1262,
+      "step": 7309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647434210532397e-05,
+      "loss": 1.1285,
+      "step": 7310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864700590777012e-05,
+      "loss": 1.3895,
+      "step": 7311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646577542125856e-05,
+      "loss": 1.4036,
+      "step": 7312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864614911360272e-05,
+      "loss": 1.1957,
+      "step": 7313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645720622203825e-05,
+      "loss": 0.7131,
+      "step": 7314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864529206793229e-05,
+      "loss": 0.9621,
+      "step": 7315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644863450791233e-05,
+      "loss": 1.044,
+      "step": 7316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644434770783767e-05,
+      "loss": 1.0511,
+      "step": 7317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864400602791301e-05,
+      "loss": 1.0603,
+      "step": 7318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8643577222182088e-05,
+      "loss": 1.1839,
+      "step": 7319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864314835359411e-05,
+      "loss": 0.8204,
+      "step": 7320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642719422152194e-05,
+      "loss": 1.2308,
+      "step": 7321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642290427859464e-05,
+      "loss": 0.8336,
+      "step": 7322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641861370719038e-05,
+      "loss": 1.2172,
+      "step": 7323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864143225073404e-05,
+      "loss": 1.0623,
+      "step": 7324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641003067907585e-05,
+      "loss": 1.1246,
+      "step": 7325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640573822242792e-05,
+      "loss": 1.221,
+      "step": 7326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640144513742794e-05,
+      "loss": 1.15,
+      "step": 7327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639715142410702e-05,
+      "loss": 0.8878,
+      "step": 7328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863928570824964e-05,
+      "loss": 0.6089,
+      "step": 7329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638856211262734e-05,
+      "loss": 0.971,
+      "step": 7330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863842665145311e-05,
+      "loss": 1.3282,
+      "step": 7331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637997028823883e-05,
+      "loss": 1.1603,
+      "step": 7332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637567343378187e-05,
+      "loss": 1.0787,
+      "step": 7333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637137595119138e-05,
+      "loss": 1.218,
+      "step": 7334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636707784049867e-05,
+      "loss": 1.1492,
+      "step": 7335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636277910173497e-05,
+      "loss": 1.0264,
+      "step": 7336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635847973493158e-05,
+      "loss": 1.0068,
+      "step": 7337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635417974011972e-05,
+      "loss": 0.9378,
+      "step": 7338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634987911733066e-05,
+      "loss": 0.896,
+      "step": 7339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863455778665957e-05,
+      "loss": 1.2034,
+      "step": 7340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863412759879461e-05,
+      "loss": 0.9526,
+      "step": 7341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633697348141315e-05,
+      "loss": 1.4491,
+      "step": 7342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633267034702818e-05,
+      "loss": 1.2308,
+      "step": 7343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863283665848224e-05,
+      "loss": 1.0688,
+      "step": 7344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863240621948272e-05,
+      "loss": 1.0405,
+      "step": 7345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863197571770738e-05,
+      "loss": 0.8476,
+      "step": 7346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863154515315935e-05,
+      "loss": 0.9617,
+      "step": 7347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631114525841774e-05,
+      "loss": 0.8591,
+      "step": 7348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630683835757768e-05,
+      "loss": 1.1835,
+      "step": 7349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630253082910473e-05,
+      "loss": 0.9807,
+      "step": 7350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629822267303018e-05,
+      "loss": 1.0889,
+      "step": 7351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862939138893854e-05,
+      "loss": 1.1768,
+      "step": 7352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862896044782017e-05,
+      "loss": 0.7514,
+      "step": 7353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628529443951037e-05,
+      "loss": 1.1824,
+      "step": 7354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628098377334282e-05,
+      "loss": 0.7934,
+      "step": 7355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627667247973037e-05,
+      "loss": 0.6807,
+      "step": 7356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862723605587044e-05,
+      "loss": 1.2966,
+      "step": 7357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626804801029622e-05,
+      "loss": 1.0367,
+      "step": 7358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626373483453722e-05,
+      "loss": 1.0974,
+      "step": 7359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625942103145877e-05,
+      "loss": 0.8298,
+      "step": 7360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625510660109222e-05,
+      "loss": 0.8292,
+      "step": 7361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625079154346894e-05,
+      "loss": 1.1954,
+      "step": 7362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8624647585862034e-05,
+      "loss": 0.9877,
+      "step": 7363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862421595465778e-05,
+      "loss": 1.1996,
+      "step": 7364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623784260737272e-05,
+      "loss": 0.9315,
+      "step": 7365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623352504103643e-05,
+      "loss": 1.0208,
+      "step": 7366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862292068476004e-05,
+      "loss": 0.839,
+      "step": 7367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86224888027096e-05,
+      "loss": 0.937,
+      "step": 7368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622056857955463e-05,
+      "loss": 0.9274,
+      "step": 7369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621624850500768e-05,
+      "loss": 1.1504,
+      "step": 7370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862119278034866e-05,
+      "loss": 1.1643,
+      "step": 7371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620760647502285e-05,
+      "loss": 1.1153,
+      "step": 7372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862032845196478e-05,
+      "loss": 0.9826,
+      "step": 7373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861989619373929e-05,
+      "loss": 0.9267,
+      "step": 7374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861946387282896e-05,
+      "loss": 1.0117,
+      "step": 7375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8619031489236924e-05,
+      "loss": 1.1623,
+      "step": 7376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618599042966338e-05,
+      "loss": 1.0365,
+      "step": 7377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618166534020342e-05,
+      "loss": 1.256,
+      "step": 7378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617733962402084e-05,
+      "loss": 1.2346,
+      "step": 7379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617301328114704e-05,
+      "loss": 1.3846,
+      "step": 7380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616868631161356e-05,
+      "loss": 0.8889,
+      "step": 7381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861643587154518e-05,
+      "loss": 1.1474,
+      "step": 7382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616003049269326e-05,
+      "loss": 0.7729,
+      "step": 7383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861557016433694e-05,
+      "loss": 0.8164,
+      "step": 7384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8615137216751173e-05,
+      "loss": 1.0924,
+      "step": 7385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861470420651517e-05,
+      "loss": 0.8044,
+      "step": 7386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861427113363208e-05,
+      "loss": 1.139,
+      "step": 7387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613837998105057e-05,
+      "loss": 1.0535,
+      "step": 7388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613404799937246e-05,
+      "loss": 1.0667,
+      "step": 7389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86129715391318e-05,
+      "loss": 1.0012,
+      "step": 7390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612538215691867e-05,
+      "loss": 1.2494,
+      "step": 7391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612104829620598e-05,
+      "loss": 1.2985,
+      "step": 7392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861167138092115e-05,
+      "loss": 1.0084,
+      "step": 7393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861123786959667e-05,
+      "loss": 1.1208,
+      "step": 7394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861080429565031e-05,
+      "loss": 1.0021,
+      "step": 7395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861037065908523e-05,
+      "loss": 1.1471,
+      "step": 7396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609936959904575e-05,
+      "loss": 1.0271,
+      "step": 7397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86095031981115e-05,
+      "loss": 1.1501,
+      "step": 7398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609069373709164e-05,
+      "loss": 0.9304,
+      "step": 7399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860863548670072e-05,
+      "loss": 1.1354,
+      "step": 7400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8608201537089323e-05,
+      "loss": 0.9997,
+      "step": 7401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8607767524878126e-05,
+      "loss": 0.8139,
+      "step": 7402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860733345007029e-05,
+      "loss": 1.1755,
+      "step": 7403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860689931266897e-05,
+      "loss": 0.825,
+      "step": 7404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860646511267732e-05,
+      "loss": 0.7077,
+      "step": 7405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86060308500985e-05,
+      "loss": 1.0961,
+      "step": 7406
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860559652493567e-05,
+      "loss": 0.8902,
+      "step": 7407
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605162137191987e-05,
+      "loss": 1.2842,
+      "step": 7408
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860472768687061e-05,
+      "loss": 1.1312,
+      "step": 7409
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604293173974694e-05,
+      "loss": 0.978,
+      "step": 7410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8603858598507404e-05,
+      "loss": 0.9842,
+      "step": 7411
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86034239604719e-05,
+      "loss": 1.009,
+      "step": 7412
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860298925987134e-05,
+      "loss": 0.8814,
+      "step": 7413
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602554496708887e-05,
+      "loss": 1.3049,
+      "step": 7414
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602119670987705e-05,
+      "loss": 1.0217,
+      "step": 7415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860168478271095e-05,
+      "loss": 0.9659,
+      "step": 7416
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860124983188179e-05,
+      "loss": 1.2718,
+      "step": 7417
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8600814818503385e-05,
+      "loss": 0.7774,
+      "step": 7418
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8600379742578902e-05,
+      "loss": 1.2559,
+      "step": 7419
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85999446041115e-05,
+      "loss": 0.7925,
+      "step": 7420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8599509403104347e-05,
+      "loss": 1.3677,
+      "step": 7421
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8599074139560606e-05,
+      "loss": 0.9662,
+      "step": 7422
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598638813483442e-05,
+      "loss": 1.1015,
+      "step": 7423
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598203424876025e-05,
+      "loss": 0.8859,
+      "step": 7424
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597767973741514e-05,
+      "loss": 1.0147,
+      "step": 7425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859733246008308e-05,
+      "loss": 1.0423,
+      "step": 7426
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596896883903892e-05,
+      "loss": 1.3604,
+      "step": 7427
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596461245207114e-05,
+      "loss": 1.1521,
+      "step": 7428
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596025543995914e-05,
+      "loss": 1.487,
+      "step": 7429
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595589780273465e-05,
+      "loss": 0.9543,
+      "step": 7430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595153954042932e-05,
+      "loss": 0.889,
+      "step": 7431
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594718065307483e-05,
+      "loss": 0.9473,
+      "step": 7432
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859428211407029e-05,
+      "loss": 0.9063,
+      "step": 7433
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8593846100334523e-05,
+      "loss": 1.1971,
+      "step": 7434
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8593410024103354e-05,
+      "loss": 1.337,
+      "step": 7435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859297388537995e-05,
+      "loss": 1.2439,
+      "step": 7436
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8592537684167488e-05,
+      "loss": 1.0754,
+      "step": 7437
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8592101420469138e-05,
+      "loss": 1.035,
+      "step": 7438
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859166509428807e-05,
+      "loss": 0.8167,
+      "step": 7439
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591228705627464e-05,
+      "loss": 1.2154,
+      "step": 7440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590792254490483e-05,
+      "loss": 0.94,
+      "step": 7441
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590355740880313e-05,
+      "loss": 1.3797,
+      "step": 7442
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589919164800118e-05,
+      "loss": 1.0932,
+      "step": 7443
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858948252625308e-05,
+      "loss": 1.0432,
+      "step": 7444
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589045825242368e-05,
+      "loss": 0.4925,
+      "step": 7445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858860906177116e-05,
+      "loss": 0.9921,
+      "step": 7446
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858817223584264e-05,
+      "loss": 1.1971,
+      "step": 7447
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587735347459973e-05,
+      "loss": 1.1628,
+      "step": 7448
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587298396626337e-05,
+      "loss": 1.0879,
+      "step": 7449
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858686138334492e-05,
+      "loss": 0.8539,
+      "step": 7450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858642430761889e-05,
+      "loss": 0.855,
+      "step": 7451
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858598716945143e-05,
+      "loss": 0.9872,
+      "step": 7452
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858554996884572e-05,
+      "loss": 1.5166,
+      "step": 7453
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585112705804934e-05,
+      "loss": 1.5057,
+      "step": 7454
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584675380332254e-05,
+      "loss": 1.0536,
+      "step": 7455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584237992430864e-05,
+      "loss": 1.2536,
+      "step": 7456
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8583800542103942e-05,
+      "loss": 0.875,
+      "step": 7457
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858336302935467e-05,
+      "loss": 1.2031,
+      "step": 7458
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582925454186225e-05,
+      "loss": 0.9218,
+      "step": 7459
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582487816601796e-05,
+      "loss": 1.1582,
+      "step": 7460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858205011660456e-05,
+      "loss": 1.4577,
+      "step": 7461
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8581612354197707e-05,
+      "loss": 1.1234,
+      "step": 7462
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8581174529384414e-05,
+      "loss": 0.7936,
+      "step": 7463
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580736642167866e-05,
+      "loss": 1.2901,
+      "step": 7464
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858029869255125e-05,
+      "loss": 0.9835,
+      "step": 7465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579860680537748e-05,
+      "loss": 0.7817,
+      "step": 7466
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579422606130547e-05,
+      "loss": 0.9249,
+      "step": 7467
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857898446933283e-05,
+      "loss": 1.055,
+      "step": 7468
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578546270147787e-05,
+      "loss": 0.9336,
+      "step": 7469
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578108008578603e-05,
+      "loss": 0.8361,
+      "step": 7470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8577669684628463e-05,
+      "loss": 0.9931,
+      "step": 7471
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857723129830056e-05,
+      "loss": 0.9531,
+      "step": 7472
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576792849598074e-05,
+      "loss": 1.3007,
+      "step": 7473
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85763543385242e-05,
+      "loss": 1.0467,
+      "step": 7474
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575915765082127e-05,
+      "loss": 1.0227,
+      "step": 7475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857547712927504e-05,
+      "loss": 1.1488,
+      "step": 7476
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575038431106134e-05,
+      "loss": 1.0024,
+      "step": 7477
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574599670578594e-05,
+      "loss": 0.9777,
+      "step": 7478
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857416084769561e-05,
+      "loss": 1.0536,
+      "step": 7479
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857372196246038e-05,
+      "loss": 0.9456,
+      "step": 7480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857328301487609e-05,
+      "loss": 0.8296,
+      "step": 7481
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572844004945937e-05,
+      "loss": 1.1087,
+      "step": 7482
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572404932673107e-05,
+      "loss": 0.9696,
+      "step": 7483
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571965798060798e-05,
+      "loss": 0.9984,
+      "step": 7484
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571526601112202e-05,
+      "loss": 1.1211,
+      "step": 7485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857108734183051e-05,
+      "loss": 1.2177,
+      "step": 7486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8570648020218923e-05,
+      "loss": 1.0534,
+      "step": 7487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857020863628063e-05,
+      "loss": 1.032,
+      "step": 7488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569769190018826e-05,
+      "loss": 0.8861,
+      "step": 7489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569329681436713e-05,
+      "loss": 0.8053,
+      "step": 7490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856889011053748e-05,
+      "loss": 1.0173,
+      "step": 7491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856845047732433e-05,
+      "loss": 1.6772,
+      "step": 7492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8568010781800453e-05,
+      "loss": 0.8274,
+      "step": 7493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567571023969053e-05,
+      "loss": 1.0746,
+      "step": 7494
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856713120383332e-05,
+      "loss": 0.8259,
+      "step": 7495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566691321396462e-05,
+      "loss": 1.0912,
+      "step": 7496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566251376661672e-05,
+      "loss": 0.7215,
+      "step": 7497
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565811369632152e-05,
+      "loss": 1.0422,
+      "step": 7498
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85653713003111e-05,
+      "loss": 0.9757,
+      "step": 7499
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564931168701713e-05,
+      "loss": 1.17,
+      "step": 7500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85644909748072e-05,
+      "loss": 0.9524,
+      "step": 7501
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564050718630756e-05,
+      "loss": 1.1581,
+      "step": 7502
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563610400175583e-05,
+      "loss": 0.749,
+      "step": 7503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563170019444888e-05,
+      "loss": 1.2462,
+      "step": 7504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856272957644187e-05,
+      "loss": 1.091,
+      "step": 7505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562289071169724e-05,
+      "loss": 1.1987,
+      "step": 7506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856184850363167e-05,
+      "loss": 0.9793,
+      "step": 7507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85614078738309e-05,
+      "loss": 1.0925,
+      "step": 7508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8560967181770618e-05,
+      "loss": 0.9295,
+      "step": 7509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8560526427454032e-05,
+      "loss": 1.0496,
+      "step": 7510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856008561088435e-05,
+      "loss": 1.229,
+      "step": 7511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559644732064774e-05,
+      "loss": 1.1162,
+      "step": 7512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855920379099851e-05,
+      "loss": 1.0899,
+      "step": 7513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558762787688767e-05,
+      "loss": 1.0911,
+      "step": 7514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855832172213875e-05,
+      "loss": 0.9525,
+      "step": 7515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557880594351668e-05,
+      "loss": 1.5383,
+      "step": 7516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855743940433073e-05,
+      "loss": 0.9887,
+      "step": 7517
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556998152079138e-05,
+      "loss": 1.2551,
+      "step": 7518
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556556837600105e-05,
+      "loss": 1.0747,
+      "step": 7519
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556115460896845e-05,
+      "loss": 1.0643,
+      "step": 7520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855567402197256e-05,
+      "loss": 0.9818,
+      "step": 7521
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555232520830465e-05,
+      "loss": 0.9588,
+      "step": 7522
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855479095747377e-05,
+      "loss": 0.6937,
+      "step": 7523
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855434933190568e-05,
+      "loss": 1.114,
+      "step": 7524
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553907644129418e-05,
+      "loss": 0.9318,
+      "step": 7525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553465894148185e-05,
+      "loss": 0.8151,
+      "step": 7526
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553024081965202e-05,
+      "loss": 1.291,
+      "step": 7527
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552582207583672e-05,
+      "loss": 1.0407,
+      "step": 7528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855214027100682e-05,
+      "loss": 1.17,
+      "step": 7529
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855169827223785e-05,
+      "loss": 0.727,
+      "step": 7530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551256211279988e-05,
+      "loss": 1.0609,
+      "step": 7531
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550814088136435e-05,
+      "loss": 1.0451,
+      "step": 7532
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550371902810414e-05,
+      "loss": 0.7401,
+      "step": 7533
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8549929655305138e-05,
+      "loss": 0.8586,
+      "step": 7534
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8549487345623826e-05,
+      "loss": 0.9337,
+      "step": 7535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854904497376969e-05,
+      "loss": 1.2314,
+      "step": 7536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854860253974595e-05,
+      "loss": 1.1013,
+      "step": 7537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548160043555827e-05,
+      "loss": 0.8656,
+      "step": 7538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854771748520253e-05,
+      "loss": 0.9631,
+      "step": 7539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547274864689288e-05,
+      "loss": 0.9729,
+      "step": 7540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854683218201931e-05,
+      "loss": 1.1815,
+      "step": 7541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546389437195818e-05,
+      "loss": 0.8552,
+      "step": 7542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545946630222038e-05,
+      "loss": 0.9031,
+      "step": 7543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854550376110118e-05,
+      "loss": 1.1217,
+      "step": 7544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854506082983647e-05,
+      "loss": 0.9179,
+      "step": 7545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854461783643113e-05,
+      "loss": 1.1016,
+      "step": 7546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854417478088838e-05,
+      "loss": 0.9312,
+      "step": 7547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543731663211444e-05,
+      "loss": 1.3022,
+      "step": 7548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543288483403542e-05,
+      "loss": 0.9761,
+      "step": 7549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542845241467895e-05,
+      "loss": 0.8306,
+      "step": 7550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542401937407734e-05,
+      "loss": 1.1133,
+      "step": 7551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854195857122627e-05,
+      "loss": 1.099,
+      "step": 7552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854151514292674e-05,
+      "loss": 1.1367,
+      "step": 7553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854107165251236e-05,
+      "loss": 0.9363,
+      "step": 7554
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854062809998636e-05,
+      "loss": 1.0808,
+      "step": 7555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540184485351968e-05,
+      "loss": 0.911,
+      "step": 7556
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85397408086124e-05,
+      "loss": 1.0939,
+      "step": 7557
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539297069770892e-05,
+      "loss": 1.2504,
+      "step": 7558
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8538853268830664e-05,
+      "loss": 1.0848,
+      "step": 7559
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8538409405794952e-05,
+      "loss": 0.8154,
+      "step": 7560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537965480666974e-05,
+      "loss": 1.1639,
+      "step": 7561
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537521493449968e-05,
+      "loss": 1.2738,
+      "step": 7562
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537077444147153e-05,
+      "loss": 1.0728,
+      "step": 7563
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536633332761767e-05,
+      "loss": 1.2091,
+      "step": 7564
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853618915929703e-05,
+      "loss": 0.7669,
+      "step": 7565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535744923756182e-05,
+      "loss": 1.1372,
+      "step": 7566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535300626142447e-05,
+      "loss": 1.0959,
+      "step": 7567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8534856266459056e-05,
+      "loss": 0.8688,
+      "step": 7568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8534411844709245e-05,
+      "loss": 0.7777,
+      "step": 7569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533967360896245e-05,
+      "loss": 1.3808,
+      "step": 7570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533522815023282e-05,
+      "loss": 1.1463,
+      "step": 7571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533078207093596e-05,
+      "loss": 0.8854,
+      "step": 7572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8532633537110422e-05,
+      "loss": 1.5071,
+      "step": 7573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8532188805076985e-05,
+      "loss": 1.2515,
+      "step": 7574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531744010996525e-05,
+      "loss": 0.6199,
+      "step": 7575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853129915487227e-05,
+      "loss": 1.1456,
+      "step": 7576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8530854236707467e-05,
+      "loss": 1.0191,
+      "step": 7577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853040925650534e-05,
+      "loss": 1.2323,
+      "step": 7578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529964214269132e-05,
+      "loss": 0.7511,
+      "step": 7579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529519110002076e-05,
+      "loss": 1.0504,
+      "step": 7580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852907394370741e-05,
+      "loss": 1.2755,
+      "step": 7581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852862871538837e-05,
+      "loss": 1.2221,
+      "step": 7582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528183425048197e-05,
+      "loss": 0.9975,
+      "step": 7583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8527738072690127e-05,
+      "loss": 0.8971,
+      "step": 7584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85272926583174e-05,
+      "loss": 0.9227,
+      "step": 7585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852684718193325e-05,
+      "loss": 0.9292,
+      "step": 7586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8526401643540924e-05,
+      "loss": 0.9773,
+      "step": 7587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852595604314366e-05,
+      "loss": 1.0023,
+      "step": 7588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525510380744693e-05,
+      "loss": 1.0296,
+      "step": 7589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525064656347265e-05,
+      "loss": 1.1476,
+      "step": 7590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524618869954626e-05,
+      "loss": 1.0564,
+      "step": 7591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852417302157001e-05,
+      "loss": 1.1247,
+      "step": 7592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852372711119666e-05,
+      "loss": 0.9329,
+      "step": 7593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852328113883782e-05,
+      "loss": 0.9788,
+      "step": 7594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852283510449673e-05,
+      "loss": 1.0695,
+      "step": 7595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522389008176643e-05,
+      "loss": 1.1381,
+      "step": 7596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521942849880795e-05,
+      "loss": 1.1152,
+      "step": 7597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521496629612432e-05,
+      "loss": 1.3692,
+      "step": 7598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85210503473748e-05,
+      "loss": 0.8437,
+      "step": 7599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852060400317114e-05,
+      "loss": 1.2995,
+      "step": 7600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520157597004703e-05,
+      "loss": 0.9827,
+      "step": 7601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519711128878738e-05,
+      "loss": 1.0787,
+      "step": 7602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519264598796483e-05,
+      "loss": 0.9564,
+      "step": 7603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851881800676119e-05,
+      "loss": 0.9168,
+      "step": 7604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518371352776107e-05,
+      "loss": 1.0357,
+      "step": 7605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517924636844483e-05,
+      "loss": 0.7441,
+      "step": 7606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517477858969563e-05,
+      "loss": 1.2743,
+      "step": 7607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85170310191546e-05,
+      "loss": 1.311,
+      "step": 7608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516584117402844e-05,
+      "loss": 1.0185,
+      "step": 7609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851613715371754e-05,
+      "loss": 0.9913,
+      "step": 7610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515690128101938e-05,
+      "loss": 1.0557,
+      "step": 7611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515243040559295e-05,
+      "loss": 0.8261,
+      "step": 7612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514795891092857e-05,
+      "loss": 1.4321,
+      "step": 7613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851434867970588e-05,
+      "loss": 1.0253,
+      "step": 7614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8513901406401612e-05,
+      "loss": 1.0304,
+      "step": 7615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8513454071183306e-05,
+      "loss": 0.9923,
+      "step": 7616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851300667405422e-05,
+      "loss": 1.1412,
+      "step": 7617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512559215017603e-05,
+      "loss": 0.7933,
+      "step": 7618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512111694076712e-05,
+      "loss": 1.3874,
+      "step": 7619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85116641112348e-05,
+      "loss": 1.0654,
+      "step": 7620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511216466495115e-05,
+      "loss": 0.9606,
+      "step": 7621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8510768759860926e-05,
+      "loss": 1.3757,
+      "step": 7622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851032099133548e-05,
+      "loss": 1.2038,
+      "step": 7623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850987316092203e-05,
+      "loss": 1.0084,
+      "step": 7624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850942526862384e-05,
+      "loss": 1.089,
+      "step": 7625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508977314444165e-05,
+      "loss": 1.0902,
+      "step": 7626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850852929838626e-05,
+      "loss": 1.0531,
+      "step": 7627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508081220453387e-05,
+      "loss": 0.84,
+      "step": 7628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507633080648805e-05,
+      "loss": 1.2158,
+      "step": 7629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507184878975767e-05,
+      "loss": 0.8787,
+      "step": 7630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8506736615437536e-05,
+      "loss": 1.015,
+      "step": 7631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8506288290037373e-05,
+      "loss": 0.9402,
+      "step": 7632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850583990277854e-05,
+      "loss": 0.8005,
+      "step": 7633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505391453664287e-05,
+      "loss": 0.8734,
+      "step": 7634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850494294269789e-05,
+      "loss": 1.151,
+      "step": 7635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504494369882597e-05,
+      "loss": 1.0653,
+      "step": 7636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504045735221682e-05,
+      "loss": 0.8442,
+      "step": 7637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503597038718403e-05,
+      "loss": 1.0008,
+      "step": 7638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503148280376017e-05,
+      "loss": 0.7637,
+      "step": 7639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8502699460197793e-05,
+      "loss": 1.056,
+      "step": 7640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8502250578186998e-05,
+      "loss": 1.1051,
+      "step": 7641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850180163434689e-05,
+      "loss": 0.9776,
+      "step": 7642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850135262868074e-05,
+      "loss": 0.855,
+      "step": 7643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500903561191802e-05,
+      "loss": 1.0688,
+      "step": 7644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500454431883358e-05,
+      "loss": 0.864,
+      "step": 7645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850000524075866e-05,
+      "loss": 1.0701,
+      "step": 7646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8499555987820982e-05,
+      "loss": 1.1015,
+      "step": 7647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8499106673073586e-05,
+      "loss": 0.9492,
+      "step": 7648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498657296519745e-05,
+      "loss": 1.2551,
+      "step": 7649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498207858162724e-05,
+      "loss": 0.841,
+      "step": 7650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497758358005793e-05,
+      "loss": 1.4303,
+      "step": 7651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497308796052218e-05,
+      "loss": 1.1168,
+      "step": 7652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8496859172305272e-05,
+      "loss": 1.0591,
+      "step": 7653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8496409486768218e-05,
+      "loss": 0.8341,
+      "step": 7654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8495959739444335e-05,
+      "loss": 1.2766,
+      "step": 7655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8495509930336887e-05,
+      "loss": 1.3405,
+      "step": 7656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849506005944915e-05,
+      "loss": 0.9322,
+      "step": 7657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849461012678439e-05,
+      "loss": 0.9129,
+      "step": 7658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8494160132345883e-05,
+      "loss": 0.9632,
+      "step": 7659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493710076136898e-05,
+      "loss": 0.9724,
+      "step": 7660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493259958160715e-05,
+      "loss": 1.015,
+      "step": 7661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8492809778420603e-05,
+      "loss": 0.9449,
+      "step": 7662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849235953691983e-05,
+      "loss": 0.94,
+      "step": 7663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491909233661678e-05,
+      "loss": 0.868,
+      "step": 7664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491458868649417e-05,
+      "loss": 0.9442,
+      "step": 7665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491008441886326e-05,
+      "loss": 0.9048,
+      "step": 7666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849055795337568e-05,
+      "loss": 0.9482,
+      "step": 7667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849010740312075e-05,
+      "loss": 1.28,
+      "step": 7668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848965679112482e-05,
+      "loss": 1.159,
+      "step": 7669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8489206117391162e-05,
+      "loss": 0.9267,
+      "step": 7670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848875538192305e-05,
+      "loss": 1.0227,
+      "step": 7671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488304584723772e-05,
+      "loss": 1.0665,
+      "step": 7672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8487853725796597e-05,
+      "loss": 1.028,
+      "step": 7673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848740280514481e-05,
+      "loss": 1.478,
+      "step": 7674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8486951822771684e-05,
+      "loss": 1.1571,
+      "step": 7675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8486500778680505e-05,
+      "loss": 1.1194,
+      "step": 7676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848604967287455e-05,
+      "loss": 1.4933,
+      "step": 7677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8485598505357098e-05,
+      "loss": 0.8664,
+      "step": 7678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848514727613143e-05,
+      "loss": 1.0458,
+      "step": 7679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484695985200832e-05,
+      "loss": 1.0992,
+      "step": 7680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848424463256858e-05,
+      "loss": 0.8124,
+      "step": 7681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848379321823796e-05,
+      "loss": 0.9421,
+      "step": 7682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848334174221225e-05,
+      "loss": 1.2392,
+      "step": 7683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848289020449474e-05,
+      "loss": 0.9036,
+      "step": 7684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8482438605088714e-05,
+      "loss": 1.2179,
+      "step": 7685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848198694399745e-05,
+      "loss": 0.8815,
+      "step": 7686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848153522122423e-05,
+      "loss": 0.6358,
+      "step": 7687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848108343677235e-05,
+      "loss": 1.134,
+      "step": 7688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8480631590645085e-05,
+      "loss": 1.2894,
+      "step": 7689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8480179682845733e-05,
+      "loss": 1.1355,
+      "step": 7690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479727713377564e-05,
+      "loss": 0.9363,
+      "step": 7691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847927568224388e-05,
+      "loss": 0.8744,
+      "step": 7692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8478823589447958e-05,
+      "loss": 1.2669,
+      "step": 7693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847837143499309e-05,
+      "loss": 0.875,
+      "step": 7694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847791921888256e-05,
+      "loss": 0.72,
+      "step": 7695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8477466941119667e-05,
+      "loss": 1.0781,
+      "step": 7696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8477014601707687e-05,
+      "loss": 0.883,
+      "step": 7697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847656220064992e-05,
+      "loss": 1.0942,
+      "step": 7698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8476109737949648e-05,
+      "loss": 0.8902,
+      "step": 7699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475657213610165e-05,
+      "loss": 0.7908,
+      "step": 7700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475204627634764e-05,
+      "loss": 0.8954,
+      "step": 7701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847475198002673e-05,
+      "loss": 1.2331,
+      "step": 7702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847429927078936e-05,
+      "loss": 1.044,
+      "step": 7703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473846499925945e-05,
+      "loss": 1.1278,
+      "step": 7704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847339366743978e-05,
+      "loss": 0.9227,
+      "step": 7705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8472940773334152e-05,
+      "loss": 1.316,
+      "step": 7706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847248781761236e-05,
+      "loss": 1.1315,
+      "step": 7707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8472034800277692e-05,
+      "loss": 0.765,
+      "step": 7708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847158172133345e-05,
+      "loss": 1.315,
+      "step": 7709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471128580782923e-05,
+      "loss": 0.932,
+      "step": 7710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847067537862941e-05,
+      "loss": 0.8394,
+      "step": 7711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8470222114876202e-05,
+      "loss": 1.1464,
+      "step": 7712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84697687895266e-05,
+      "loss": 1.1181,
+      "step": 7713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469315402583898e-05,
+      "loss": 0.9442,
+      "step": 7714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8468861954051398e-05,
+      "loss": 1.0053,
+      "step": 7715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846840844393239e-05,
+      "loss": 0.9376,
+      "step": 7716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8467954872230173e-05,
+      "loss": 1.0042,
+      "step": 7717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846750123894805e-05,
+      "loss": 0.9141,
+      "step": 7718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846704754408932e-05,
+      "loss": 0.9678,
+      "step": 7719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466593787657277e-05,
+      "loss": 0.9417,
+      "step": 7720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466139969655226e-05,
+      "loss": 1.6513,
+      "step": 7721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8465686090086464e-05,
+      "loss": 0.9407,
+      "step": 7722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846523214895429e-05,
+      "loss": 1.2947,
+      "step": 7723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464778146262013e-05,
+      "loss": 1.1499,
+      "step": 7724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464324082012926e-05,
+      "loss": 0.8589,
+      "step": 7725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463869956210337e-05,
+      "loss": 1.1483,
+      "step": 7726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463415768857543e-05,
+      "loss": 1.2087,
+      "step": 7727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846296151995785e-05,
+      "loss": 0.9217,
+      "step": 7728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846250720951456e-05,
+      "loss": 1.2872,
+      "step": 7729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846205283753098e-05,
+      "loss": 1.1494,
+      "step": 7730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846159840401041e-05,
+      "loss": 0.9793,
+      "step": 7731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846114390895616e-05,
+      "loss": 0.9059,
+      "step": 7732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846068935237153e-05,
+      "loss": 1.2364,
+      "step": 7733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8460234734259827e-05,
+      "loss": 0.9012,
+      "step": 7734
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8459780054624356e-05,
+      "loss": 0.9941,
+      "step": 7735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845932531346843e-05,
+      "loss": 1.1772,
+      "step": 7736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8458870510795347e-05,
+      "loss": 1.163,
+      "step": 7737
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8458415646608418e-05,
+      "loss": 0.9505,
+      "step": 7738
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457960720910953e-05,
+      "loss": 0.9765,
+      "step": 7739
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845750573370626e-05,
+      "loss": 1.137,
+      "step": 7740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457050684997645e-05,
+      "loss": 0.8149,
+      "step": 7741
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456595574788415e-05,
+      "loss": 0.9886,
+      "step": 7742
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456140403081886e-05,
+      "loss": 0.9512,
+      "step": 7743
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8455685169881364e-05,
+      "loss": 1.2028,
+      "step": 7744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845522987519016e-05,
+      "loss": 1.1377,
+      "step": 7745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454774519011585e-05,
+      "loss": 1.1311,
+      "step": 7746
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845431910134895e-05,
+      "loss": 0.9952,
+      "step": 7747
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453863622205573e-05,
+      "loss": 0.7461,
+      "step": 7748
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453408081584757e-05,
+      "loss": 1.414,
+      "step": 7749
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845295247948982e-05,
+      "loss": 0.9962,
+      "step": 7750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8452496815924072e-05,
+      "loss": 1.3749,
+      "step": 7751
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845204109089083e-05,
+      "loss": 1.0158,
+      "step": 7752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8451585304393406e-05,
+      "loss": 1.1059,
+      "step": 7753
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8451129456435114e-05,
+      "loss": 1.1588,
+      "step": 7754
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8450673547019273e-05,
+      "loss": 0.9979,
+      "step": 7755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8450217576149193e-05,
+      "loss": 1.1305,
+      "step": 7756
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449761543828194e-05,
+      "loss": 0.8178,
+      "step": 7757
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844930545005959e-05,
+      "loss": 0.9839,
+      "step": 7758
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8448849294846696e-05,
+      "loss": 1.4533,
+      "step": 7759
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8448393078192835e-05,
+      "loss": 0.8853,
+      "step": 7760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8447936800101323e-05,
+      "loss": 1.2063,
+      "step": 7761
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8447480460575474e-05,
+      "loss": 0.9731,
+      "step": 7762
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844702405961861e-05,
+      "loss": 1.1079,
+      "step": 7763
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446567597234048e-05,
+      "loss": 1.4411,
+      "step": 7764
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446111073425105e-05,
+      "loss": 1.0611,
+      "step": 7765
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844565448819511e-05,
+      "loss": 0.8519,
+      "step": 7766
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8445197841547378e-05,
+      "loss": 0.9254,
+      "step": 7767
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8444741133485225e-05,
+      "loss": 0.9438,
+      "step": 7768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844428436401198e-05,
+      "loss": 1.0237,
+      "step": 7769
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844382753313096e-05,
+      "loss": 0.9055,
+      "step": 7770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844337064084549e-05,
+      "loss": 0.9174,
+      "step": 7771
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844291368715889e-05,
+      "loss": 0.9476,
+      "step": 7772
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8442456672074482e-05,
+      "loss": 1.0955,
+      "step": 7773
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8441999595595592e-05,
+      "loss": 1.1133,
+      "step": 7774
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8441542457725547e-05,
+      "loss": 0.965,
+      "step": 7775
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8441085258467666e-05,
+      "loss": 1.2048,
+      "step": 7776
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8440627997825274e-05,
+      "loss": 1.2265,
+      "step": 7777
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84401706758017e-05,
+      "loss": 1.1532,
+      "step": 7778
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439713292400267e-05,
+      "loss": 0.9688,
+      "step": 7779
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439255847624303e-05,
+      "loss": 1.0052,
+      "step": 7780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843879834147713e-05,
+      "loss": 1.2621,
+      "step": 7781
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843834077396208e-05,
+      "loss": 1.0765,
+      "step": 7782
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437883145082482e-05,
+      "loss": 1.3367,
+      "step": 7783
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843742545484166e-05,
+      "loss": 1.0538,
+      "step": 7784
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8436967703242938e-05,
+      "loss": 1.2447,
+      "step": 7785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8436509890289653e-05,
+      "loss": 1.0007,
+      "step": 7786
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8436052015985135e-05,
+      "loss": 0.7087,
+      "step": 7787
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8435594080332703e-05,
+      "loss": 0.9983,
+      "step": 7788
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8435136083335697e-05,
+      "loss": 0.9914,
+      "step": 7789
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434678024997445e-05,
+      "loss": 0.9445,
+      "step": 7790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434219905321278e-05,
+      "loss": 0.9662,
+      "step": 7791
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433761724310527e-05,
+      "loss": 1.2372,
+      "step": 7792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433303481968523e-05,
+      "loss": 1.0306,
+      "step": 7793
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432845178298603e-05,
+      "loss": 1.3047,
+      "step": 7794
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432386813304094e-05,
+      "loss": 1.1269,
+      "step": 7795
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431928386988332e-05,
+      "loss": 0.8408,
+      "step": 7796
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431469899354647e-05,
+      "loss": 0.9991,
+      "step": 7797
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431011350406384e-05,
+      "loss": 1.0738,
+      "step": 7798
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430552740146864e-05,
+      "loss": 0.6234,
+      "step": 7799
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843009406857943e-05,
+      "loss": 0.9654,
+      "step": 7800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8429635335707417e-05,
+      "loss": 1.207,
+      "step": 7801
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842917654153416e-05,
+      "loss": 1.2223,
+      "step": 7802
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428717686062994e-05,
+      "loss": 1.2927,
+      "step": 7803
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428258769297257e-05,
+      "loss": 0.948,
+      "step": 7804
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427799791240287e-05,
+      "loss": 0.9007,
+      "step": 7805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427340751895418e-05,
+      "loss": 1.0886,
+      "step": 7806
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426881651265994e-05,
+      "loss": 0.7946,
+      "step": 7807
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426422489355352e-05,
+      "loss": 0.8835,
+      "step": 7808
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8425963266166827e-05,
+      "loss": 1.153,
+      "step": 7809
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8425503981703766e-05,
+      "loss": 1.3691,
+      "step": 7810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84250446359695e-05,
+      "loss": 1.1474,
+      "step": 7811
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424585228967377e-05,
+      "loss": 1.3359,
+      "step": 7812
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842412576070073e-05,
+      "loss": 1.0876,
+      "step": 7813
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423666231172908e-05,
+      "loss": 1.2909,
+      "step": 7814
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842320664038725e-05,
+      "loss": 0.9205,
+      "step": 7815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422746988347098e-05,
+      "loss": 1.1926,
+      "step": 7816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422287275055795e-05,
+      "loss": 0.9294,
+      "step": 7817
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8421827500516682e-05,
+      "loss": 1.0405,
+      "step": 7818
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8421367664733106e-05,
+      "loss": 1.1506,
+      "step": 7819
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420907767708408e-05,
+      "loss": 1.0992,
+      "step": 7820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842044780944593e-05,
+      "loss": 0.9641,
+      "step": 7821
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419987789949026e-05,
+      "loss": 1.1623,
+      "step": 7822
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419527709221035e-05,
+      "loss": 0.9623,
+      "step": 7823
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419067567265304e-05,
+      "loss": 1.2053,
+      "step": 7824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418607364085175e-05,
+      "loss": 1.0064,
+      "step": 7825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418147099684002e-05,
+      "loss": 1.0154,
+      "step": 7826
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417686774065128e-05,
+      "loss": 1.3597,
+      "step": 7827
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417226387231898e-05,
+      "loss": 1.0617,
+      "step": 7828
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416765939187662e-05,
+      "loss": 0.8413,
+      "step": 7829
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416305429935776e-05,
+      "loss": 0.9768,
+      "step": 7830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415844859479576e-05,
+      "loss": 1.0502,
+      "step": 7831
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415384227822423e-05,
+      "loss": 1.0082,
+      "step": 7832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414923534967658e-05,
+      "loss": 1.1307,
+      "step": 7833
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414462780918635e-05,
+      "loss": 1.2617,
+      "step": 7834
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841400196567871e-05,
+      "loss": 1.3071,
+      "step": 7835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841354108925122e-05,
+      "loss": 1.1967,
+      "step": 7836
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413080151639525e-05,
+      "loss": 0.9658,
+      "step": 7837
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412619152846983e-05,
+      "loss": 0.792,
+      "step": 7838
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412158092876934e-05,
+      "loss": 0.9592,
+      "step": 7839
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411696971732742e-05,
+      "loss": 1.2377,
+      "step": 7840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411235789417753e-05,
+      "loss": 0.9034,
+      "step": 7841
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8410774545935325e-05,
+      "loss": 1.0707,
+      "step": 7842
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841031324128881e-05,
+      "loss": 1.2819,
+      "step": 7843
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409851875481563e-05,
+      "loss": 1.0615,
+      "step": 7844
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840939044851694e-05,
+      "loss": 0.7618,
+      "step": 7845
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408928960398294e-05,
+      "loss": 1.0009,
+      "step": 7846
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408467411128984e-05,
+      "loss": 1.2846,
+      "step": 7847
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408005800712365e-05,
+      "loss": 1.1764,
+      "step": 7848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8407544129151796e-05,
+      "loss": 0.9434,
+      "step": 7849
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840708239645063e-05,
+      "loss": 1.0719,
+      "step": 7850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8406620602612228e-05,
+      "loss": 1.3031,
+      "step": 7851
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8406158747639945e-05,
+      "loss": 0.9521,
+      "step": 7852
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8405696831537145e-05,
+      "loss": 0.9696,
+      "step": 7853
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8405234854307182e-05,
+      "loss": 1.0698,
+      "step": 7854
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840477281595342e-05,
+      "loss": 0.9426,
+      "step": 7855
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404310716479217e-05,
+      "loss": 0.5899,
+      "step": 7856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840384855588793e-05,
+      "loss": 0.8154,
+      "step": 7857
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403386334182927e-05,
+      "loss": 1.0442,
+      "step": 7858
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402924051367566e-05,
+      "loss": 0.9164,
+      "step": 7859
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402461707445206e-05,
+      "loss": 1.2248,
+      "step": 7860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840199930241921e-05,
+      "loss": 1.1184,
+      "step": 7861
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401536836292944e-05,
+      "loss": 1.2064,
+      "step": 7862
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401074309069767e-05,
+      "loss": 1.1565,
+      "step": 7863
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840061172075305e-05,
+      "loss": 0.9745,
+      "step": 7864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8400149071346145e-05,
+      "loss": 0.9589,
+      "step": 7865
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839968636085243e-05,
+      "loss": 1.1085,
+      "step": 7866
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8399223589275256e-05,
+      "loss": 0.6519,
+      "step": 7867
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398760756617996e-05,
+      "loss": 0.8097,
+      "step": 7868
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398297862884022e-05,
+      "loss": 1.0308,
+      "step": 7869
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8397834908076686e-05,
+      "loss": 0.9999,
+      "step": 7870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8397371892199365e-05,
+      "loss": 0.733,
+      "step": 7871
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396908815255424e-05,
+      "loss": 1.1734,
+      "step": 7872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396445677248227e-05,
+      "loss": 0.9669,
+      "step": 7873
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8395982478181146e-05,
+      "loss": 1.1552,
+      "step": 7874
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839551921805755e-05,
+      "loss": 1.0762,
+      "step": 7875
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8395055896880802e-05,
+      "loss": 1.1589,
+      "step": 7876
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394592514654278e-05,
+      "loss": 1.1878,
+      "step": 7877
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394129071381343e-05,
+      "loss": 1.3691,
+      "step": 7878
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8393665567065374e-05,
+      "loss": 1.235,
+      "step": 7879
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8393202001709733e-05,
+      "loss": 0.9698,
+      "step": 7880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8392738375317797e-05,
+      "loss": 0.8772,
+      "step": 7881
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8392274687892935e-05,
+      "loss": 0.6081,
+      "step": 7882
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8391810939438516e-05,
+      "loss": 0.9498,
+      "step": 7883
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839134712995792e-05,
+      "loss": 0.9905,
+      "step": 7884
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390883259454514e-05,
+      "loss": 0.9989,
+      "step": 7885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390419327931673e-05,
+      "loss": 1.078,
+      "step": 7886
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389955335392774e-05,
+      "loss": 1.3093,
+      "step": 7887
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389491281841187e-05,
+      "loss": 1.0597,
+      "step": 7888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389027167280283e-05,
+      "loss": 0.8386,
+      "step": 7889
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388562991713447e-05,
+      "loss": 1.2348,
+      "step": 7890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388098755144047e-05,
+      "loss": 1.2896,
+      "step": 7891
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838763445757546e-05,
+      "loss": 0.81,
+      "step": 7892
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387170099011067e-05,
+      "loss": 0.7969,
+      "step": 7893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8386705679454243e-05,
+      "loss": 1.3204,
+      "step": 7894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838624119890836e-05,
+      "loss": 1.0757,
+      "step": 7895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8385776657376802e-05,
+      "loss": 1.0876,
+      "step": 7896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8385312054862945e-05,
+      "loss": 0.9695,
+      "step": 7897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384847391370166e-05,
+      "loss": 1.2746,
+      "step": 7898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384382666901847e-05,
+      "loss": 1.2363,
+      "step": 7899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8383917881461366e-05,
+      "loss": 0.9406,
+      "step": 7900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83834530350521e-05,
+      "loss": 0.9306,
+      "step": 7901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8382988127677437e-05,
+      "loss": 1.1294,
+      "step": 7902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8382523159340753e-05,
+      "loss": 0.7248,
+      "step": 7903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838205813004543e-05,
+      "loss": 0.9345,
+      "step": 7904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8381593039794846e-05,
+      "loss": 1.0311,
+      "step": 7905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8381127888592388e-05,
+      "loss": 1.0222,
+      "step": 7906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8380662676441438e-05,
+      "loss": 0.6687,
+      "step": 7907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838019740334538e-05,
+      "loss": 1.1214,
+      "step": 7908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837973206930759e-05,
+      "loss": 0.7969,
+      "step": 7909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8379266674331466e-05,
+      "loss": 0.8393,
+      "step": 7910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837880121842038e-05,
+      "loss": 0.7814,
+      "step": 7911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837833570157772e-05,
+      "loss": 0.8619,
+      "step": 7912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377870123806873e-05,
+      "loss": 1.1452,
+      "step": 7913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377404485111228e-05,
+      "loss": 0.718,
+      "step": 7914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837693878549416e-05,
+      "loss": 1.0587,
+      "step": 7915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8376473024959068e-05,
+      "loss": 1.169,
+      "step": 7916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8376007203509336e-05,
+      "loss": 1.131,
+      "step": 7917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8375541321148347e-05,
+      "loss": 1.324,
+      "step": 7918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8375075377879488e-05,
+      "loss": 1.2296,
+      "step": 7919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374609373706156e-05,
+      "loss": 1.3554,
+      "step": 7920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374143308631733e-05,
+      "loss": 0.9941,
+      "step": 7921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837367718265961e-05,
+      "loss": 0.867,
+      "step": 7922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373210995793176e-05,
+      "loss": 1.1844,
+      "step": 7923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8372744748035824e-05,
+      "loss": 1.3052,
+      "step": 7924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837227843939094e-05,
+      "loss": 1.2413,
+      "step": 7925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371812069861917e-05,
+      "loss": 0.9436,
+      "step": 7926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371345639452146e-05,
+      "loss": 0.9295,
+      "step": 7927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370879148165022e-05,
+      "loss": 1.0524,
+      "step": 7928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370412596003934e-05,
+      "loss": 0.9372,
+      "step": 7929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8369945982972277e-05,
+      "loss": 0.9031,
+      "step": 7930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8369479309073442e-05,
+      "loss": 1.2895,
+      "step": 7931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8369012574310828e-05,
+      "loss": 0.9922,
+      "step": 7932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368545778687822e-05,
+      "loss": 1.1863,
+      "step": 7933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836807892220782e-05,
+      "loss": 1.0056,
+      "step": 7934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8367612004874224e-05,
+      "loss": 1.0106,
+      "step": 7935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836714502669042e-05,
+      "loss": 0.9766,
+      "step": 7936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366677987659812e-05,
+      "loss": 1.2162,
+      "step": 7937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836621088778579e-05,
+      "loss": 0.8623,
+      "step": 7938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836574372707175e-05,
+      "loss": 0.9279,
+      "step": 7939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83652765055211e-05,
+      "loss": 1.185,
+      "step": 7940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8364809223137225e-05,
+      "loss": 1.1695,
+      "step": 7941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836434187992353e-05,
+      "loss": 0.8594,
+      "step": 7942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8363874475883416e-05,
+      "loss": 1.0301,
+      "step": 7943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836340701102027e-05,
+      "loss": 0.9869,
+      "step": 7944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8362939485337502e-05,
+      "loss": 1.2532,
+      "step": 7945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8362471898838515e-05,
+      "loss": 1.2846,
+      "step": 7946
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83620042515267e-05,
+      "loss": 0.7048,
+      "step": 7947
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8361536543405458e-05,
+      "loss": 0.9855,
+      "step": 7948
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8361068774478197e-05,
+      "loss": 1.0103,
+      "step": 7949
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8360600944748316e-05,
+      "loss": 1.0391,
+      "step": 7950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8360133054219217e-05,
+      "loss": 0.8611,
+      "step": 7951
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83596651028943e-05,
+      "loss": 0.9008,
+      "step": 7952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8359197090776972e-05,
+      "loss": 1.0728,
+      "step": 7953
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8358729017870634e-05,
+      "loss": 1.1764,
+      "step": 7954
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835826088417869e-05,
+      "loss": 1.0229,
+      "step": 7955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8357792689704543e-05,
+      "loss": 1.0903,
+      "step": 7956
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83573244344516e-05,
+      "loss": 0.9241,
+      "step": 7957
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835685611842327e-05,
+      "loss": 0.9863,
+      "step": 7958
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8356387741622947e-05,
+      "loss": 0.9966,
+      "step": 7959
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8355919304054047e-05,
+      "loss": 1.2213,
+      "step": 7960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835545080571998e-05,
+      "loss": 1.0063,
+      "step": 7961
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835498224662414e-05,
+      "loss": 0.8688,
+      "step": 7962
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8354513626769945e-05,
+      "loss": 0.9954,
+      "step": 7963
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8354044946160797e-05,
+      "loss": 1.2103,
+      "step": 7964
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353576204800106e-05,
+      "loss": 1.0397,
+      "step": 7965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353107402691282e-05,
+      "loss": 1.3391,
+      "step": 7966
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352638539837735e-05,
+      "loss": 0.9079,
+      "step": 7967
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352169616242868e-05,
+      "loss": 0.9803,
+      "step": 7968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83517006319101e-05,
+      "loss": 1.2144,
+      "step": 7969
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835123158684284e-05,
+      "loss": 1.0951,
+      "step": 7970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835076248104449e-05,
+      "loss": 1.1191,
+      "step": 7971
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8350293314518475e-05,
+      "loss": 0.9924,
+      "step": 7972
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83498240872682e-05,
+      "loss": 1.0539,
+      "step": 7973
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834935479929707e-05,
+      "loss": 0.9433,
+      "step": 7974
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348885450608506e-05,
+      "loss": 0.9605,
+      "step": 7975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348416041205923e-05,
+      "loss": 0.9076,
+      "step": 7976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834794657109273e-05,
+      "loss": 0.9594,
+      "step": 7977
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347477040272345e-05,
+      "loss": 1.026,
+      "step": 7978
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347007448748177e-05,
+      "loss": 1.0829,
+      "step": 7979
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8346537796523643e-05,
+      "loss": 1.2838,
+      "step": 7980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834606808360216e-05,
+      "loss": 1.0432,
+      "step": 7981
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834559830998715e-05,
+      "loss": 1.1754,
+      "step": 7982
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8345128475682016e-05,
+      "loss": 0.7493,
+      "step": 7983
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834465858069018e-05,
+      "loss": 1.0317,
+      "step": 7984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834418862501506e-05,
+      "loss": 1.1723,
+      "step": 7985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343718608660076e-05,
+      "loss": 1.045,
+      "step": 7986
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8343248531628644e-05,
+      "loss": 0.936,
+      "step": 7987
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834277839392418e-05,
+      "loss": 0.6814,
+      "step": 7988
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834230819555011e-05,
+      "loss": 1.1587,
+      "step": 7989
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8341837936509844e-05,
+      "loss": 0.7587,
+      "step": 7990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8341367616806807e-05,
+      "loss": 0.9901,
+      "step": 7991
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834089723644442e-05,
+      "loss": 0.9609,
+      "step": 7992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83404267954261e-05,
+      "loss": 1.0264,
+      "step": 7993
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833995629375527e-05,
+      "loss": 1.2289,
+      "step": 7994
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833948573143535e-05,
+      "loss": 1.0225,
+      "step": 7995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8339015108469768e-05,
+      "loss": 0.9347,
+      "step": 7996
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8338544424861942e-05,
+      "loss": 0.793,
+      "step": 7997
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8338073680615292e-05,
+      "loss": 1.0759,
+      "step": 7998
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337602875733243e-05,
+      "loss": 1.0266,
+      "step": 7999
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337132010219224e-05,
+      "loss": 1.1884,
+      "step": 8000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336661084076653e-05,
+      "loss": 0.6741,
+      "step": 8001
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833619009730896e-05,
+      "loss": 0.7548,
+      "step": 8002
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833571904991956e-05,
+      "loss": 1.2341,
+      "step": 8003
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833524794191189e-05,
+      "loss": 1.0845,
+      "step": 8004
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8334776773289375e-05,
+      "loss": 1.0601,
+      "step": 8005
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8334305544055434e-05,
+      "loss": 1.0625,
+      "step": 8006
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333834254213497e-05,
+      "loss": 0.7798,
+      "step": 8007
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333362903766993e-05,
+      "loss": 1.0012,
+      "step": 8008
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332891492719347e-05,
+      "loss": 1.1388,
+      "step": 8009
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332420021073992e-05,
+      "loss": 0.8211,
+      "step": 8010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331948488834346e-05,
+      "loss": 1.2187,
+      "step": 8011
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331476896003852e-05,
+      "loss": 1.0159,
+      "step": 8012
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833100524258593e-05,
+      "loss": 0.8897,
+      "step": 8013
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8330533528584016e-05,
+      "loss": 1.4206,
+      "step": 8014
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8330061754001532e-05,
+      "loss": 0.9822,
+      "step": 8015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832958991884192e-05,
+      "loss": 0.9232,
+      "step": 8016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8329118023108603e-05,
+      "loss": 0.9953,
+      "step": 8017
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328646066805013e-05,
+      "loss": 1.2748,
+      "step": 8018
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328174049934585e-05,
+      "loss": 1.255,
+      "step": 8019
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832770197250075e-05,
+      "loss": 1.385,
+      "step": 8020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327229834506938e-05,
+      "loss": 0.9716,
+      "step": 8021
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8326757635956593e-05,
+      "loss": 1.2233,
+      "step": 8022
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832628537685314e-05,
+      "loss": 0.97,
+      "step": 8023
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8325813057200012e-05,
+      "loss": 1.2184,
+      "step": 8024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8325340677000646e-05,
+      "loss": 1.0288,
+      "step": 8025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324868236258482e-05,
+      "loss": 1.1998,
+      "step": 8026
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324395734976953e-05,
+      "loss": 1.3069,
+      "step": 8027
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832392317315949e-05,
+      "loss": 1.1629,
+      "step": 8028
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8323450550809535e-05,
+      "loss": 0.8167,
+      "step": 8029
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832297786793052e-05,
+      "loss": 1.2497,
+      "step": 8030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322505124525888e-05,
+      "loss": 1.1804,
+      "step": 8031
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322032320599076e-05,
+      "loss": 1.1821,
+      "step": 8032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8321559456153523e-05,
+      "loss": 0.789,
+      "step": 8033
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832108653119266e-05,
+      "loss": 0.8882,
+      "step": 8034
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832061354571993e-05,
+      "loss": 0.9974,
+      "step": 8035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832014049973878e-05,
+      "loss": 0.8967,
+      "step": 8036
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8319667393252644e-05,
+      "loss": 0.7161,
+      "step": 8037
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8319194226264963e-05,
+      "loss": 1.0656,
+      "step": 8038
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8318720998779174e-05,
+      "loss": 1.2152,
+      "step": 8039
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8318247710798728e-05,
+      "loss": 1.52,
+      "step": 8040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317774362327056e-05,
+      "loss": 1.135,
+      "step": 8041
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831730095336761e-05,
+      "loss": 0.9181,
+      "step": 8042
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8316827483923822e-05,
+      "loss": 1.2204,
+      "step": 8043
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8316353953999142e-05,
+      "loss": 0.5923,
+      "step": 8044
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8315880363597015e-05,
+      "loss": 1.1595,
+      "step": 8045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8315406712720878e-05,
+      "loss": 0.8599,
+      "step": 8046
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314933001374186e-05,
+      "loss": 1.0763,
+      "step": 8047
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314459229560376e-05,
+      "loss": 1.3488,
+      "step": 8048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313985397282894e-05,
+      "loss": 1.1303,
+      "step": 8049
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313511504545184e-05,
+      "loss": 0.9631,
+      "step": 8050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83130375513507e-05,
+      "loss": 1.2916,
+      "step": 8051
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312563537702885e-05,
+      "loss": 1.4151,
+      "step": 8052
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831208946360518e-05,
+      "loss": 1.0618,
+      "step": 8053
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831161532906104e-05,
+      "loss": 1.3949,
+      "step": 8054
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831114113407391e-05,
+      "loss": 1.1088,
+      "step": 8055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831066687864724e-05,
+      "loss": 1.3815,
+      "step": 8056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8310192562784473e-05,
+      "loss": 1.0642,
+      "step": 8057
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830971818648907e-05,
+      "loss": 1.1362,
+      "step": 8058
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830924374976447e-05,
+      "loss": 0.9041,
+      "step": 8059
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8308769252614124e-05,
+      "loss": 1.2005,
+      "step": 8060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8308294695041486e-05,
+      "loss": 1.2091,
+      "step": 8061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830782007705001e-05,
+      "loss": 0.9884,
+      "step": 8062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830734539864314e-05,
+      "loss": 1.1011,
+      "step": 8063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306870659824335e-05,
+      "loss": 0.8289,
+      "step": 8064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306395860597043e-05,
+      "loss": 1.0723,
+      "step": 8065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830592100096472e-05,
+      "loss": 1.3052,
+      "step": 8066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830544608093081e-05,
+      "loss": 1.462,
+      "step": 8067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830497110049878e-05,
+      "loss": 1.1877,
+      "step": 8068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830449605967208e-05,
+      "loss": 1.0548,
+      "step": 8069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8304020958454156e-05,
+      "loss": 0.8241,
+      "step": 8070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8303545796848473e-05,
+      "loss": 1.0308,
+      "step": 8071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8303070574858485e-05,
+      "loss": 0.6769,
+      "step": 8072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830259529248764e-05,
+      "loss": 0.8911,
+      "step": 8073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8302119949739406e-05,
+      "loss": 1.1486,
+      "step": 8074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830164454661723e-05,
+      "loss": 1.0848,
+      "step": 8075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8301169083124573e-05,
+      "loss": 1.3681,
+      "step": 8076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8300693559264894e-05,
+      "loss": 1.005,
+      "step": 8077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830021797504165e-05,
+      "loss": 0.8725,
+      "step": 8078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8299742330458295e-05,
+      "loss": 0.9753,
+      "step": 8079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82992666255183e-05,
+      "loss": 0.9845,
+      "step": 8080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829879086022511e-05,
+      "loss": 1.1532,
+      "step": 8081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8298315034582192e-05,
+      "loss": 1.0872,
+      "step": 8082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8297839148593007e-05,
+      "loss": 1.035,
+      "step": 8083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8297363202261013e-05,
+      "loss": 1.0593,
+      "step": 8084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8296887195589678e-05,
+      "loss": 1.1946,
+      "step": 8085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829641112858245e-05,
+      "loss": 1.0161,
+      "step": 8086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8295935001242805e-05,
+      "loss": 1.162,
+      "step": 8087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8295458813574193e-05,
+      "loss": 1.1313,
+      "step": 8088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829498256558009e-05,
+      "loss": 1.323,
+      "step": 8089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829450625726395e-05,
+      "loss": 1.1125,
+      "step": 8090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8294029888629235e-05,
+      "loss": 0.9606,
+      "step": 8091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829355345967942e-05,
+      "loss": 1.2212,
+      "step": 8092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8293076970417964e-05,
+      "loss": 0.9942,
+      "step": 8093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829260042084833e-05,
+      "loss": 1.0162,
+      "step": 8094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829212381097398e-05,
+      "loss": 1.4019,
+      "step": 8095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829164714079839e-05,
+      "loss": 1.2975,
+      "step": 8096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8291170410325016e-05,
+      "loss": 1.133,
+      "step": 8097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290693619557335e-05,
+      "loss": 1.0828,
+      "step": 8098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290216768498807e-05,
+      "loss": 0.8693,
+      "step": 8099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8289739857152903e-05,
+      "loss": 0.9975,
+      "step": 8100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828926288552309e-05,
+      "loss": 1.1445,
+      "step": 8101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288785853612837e-05,
+      "loss": 0.9188,
+      "step": 8102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288308761425612e-05,
+      "loss": 0.8046,
+      "step": 8103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8287831608964887e-05,
+      "loss": 1.0873,
+      "step": 8104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828735439623413e-05,
+      "loss": 0.9565,
+      "step": 8105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828687712323681e-05,
+      "loss": 1.1449,
+      "step": 8106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82863997899764e-05,
+      "loss": 1.1048,
+      "step": 8107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828592239645637e-05,
+      "loss": 0.7024,
+      "step": 8108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8285444942680195e-05,
+      "loss": 0.8871,
+      "step": 8109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8284967428651338e-05,
+      "loss": 1.2278,
+      "step": 8110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828448985437328e-05,
+      "loss": 1.0583,
+      "step": 8111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8284012219849495e-05,
+      "loss": 1.1486,
+      "step": 8112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8283534525083454e-05,
+      "loss": 0.9203,
+      "step": 8113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828305677007863e-05,
+      "loss": 1.1156,
+      "step": 8114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282578954838493e-05,
+      "loss": 0.9166,
+      "step": 8115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282101079366525e-05,
+      "loss": 1.2084,
+      "step": 8116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8281623143666198e-05,
+      "loss": 1.0947,
+      "step": 8117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8281145147740988e-05,
+      "loss": 0.9784,
+      "step": 8118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280667091594368e-05,
+      "loss": 1.0949,
+      "step": 8119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828018897522982e-05,
+      "loss": 1.3787,
+      "step": 8120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827971079865082e-05,
+      "loss": 0.9102,
+      "step": 8121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8279232561860844e-05,
+      "loss": 1.1923,
+      "step": 8122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8278754264863363e-05,
+      "loss": 1.0757,
+      "step": 8123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8278275907661867e-05,
+      "loss": 1.2549,
+      "step": 8124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827779749025983e-05,
+      "loss": 1.1213,
+      "step": 8125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277319012660727e-05,
+      "loss": 0.7919,
+      "step": 8126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8276840474868043e-05,
+      "loss": 1.1492,
+      "step": 8127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8276361876885255e-05,
+      "loss": 0.8492,
+      "step": 8128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275883218715842e-05,
+      "loss": 1.0745,
+      "step": 8129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275404500363293e-05,
+      "loss": 0.6171,
+      "step": 8130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827492572183108e-05,
+      "loss": 1.277,
+      "step": 8131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8274446883122688e-05,
+      "loss": 0.7592,
+      "step": 8132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82739679842416e-05,
+      "loss": 1.1007,
+      "step": 8133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82734890251913e-05,
+      "loss": 1.2117,
+      "step": 8134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8273010005975265e-05,
+      "loss": 1.2103,
+      "step": 8135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272530926596985e-05,
+      "loss": 0.9881,
+      "step": 8136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272051787059945e-05,
+      "loss": 1.1515,
+      "step": 8137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827157258736762e-05,
+      "loss": 0.7873,
+      "step": 8138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8271093327523503e-05,
+      "loss": 1.05,
+      "step": 8139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8270614007531074e-05,
+      "loss": 0.9677,
+      "step": 8140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8270134627393824e-05,
+      "loss": 1.199,
+      "step": 8141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826965518711524e-05,
+      "loss": 0.9542,
+      "step": 8142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82691756866988e-05,
+      "loss": 1.2839,
+      "step": 8143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8268696126148e-05,
+      "loss": 0.9088,
+      "step": 8144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8268216505466318e-05,
+      "loss": 1.1726,
+      "step": 8145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826773682465725e-05,
+      "loss": 1.0692,
+      "step": 8146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8267257083724287e-05,
+      "loss": 0.9204,
+      "step": 8147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8266777282670903e-05,
+      "loss": 1.1016,
+      "step": 8148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8266297421500603e-05,
+      "loss": 0.6332,
+      "step": 8149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265817500216873e-05,
+      "loss": 0.8406,
+      "step": 8150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265337518823192e-05,
+      "loss": 0.9046,
+      "step": 8151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8264857477323063e-05,
+      "loss": 1.2235,
+      "step": 8152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8264377375719974e-05,
+      "loss": 1.1288,
+      "step": 8153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826389721401741e-05,
+      "loss": 1.2238,
+      "step": 8154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8263416992218873e-05,
+      "loss": 1.0387,
+      "step": 8155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8262936710327846e-05,
+      "loss": 1.424,
+      "step": 8156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826245636834783e-05,
+      "loss": 1.0184,
+      "step": 8157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261975966282312e-05,
+      "loss": 0.7936,
+      "step": 8158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261495504134787e-05,
+      "loss": 1.3472,
+      "step": 8159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826101498190875e-05,
+      "loss": 1.1013,
+      "step": 8160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8260534399607693e-05,
+      "loss": 1.2102,
+      "step": 8161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8260053757235115e-05,
+      "loss": 1.1176,
+      "step": 8162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259573054794507e-05,
+      "loss": 1.09,
+      "step": 8163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259092292289367e-05,
+      "loss": 1.0298,
+      "step": 8164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.825861146972319e-05,
+      "loss": 0.8205,
+      "step": 8165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825813058709947e-05,
+      "loss": 1.0863,
+      "step": 8166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257649644421714e-05,
+      "loss": 1.0502,
+      "step": 8167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825716864169341e-05,
+      "loss": 0.7816,
+      "step": 8168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8256687578918056e-05,
+      "loss": 0.8157,
+      "step": 8169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8256206456099157e-05,
+      "loss": 1.0551,
+      "step": 8170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825572527324021e-05,
+      "loss": 1.3562,
+      "step": 8171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8255244030344706e-05,
+      "loss": 0.9851,
+      "step": 8172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8254762727416153e-05,
+      "loss": 0.9373,
+      "step": 8173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8254281364458047e-05,
+      "loss": 0.9172,
+      "step": 8174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8253799941473894e-05,
+      "loss": 1.317,
+      "step": 8175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8253318458467187e-05,
+      "loss": 0.9001,
+      "step": 8176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8252836915441436e-05,
+      "loss": 0.9766,
+      "step": 8177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8252355312400135e-05,
+      "loss": 1.383,
+      "step": 8178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251873649346794e-05,
+      "loss": 1.0629,
+      "step": 8179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251391926284905e-05,
+      "loss": 0.8883,
+      "step": 8180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250910143217983e-05,
+      "loss": 1.4124,
+      "step": 8181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250428300149527e-05,
+      "loss": 1.0013,
+      "step": 8182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8249946397083037e-05,
+      "loss": 0.8865,
+      "step": 8183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824946443402202e-05,
+      "loss": 0.8854,
+      "step": 8184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8248982410969986e-05,
+      "loss": 1.0025,
+      "step": 8185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824850032793043e-05,
+      "loss": 0.8656,
+      "step": 8186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8248018184906867e-05,
+      "loss": 1.0773,
+      "step": 8187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82475359819028e-05,
+      "loss": 0.9559,
+      "step": 8188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247053718921733e-05,
+      "loss": 1.1503,
+      "step": 8189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824657139596718e-05,
+      "loss": 1.1509,
+      "step": 8190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8246089013042642e-05,
+      "loss": 1.0965,
+      "step": 8191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824560657015163e-05,
+      "loss": 1.5407,
+      "step": 8192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245124067297647e-05,
+      "loss": 0.9105,
+      "step": 8193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824464150448421e-05,
+      "loss": 1.17,
+      "step": 8194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8244158881714825e-05,
+      "loss": 0.9984,
+      "step": 8195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8243676198993e-05,
+      "loss": 1.1918,
+      "step": 8196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8243193456322242e-05,
+      "loss": 1.0366,
+      "step": 8197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8242710653706074e-05,
+      "loss": 0.987,
+      "step": 8198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8242227791147994e-05,
+      "loss": 1.0513,
+      "step": 8199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8241744868651517e-05,
+      "loss": 1.3218,
+      "step": 8200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8241261886220155e-05,
+      "loss": 1.4057,
+      "step": 8201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8240778843857427e-05,
+      "loss": 1.0259,
+      "step": 8202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8240295741566837e-05,
+      "loss": 1.064,
+      "step": 8203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8239812579351902e-05,
+      "loss": 1.0832,
+      "step": 8204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8239329357216135e-05,
+      "loss": 1.0129,
+      "step": 8205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823884607516305e-05,
+      "loss": 1.0301,
+      "step": 8206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823836273319616e-05,
+      "loss": 1.0236,
+      "step": 8207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8237879331318983e-05,
+      "loss": 1.0899,
+      "step": 8208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823739586953503e-05,
+      "loss": 1.3126,
+      "step": 8209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8236912347847823e-05,
+      "loss": 1.1127,
+      "step": 8210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8236428766260875e-05,
+      "loss": 1.2094,
+      "step": 8211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82359451247777e-05,
+      "loss": 1.0444,
+      "step": 8212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235461423401818e-05,
+      "loss": 1.4698,
+      "step": 8213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8234977662136745e-05,
+      "loss": 0.9205,
+      "step": 8214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8234493840986e-05,
+      "loss": 1.4985,
+      "step": 8215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82340099599531e-05,
+      "loss": 1.0558,
+      "step": 8216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823352601904157e-05,
+      "loss": 0.9876,
+      "step": 8217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233042018254923e-05,
+      "loss": 1.0897,
+      "step": 8218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823255795759668e-05,
+      "loss": 0.7423,
+      "step": 8219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823207383707036e-05,
+      "loss": 0.9027,
+      "step": 8220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8231589656679484e-05,
+      "loss": 1.0996,
+      "step": 8221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8231105416427576e-05,
+      "loss": 0.9603,
+      "step": 8222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230621116318158e-05,
+      "loss": 0.8951,
+      "step": 8223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230136756354745e-05,
+      "loss": 1.0316,
+      "step": 8224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229652336540865e-05,
+      "loss": 0.8868,
+      "step": 8225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8229167856880035e-05,
+      "loss": 1.2036,
+      "step": 8226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8228683317375792e-05,
+      "loss": 1.3245,
+      "step": 8227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822819871803164e-05,
+      "loss": 0.9781,
+      "step": 8228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227714058851117e-05,
+      "loss": 1.3202,
+      "step": 8229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227229339837746e-05,
+      "loss": 1.0385,
+      "step": 8230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822674456099505e-05,
+      "loss": 1.138,
+      "step": 8231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8226259722326553e-05,
+      "loss": 1.4538,
+      "step": 8232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822577482383578e-05,
+      "loss": 0.89,
+      "step": 8233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8225289865526258e-05,
+      "loss": 0.9206,
+      "step": 8234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224804847401518e-05,
+      "loss": 1.5744,
+      "step": 8235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822431976946508e-05,
+      "loss": 1.208,
+      "step": 8236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8223834631720482e-05,
+      "loss": 1.2182,
+      "step": 8237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822334943417124e-05,
+      "loss": 0.9979,
+      "step": 8238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822286417682089e-05,
+      "loss": 1.1239,
+      "step": 8239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222378859672956e-05,
+      "loss": 1.2442,
+      "step": 8240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822189348273097e-05,
+      "loss": 1.1239,
+      "step": 8241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221408045998463e-05,
+      "loss": 0.9584,
+      "step": 8242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220922549478964e-05,
+      "loss": 1.256,
+      "step": 8243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220436993176e-05,
+      "loss": 0.8428,
+      "step": 8244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821995137709311e-05,
+      "loss": 0.944,
+      "step": 8245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821946570123382e-05,
+      "loss": 1.1905,
+      "step": 8246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821897996560166e-05,
+      "loss": 0.8485,
+      "step": 8247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8218494170200167e-05,
+      "loss": 0.9782,
+      "step": 8248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8218008315032874e-05,
+      "loss": 1.0671,
+      "step": 8249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821752240010331e-05,
+      "loss": 0.8703,
+      "step": 8250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821703642541501e-05,
+      "loss": 0.9654,
+      "step": 8251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8216550390971507e-05,
+      "loss": 0.6856,
+      "step": 8252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8216064296776338e-05,
+      "loss": 1.218,
+      "step": 8253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821557814283304e-05,
+      "loss": 1.0225,
+      "step": 8254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821509192914514e-05,
+      "loss": 1.2424,
+      "step": 8255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8214605655716184e-05,
+      "loss": 1.0326,
+      "step": 8256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82141193225497e-05,
+      "loss": 1.3263,
+      "step": 8257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8213632929649233e-05,
+      "loss": 0.8217,
+      "step": 8258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821314647701831e-05,
+      "loss": 0.894,
+      "step": 8259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212659964660477e-05,
+      "loss": 0.9465,
+      "step": 8260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212173392579267e-05,
+      "loss": 1.3689,
+      "step": 8261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211686760778222e-05,
+      "loss": 1.1022,
+      "step": 8262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821120006926088e-05,
+      "loss": 1.0766,
+      "step": 8263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8210713318030775e-05,
+      "loss": 1.143,
+      "step": 8264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8210226507091454e-05,
+      "loss": 0.993,
+      "step": 8265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820973963644645e-05,
+      "loss": 1.0349,
+      "step": 8266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8209252706099313e-05,
+      "loss": 0.988,
+      "step": 8267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8208765716053575e-05,
+      "loss": 0.8543,
+      "step": 8268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8208278666312786e-05,
+      "loss": 1.0148,
+      "step": 8269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8207791556880477e-05,
+      "loss": 0.9707,
+      "step": 8270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82073043877602e-05,
+      "loss": 1.0002,
+      "step": 8271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206817158955493e-05,
+      "loss": 0.7571,
+      "step": 8272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206329870469903e-05,
+      "loss": 0.7529,
+      "step": 8273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8205842522306966e-05,
+      "loss": 1.0845,
+      "step": 8274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8205355114470235e-05,
+      "loss": 1.2999,
+      "step": 8275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820486764696325e-05,
+      "loss": 0.8695,
+      "step": 8276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8204380119789554e-05,
+      "loss": 0.9497,
+      "step": 8277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203892532952695e-05,
+      "loss": 0.8538,
+      "step": 8278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820340488645622e-05,
+      "loss": 0.8855,
+      "step": 8279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202917180303673e-05,
+      "loss": 0.9121,
+      "step": 8280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82024294144986e-05,
+      "loss": 1.0818,
+      "step": 8281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820194158904455e-05,
+      "loss": 0.8802,
+      "step": 8282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201453703945074e-05,
+      "loss": 1.1292,
+      "step": 8283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820096575920371e-05,
+      "loss": 0.9563,
+      "step": 8284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820047775482402e-05,
+      "loss": 1.4757,
+      "step": 8285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8199989690809536e-05,
+      "loss": 0.8904,
+      "step": 8286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8199501567163822e-05,
+      "loss": 0.6151,
+      "step": 8287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819901338389042e-05,
+      "loss": 1.1568,
+      "step": 8288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198525140992883e-05,
+      "loss": 0.9816,
+      "step": 8289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819803683847476e-05,
+      "loss": 1.0279,
+      "step": 8290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197548476339604e-05,
+      "loss": 0.8695,
+      "step": 8291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197060054590965e-05,
+      "loss": 0.9355,
+      "step": 8292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819657157323239e-05,
+      "loss": 1.1849,
+      "step": 8293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8196083032267443e-05,
+      "loss": 1.2392,
+      "step": 8294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819559443169967e-05,
+      "loss": 1.0801,
+      "step": 8295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819510577153262e-05,
+      "loss": 0.9775,
+      "step": 8296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8194617051769854e-05,
+      "loss": 0.9288,
+      "step": 8297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8194128272414923e-05,
+      "loss": 1.0862,
+      "step": 8298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819363943347138e-05,
+      "loss": 0.9268,
+      "step": 8299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8193150534942777e-05,
+      "loss": 0.916,
+      "step": 8300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819266157683268e-05,
+      "loss": 0.7577,
+      "step": 8301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192172559144638e-05,
+      "loss": 0.8106,
+      "step": 8302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81916834818822e-05,
+      "loss": 0.9075,
+      "step": 8303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8191194345048937e-05,
+      "loss": 1.0021,
+      "step": 8304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81907051486484e-05,
+      "loss": 1.253,
+      "step": 8305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819021589268414e-05,
+      "loss": 1.0386,
+      "step": 8306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8189726577159727e-05,
+      "loss": 1.1084,
+      "step": 8307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8189237202078707e-05,
+      "loss": 0.9298,
+      "step": 8308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8188747767444647e-05,
+      "loss": 1.0116,
+      "step": 8309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8188258273261104e-05,
+      "loss": 1.3137,
+      "step": 8310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8187768719531637e-05,
+      "loss": 0.8844,
+      "step": 8311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8187279106259806e-05,
+      "loss": 1.1002,
+      "step": 8312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8186789433449175e-05,
+      "loss": 0.8367,
+      "step": 8313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8186299701103302e-05,
+      "loss": 1.1131,
+      "step": 8314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8185809909225744e-05,
+      "loss": 0.8906,
+      "step": 8315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818532005782007e-05,
+      "loss": 1.0349,
+      "step": 8316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818483014688984e-05,
+      "loss": 1.1879,
+      "step": 8317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184340176438614e-05,
+      "loss": 1.0667,
+      "step": 8318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183850146469964e-05,
+      "loss": 1.0486,
+      "step": 8319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818336005698744e-05,
+      "loss": 1.591,
+      "step": 8320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8182869907994617e-05,
+      "loss": 1.0524,
+      "step": 8321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818237969949505e-05,
+      "loss": 0.9386,
+      "step": 8322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8181889431492315e-05,
+      "loss": 1.4787,
+      "step": 8323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818139910398997e-05,
+      "loss": 0.7075,
+      "step": 8324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818090871699158e-05,
+      "loss": 0.8752,
+      "step": 8325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8180418270500716e-05,
+      "loss": 0.8089,
+      "step": 8326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8179927764520943e-05,
+      "loss": 1.0805,
+      "step": 8327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817943719905582e-05,
+      "loss": 0.8851,
+      "step": 8328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817894657410893e-05,
+      "loss": 0.8,
+      "step": 8329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8178455889683827e-05,
+      "loss": 1.1234,
+      "step": 8330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177965145784086e-05,
+      "loss": 1.3084,
+      "step": 8331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817747434241327e-05,
+      "loss": 1.0747,
+      "step": 8332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817698347957496e-05,
+      "loss": 0.9518,
+      "step": 8333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8176492557272716e-05,
+      "loss": 0.6079,
+      "step": 8334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8176001575510108e-05,
+      "loss": 0.9355,
+      "step": 8335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817551053429071e-05,
+      "loss": 0.9744,
+      "step": 8336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817501943361809e-05,
+      "loss": 1.1717,
+      "step": 8337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8174528273495822e-05,
+      "loss": 0.9935,
+      "step": 8338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8174037053927478e-05,
+      "loss": 0.7437,
+      "step": 8339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8173545774916628e-05,
+      "loss": 0.981,
+      "step": 8340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8173054436466846e-05,
+      "loss": 1.2182,
+      "step": 8341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8172563038581702e-05,
+      "loss": 1.0993,
+      "step": 8342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817207158126477e-05,
+      "loss": 1.3891,
+      "step": 8343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817158006451963e-05,
+      "loss": 0.9848,
+      "step": 8344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171088488349856e-05,
+      "loss": 0.8497,
+      "step": 8345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8170596852759016e-05,
+      "loss": 0.8159,
+      "step": 8346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8170105157750685e-05,
+      "loss": 1.1515,
+      "step": 8347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8169613403328445e-05,
+      "loss": 0.8811,
+      "step": 8348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816912158949587e-05,
+      "loss": 1.2246,
+      "step": 8349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8168629716256532e-05,
+      "loss": 0.7747,
+      "step": 8350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8168137783614016e-05,
+      "loss": 1.1337,
+      "step": 8351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816764579157189e-05,
+      "loss": 0.994,
+      "step": 8352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816715374013374e-05,
+      "loss": 0.9914,
+      "step": 8353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816666162930314e-05,
+      "loss": 1.3031,
+      "step": 8354
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8166169459083673e-05,
+      "loss": 1.0708,
+      "step": 8355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816567722947891e-05,
+      "loss": 1.2417,
+      "step": 8356
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8165184940492438e-05,
+      "loss": 1.0721,
+      "step": 8357
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8164692592127835e-05,
+      "loss": 0.9563,
+      "step": 8358
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816420018438868e-05,
+      "loss": 1.0454,
+      "step": 8359
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816370771727855e-05,
+      "loss": 0.8839,
+      "step": 8360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816321519080104e-05,
+      "loss": 1.1378,
+      "step": 8361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8162722604959718e-05,
+      "loss": 0.883,
+      "step": 8362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816222995975817e-05,
+      "loss": 0.9596,
+      "step": 8363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816173725519998e-05,
+      "loss": 0.8209,
+      "step": 8364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816124449128873e-05,
+      "loss": 0.9613,
+      "step": 8365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160751668028004e-05,
+      "loss": 1.3675,
+      "step": 8366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8160258785421385e-05,
+      "loss": 0.788,
+      "step": 8367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8159765843472457e-05,
+      "loss": 0.8044,
+      "step": 8368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815927284218481e-05,
+      "loss": 1.2952,
+      "step": 8369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8158779781562022e-05,
+      "loss": 0.5934,
+      "step": 8370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815828666160768e-05,
+      "loss": 0.8987,
+      "step": 8371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8157793482325377e-05,
+      "loss": 0.6106,
+      "step": 8372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8157300243718687e-05,
+      "loss": 0.9255,
+      "step": 8373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8156806945791207e-05,
+      "loss": 1.3313,
+      "step": 8374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815631358854652e-05,
+      "loss": 1.2358,
+      "step": 8375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155820171988215e-05,
+      "loss": 1.0907,
+      "step": 8376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155326696119882e-05,
+      "loss": 0.8134,
+      "step": 8377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154833160945108e-05,
+      "loss": 1.2317,
+      "step": 8378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815433956646748e-05,
+      "loss": 1.0276,
+      "step": 8379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815384591269059e-05,
+      "loss": 1.2526,
+      "step": 8380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8153352199618025e-05,
+      "loss": 1.2494,
+      "step": 8381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815285842725338e-05,
+      "loss": 1.2294,
+      "step": 8382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815236459560024e-05,
+      "loss": 1.0663,
+      "step": 8383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151870704662202e-05,
+      "loss": 1.0051,
+      "step": 8384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151376754442856e-05,
+      "loss": 1.0694,
+      "step": 8385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815088274494579e-05,
+      "loss": 1.2512,
+      "step": 8386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8150388676174603e-05,
+      "loss": 1.0618,
+      "step": 8387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814989454813288e-05,
+      "loss": 0.907,
+      "step": 8388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814940036082422e-05,
+      "loss": 0.8085,
+      "step": 8389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8148906114252217e-05,
+      "loss": 0.7432,
+      "step": 8390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8148411808420463e-05,
+      "loss": 1.2042,
+      "step": 8391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147917443332558e-05,
+      "loss": 1.4075,
+      "step": 8392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147423018992086e-05,
+      "loss": 0.8864,
+      "step": 8393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146928535402652e-05,
+      "loss": 0.9248,
+      "step": 8394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146433992567848e-05,
+      "loss": 1.2279,
+      "step": 8395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814593939049127e-05,
+      "loss": 0.9202,
+      "step": 8396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814544472917652e-05,
+      "loss": 0.5066,
+      "step": 8397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814495000862719e-05,
+      "loss": 1.0919,
+      "step": 8398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8144455228846878e-05,
+      "loss": 1.0647,
+      "step": 8399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143960389839184e-05,
+      "loss": 0.8265,
+      "step": 8400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143465491607705e-05,
+      "loss": 0.9087,
+      "step": 8401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8142970534156043e-05,
+      "loss": 1.2322,
+      "step": 8402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8142475517487792e-05,
+      "loss": 0.7772,
+      "step": 8403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8141980441606555e-05,
+      "loss": 0.826,
+      "step": 8404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8141485306515933e-05,
+      "loss": 1.1345,
+      "step": 8405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140990112219527e-05,
+      "loss": 0.8745,
+      "step": 8406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140494858720935e-05,
+      "loss": 0.9639,
+      "step": 8407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813999954602376e-05,
+      "loss": 1.123,
+      "step": 8408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813950417413161e-05,
+      "loss": 0.8375,
+      "step": 8409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813900874304807e-05,
+      "loss": 1.1058,
+      "step": 8410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8138513252776768e-05,
+      "loss": 1.1574,
+      "step": 8411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8138017703321287e-05,
+      "loss": 1.2753,
+      "step": 8412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137522094685236e-05,
+      "loss": 0.8643,
+      "step": 8413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137026426872225e-05,
+      "loss": 1.083,
+      "step": 8414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136530699885852e-05,
+      "loss": 1.0296,
+      "step": 8415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136034913729724e-05,
+      "loss": 1.0361,
+      "step": 8416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135539068407448e-05,
+      "loss": 0.9642,
+      "step": 8417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8135043163922626e-05,
+      "loss": 1.1087,
+      "step": 8418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813454720027887e-05,
+      "loss": 0.9067,
+      "step": 8419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134051177479778e-05,
+      "loss": 1.234,
+      "step": 8420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8133555095528968e-05,
+      "loss": 1.1241,
+      "step": 8421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813305895443004e-05,
+      "loss": 1.3974,
+      "step": 8422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132562754186606e-05,
+      "loss": 1.5634,
+      "step": 8423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813206649480227e-05,
+      "loss": 0.8499,
+      "step": 8424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131570176280644e-05,
+      "loss": 1.0154,
+      "step": 8425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131073798625336e-05,
+      "loss": 1.251,
+      "step": 8426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813057736183996e-05,
+      "loss": 1.1353,
+      "step": 8427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130080865928118e-05,
+      "loss": 1.1083,
+      "step": 8428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8129584310893428e-05,
+      "loss": 0.8509,
+      "step": 8429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8129087696739497e-05,
+      "loss": 1.0808,
+      "step": 8430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8128591023469938e-05,
+      "loss": 1.051,
+      "step": 8431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812809429108836e-05,
+      "loss": 1.1639,
+      "step": 8432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8127597499598383e-05,
+      "loss": 1.0503,
+      "step": 8433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8127100649003608e-05,
+      "loss": 1.1276,
+      "step": 8434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812660373930766e-05,
+      "loss": 1.1165,
+      "step": 8435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126106770514145e-05,
+      "loss": 0.843,
+      "step": 8436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812560974262668e-05,
+      "loss": 0.9629,
+      "step": 8437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8125112655648878e-05,
+      "loss": 1.2394,
+      "step": 8438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8124615509584352e-05,
+      "loss": 0.9251,
+      "step": 8439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8124118304436724e-05,
+      "loss": 1.1421,
+      "step": 8440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123621040209604e-05,
+      "loss": 0.8013,
+      "step": 8441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812312371690661e-05,
+      "loss": 0.8404,
+      "step": 8442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122626334531356e-05,
+      "loss": 1.0084,
+      "step": 8443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122128893087464e-05,
+      "loss": 1.0548,
+      "step": 8444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8121631392578545e-05,
+      "loss": 1.069,
+      "step": 8445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8121133833008225e-05,
+      "loss": 0.8662,
+      "step": 8446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8120636214380118e-05,
+      "loss": 1.3836,
+      "step": 8447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812013853669784e-05,
+      "loss": 1.0459,
+      "step": 8448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8119640799965015e-05,
+      "loss": 0.8814,
+      "step": 8449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811914300418526e-05,
+      "loss": 0.8513,
+      "step": 8450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8118645149362194e-05,
+      "loss": 1.3449,
+      "step": 8451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811814723549944e-05,
+      "loss": 1.275,
+      "step": 8452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811764926260062e-05,
+      "loss": 1.1453,
+      "step": 8453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8117151230669348e-05,
+      "loss": 0.9387,
+      "step": 8454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116653139709252e-05,
+      "loss": 1.0833,
+      "step": 8455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116154989723957e-05,
+      "loss": 0.7245,
+      "step": 8456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115656780717077e-05,
+      "loss": 0.962,
+      "step": 8457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115158512692242e-05,
+      "loss": 0.7462,
+      "step": 8458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114660185653076e-05,
+      "loss": 0.763,
+      "step": 8459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114161799603195e-05,
+      "loss": 1.0588,
+      "step": 8460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811366335454623e-05,
+      "loss": 1.3061,
+      "step": 8461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8113164850485802e-05,
+      "loss": 1.0426,
+      "step": 8462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811266628742554e-05,
+      "loss": 1.1306,
+      "step": 8463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811216766536907e-05,
+      "loss": 1.0889,
+      "step": 8464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8111668984320014e-05,
+      "loss": 1.0534,
+      "step": 8465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8111170244282003e-05,
+      "loss": 0.7109,
+      "step": 8466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8110671445258657e-05,
+      "loss": 0.9788,
+      "step": 8467
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811017258725361e-05,
+      "loss": 0.8287,
+      "step": 8468
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8109673670270486e-05,
+      "loss": 1.0759,
+      "step": 8469
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8109174694312915e-05,
+      "loss": 0.818,
+      "step": 8470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8108675659384527e-05,
+      "loss": 0.8783,
+      "step": 8471
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8108176565488948e-05,
+      "loss": 1.2581,
+      "step": 8472
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810767741262981e-05,
+      "loss": 1.1879,
+      "step": 8473
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107178200810737e-05,
+      "loss": 0.9086,
+      "step": 8474
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810667893003537e-05,
+      "loss": 1.0801,
+      "step": 8475
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810617960030733e-05,
+      "loss": 0.9684,
+      "step": 8476
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810568021163025e-05,
+      "loss": 1.1371,
+      "step": 8477
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8105180764007766e-05,
+      "loss": 0.8132,
+      "step": 8478
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810468125744351e-05,
+      "loss": 1.1923,
+      "step": 8479
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104181691941106e-05,
+      "loss": 0.6499,
+      "step": 8480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8103682067504194e-05,
+      "loss": 1.1414,
+      "step": 8481
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810318238413641e-05,
+      "loss": 0.9675,
+      "step": 8482
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810268264184138e-05,
+      "loss": 1.0823,
+      "step": 8483
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8102182840622746e-05,
+      "loss": 1.037,
+      "step": 8484
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8101682980484137e-05,
+      "loss": 0.9566,
+      "step": 8485
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810118306142919e-05,
+      "loss": 0.9564,
+      "step": 8486
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8100683083461535e-05,
+      "loss": 1.2478,
+      "step": 8487
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8100183046584817e-05,
+      "loss": 0.9466,
+      "step": 8488
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8099682950802672e-05,
+      "loss": 0.9448,
+      "step": 8489
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8099182796118727e-05,
+      "loss": 1.0891,
+      "step": 8490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8098682582536627e-05,
+      "loss": 0.8394,
+      "step": 8491
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809818231006001e-05,
+      "loss": 1.1199,
+      "step": 8492
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8097681978692512e-05,
+      "loss": 0.8337,
+      "step": 8493
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809718158843777e-05,
+      "loss": 1.0599,
+      "step": 8494
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8096681139299424e-05,
+      "loss": 0.9283,
+      "step": 8495
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8096180631281116e-05,
+      "loss": 1.3924,
+      "step": 8496
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809568006438648e-05,
+      "loss": 1.3051,
+      "step": 8497
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809517943861916e-05,
+      "loss": 1.2887,
+      "step": 8498
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8094678753982798e-05,
+      "loss": 0.9861,
+      "step": 8499
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8094178010481033e-05,
+      "loss": 1.0552,
+      "step": 8500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809367720811751e-05,
+      "loss": 0.8055,
+      "step": 8501
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809317634689586e-05,
+      "loss": 0.8258,
+      "step": 8502
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8092675426819738e-05,
+      "loss": 1.1392,
+      "step": 8503
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8092174447892782e-05,
+      "loss": 0.9768,
+      "step": 8504
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8091673410118633e-05,
+      "loss": 0.8127,
+      "step": 8505
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8091172313500935e-05,
+      "loss": 0.9959,
+      "step": 8506
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8090671158043334e-05,
+      "loss": 1.0624,
+      "step": 8507
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8090169943749477e-05,
+      "loss": 1.1114,
+      "step": 8508
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089668670623e-05,
+      "loss": 0.8717,
+      "step": 8509
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808916733866756e-05,
+      "loss": 0.6994,
+      "step": 8510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088665947886795e-05,
+      "loss": 1.1249,
+      "step": 8511
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088164498284352e-05,
+      "loss": 0.8204,
+      "step": 8512
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808766298986388e-05,
+      "loss": 1.1199,
+      "step": 8513
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8087161422629026e-05,
+      "loss": 1.2596,
+      "step": 8514
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808665979658343e-05,
+      "loss": 0.9058,
+      "step": 8515
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8086158111730752e-05,
+      "loss": 1.2064,
+      "step": 8516
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808565636807463e-05,
+      "loss": 1.3489,
+      "step": 8517
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085154565618717e-05,
+      "loss": 1.1586,
+      "step": 8518
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084652704366668e-05,
+      "loss": 0.9344,
+      "step": 8519
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084150784322123e-05,
+      "loss": 1.1199,
+      "step": 8520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8083648805488734e-05,
+      "loss": 0.882,
+      "step": 8521
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8083146767870158e-05,
+      "loss": 0.9881,
+      "step": 8522
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082644671470037e-05,
+      "loss": 1.0555,
+      "step": 8523
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808214251629203e-05,
+      "loss": 1.1453,
+      "step": 8524
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8081640302339777e-05,
+      "loss": 0.998,
+      "step": 8525
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8081138029616947e-05,
+      "loss": 1.0439,
+      "step": 8526
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808063569812718e-05,
+      "loss": 1.3084,
+      "step": 8527
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8080133307874135e-05,
+      "loss": 1.0678,
+      "step": 8528
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8079630858861462e-05,
+      "loss": 1.3018,
+      "step": 8529
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8079128351092813e-05,
+      "loss": 1.0743,
+      "step": 8530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8078625784571852e-05,
+      "loss": 0.9317,
+      "step": 8531
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8078123159302223e-05,
+      "loss": 0.9252,
+      "step": 8532
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077620475287584e-05,
+      "loss": 1.0544,
+      "step": 8533
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077117732531596e-05,
+      "loss": 0.9385,
+      "step": 8534
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8076614931037908e-05,
+      "loss": 1.1988,
+      "step": 8535
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807611207081018e-05,
+      "loss": 1.0677,
+      "step": 8536
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807560915185207e-05,
+      "loss": 1.1119,
+      "step": 8537
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807510617416723e-05,
+      "loss": 1.2699,
+      "step": 8538
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074603137759324e-05,
+      "loss": 0.7996,
+      "step": 8539
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074100042632005e-05,
+      "loss": 0.9292,
+      "step": 8540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073596888788936e-05,
+      "loss": 1.2106,
+      "step": 8541
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073093676233774e-05,
+      "loss": 1.1107,
+      "step": 8542
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8072590404970175e-05,
+      "loss": 0.8847,
+      "step": 8543
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8072087075001807e-05,
+      "loss": 0.9932,
+      "step": 8544
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8071583686332323e-05,
+      "loss": 0.8027,
+      "step": 8545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8071080238965385e-05,
+      "loss": 0.9841,
+      "step": 8546
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8070576732904657e-05,
+      "loss": 0.9376,
+      "step": 8547
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80700731681538e-05,
+      "loss": 0.7426,
+      "step": 8548
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8069569544716472e-05,
+      "loss": 1.1463,
+      "step": 8549
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8069065862596338e-05,
+      "loss": 1.1768,
+      "step": 8550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806856212179706e-05,
+      "loss": 0.8111,
+      "step": 8551
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8068058322322307e-05,
+      "loss": 1.1439,
+      "step": 8552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8067554464175736e-05,
+      "loss": 1.4005,
+      "step": 8553
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806705054736101e-05,
+      "loss": 0.8898,
+      "step": 8554
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80665465718818e-05,
+      "loss": 0.9846,
+      "step": 8555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8066042537741764e-05,
+      "loss": 1.1002,
+      "step": 8556
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065538444944577e-05,
+      "loss": 0.975,
+      "step": 8557
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065034293493894e-05,
+      "loss": 1.114,
+      "step": 8558
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8064530083393387e-05,
+      "loss": 0.868,
+      "step": 8559
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8064025814646724e-05,
+      "loss": 0.7925,
+      "step": 8560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8063521487257568e-05,
+      "loss": 0.9971,
+      "step": 8561
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8063017101229585e-05,
+      "loss": 1.0449,
+      "step": 8562
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806251265656645e-05,
+      "loss": 1.1606,
+      "step": 8563
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062008153271828e-05,
+      "loss": 1.0189,
+      "step": 8564
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061503591349386e-05,
+      "loss": 0.9027,
+      "step": 8565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8060998970802794e-05,
+      "loss": 1.1732,
+      "step": 8566
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8060494291635725e-05,
+      "loss": 1.1126,
+      "step": 8567
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8059989553851842e-05,
+      "loss": 1.0704,
+      "step": 8568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8059484757454827e-05,
+      "loss": 1.0931,
+      "step": 8569
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8058979902448337e-05,
+      "loss": 1.0134,
+      "step": 8570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8058474988836053e-05,
+      "loss": 0.9919,
+      "step": 8571
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057970016621643e-05,
+      "loss": 1.3596,
+      "step": 8572
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805746498580878e-05,
+      "loss": 1.2645,
+      "step": 8573
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805695989640114e-05,
+      "loss": 1.0954,
+      "step": 8574
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805645474840239e-05,
+      "loss": 1.0109,
+      "step": 8575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8055949541816207e-05,
+      "loss": 1.1769,
+      "step": 8576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8055444276646263e-05,
+      "loss": 1.188,
+      "step": 8577
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054938952896236e-05,
+      "loss": 0.9526,
+      "step": 8578
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8054433570569797e-05,
+      "loss": 0.8039,
+      "step": 8579
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8053928129670624e-05,
+      "loss": 0.909,
+      "step": 8580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8053422630202392e-05,
+      "loss": 0.6695,
+      "step": 8581
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8052917072168774e-05,
+      "loss": 0.6832,
+      "step": 8582
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805241145557345e-05,
+      "loss": 0.7719,
+      "step": 8583
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8051905780420097e-05,
+      "loss": 1.3407,
+      "step": 8584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8051400046712388e-05,
+      "loss": 0.9316,
+      "step": 8585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8050894254454007e-05,
+      "loss": 1.0869,
+      "step": 8586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8050388403648624e-05,
+      "loss": 1.1988,
+      "step": 8587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8049882494299926e-05,
+      "loss": 1.2558,
+      "step": 8588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8049376526411588e-05,
+      "loss": 0.9305,
+      "step": 8589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048870499987293e-05,
+      "loss": 1.0963,
+      "step": 8590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048364415030714e-05,
+      "loss": 0.9715,
+      "step": 8591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8047858271545534e-05,
+      "loss": 0.8964,
+      "step": 8592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8047352069535437e-05,
+      "loss": 1.123,
+      "step": 8593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8046845809004107e-05,
+      "loss": 1.1614,
+      "step": 8594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8046339489955214e-05,
+      "loss": 1.0083,
+      "step": 8595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804583311239245e-05,
+      "loss": 1.0302,
+      "step": 8596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045326676319494e-05,
+      "loss": 0.8342,
+      "step": 8597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044820181740025e-05,
+      "loss": 0.9449,
+      "step": 8598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8044313628657735e-05,
+      "loss": 1.0504,
+      "step": 8599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80438070170763e-05,
+      "loss": 1.2996,
+      "step": 8600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804330034699941e-05,
+      "loss": 1.2388,
+      "step": 8601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8042793618430746e-05,
+      "loss": 0.8464,
+      "step": 8602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8042286831373988e-05,
+      "loss": 1.2367,
+      "step": 8603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8041779985832834e-05,
+      "loss": 1.1911,
+      "step": 8604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804127308181096e-05,
+      "loss": 0.6302,
+      "step": 8605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804076611931206e-05,
+      "loss": 1.2585,
+      "step": 8606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804025909833981e-05,
+      "loss": 0.842,
+      "step": 8607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039752018897903e-05,
+      "loss": 1.2539,
+      "step": 8608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803924488099003e-05,
+      "loss": 1.0694,
+      "step": 8609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8038737684619874e-05,
+      "loss": 0.9585,
+      "step": 8610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8038230429791123e-05,
+      "loss": 1.0447,
+      "step": 8611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803772311650747e-05,
+      "loss": 0.8971,
+      "step": 8612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8037215744772603e-05,
+      "loss": 0.8242,
+      "step": 8613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8036708314590207e-05,
+      "loss": 0.6531,
+      "step": 8614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803620082596398e-05,
+      "loss": 1.5725,
+      "step": 8615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035693278897607e-05,
+      "loss": 1.517,
+      "step": 8616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035185673394776e-05,
+      "loss": 1.0591,
+      "step": 8617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034678009459185e-05,
+      "loss": 1.0063,
+      "step": 8618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034170287094525e-05,
+      "loss": 1.1367,
+      "step": 8619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8033662506304487e-05,
+      "loss": 0.8396,
+      "step": 8620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803315466709276e-05,
+      "loss": 0.7867,
+      "step": 8621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8032646769463042e-05,
+      "loss": 1.2244,
+      "step": 8622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8032138813419024e-05,
+      "loss": 0.7569,
+      "step": 8623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80316307989644e-05,
+      "loss": 0.7622,
+      "step": 8624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031122726102868e-05,
+      "loss": 0.9354,
+      "step": 8625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030614594838115e-05,
+      "loss": 0.9464,
+      "step": 8626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030106405173845e-05,
+      "loss": 0.8885,
+      "step": 8627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029598157113748e-05,
+      "loss": 1.0429,
+      "step": 8628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029089850661524e-05,
+      "loss": 0.8651,
+      "step": 8629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8028581485820863e-05,
+      "loss": 1.0252,
+      "step": 8630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8028073062595468e-05,
+      "loss": 0.9425,
+      "step": 8631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802756458098903e-05,
+      "loss": 1.0972,
+      "step": 8632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8027056041005257e-05,
+      "loss": 1.1162,
+      "step": 8633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8026547442647837e-05,
+      "loss": 1.0411,
+      "step": 8634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8026038785920476e-05,
+      "loss": 1.2441,
+      "step": 8635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025530070826866e-05,
+      "loss": 0.9769,
+      "step": 8636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802502129737071e-05,
+      "loss": 1.1062,
+      "step": 8637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802451246555571e-05,
+      "loss": 1.0911,
+      "step": 8638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024003575385568e-05,
+      "loss": 0.8275,
+      "step": 8639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8023494626863976e-05,
+      "loss": 0.9261,
+      "step": 8640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802298561999464e-05,
+      "loss": 1.1719,
+      "step": 8641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8022476554781264e-05,
+      "loss": 1.1314,
+      "step": 8642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8021967431227547e-05,
+      "loss": 0.9806,
+      "step": 8643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802145824933719e-05,
+      "loss": 0.948,
+      "step": 8644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80209490091139e-05,
+      "loss": 1.7517,
+      "step": 8645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802043971056138e-05,
+      "loss": 1.6588,
+      "step": 8646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801993035368333e-05,
+      "loss": 0.9981,
+      "step": 8647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8019420938483453e-05,
+      "loss": 1.1145,
+      "step": 8648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801891146496546e-05,
+      "loss": 1.2329,
+      "step": 8649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801840193313305e-05,
+      "loss": 1.0293,
+      "step": 8650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801789234298993e-05,
+      "loss": 1.3137,
+      "step": 8651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801738269453981e-05,
+      "loss": 1.1042,
+      "step": 8652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801687298778639e-05,
+      "loss": 1.1284,
+      "step": 8653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8016363222733383e-05,
+      "loss": 0.73,
+      "step": 8654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015853399384488e-05,
+      "loss": 0.9546,
+      "step": 8655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015343517743418e-05,
+      "loss": 1.197,
+      "step": 8656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801483357781388e-05,
+      "loss": 0.9484,
+      "step": 8657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8014323579599585e-05,
+      "loss": 1.0096,
+      "step": 8658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013813523104234e-05,
+      "loss": 0.8599,
+      "step": 8659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013303408331545e-05,
+      "loss": 1.281,
+      "step": 8660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012793235285223e-05,
+      "loss": 1.079,
+      "step": 8661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012283003968977e-05,
+      "loss": 1.7133,
+      "step": 8662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011772714386523e-05,
+      "loss": 1.3784,
+      "step": 8663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011262366541566e-05,
+      "loss": 0.9851,
+      "step": 8664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8010751960437818e-05,
+      "loss": 0.9901,
+      "step": 8665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8010241496078992e-05,
+      "loss": 0.9854,
+      "step": 8666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8009730973468803e-05,
+      "loss": 0.7823,
+      "step": 8667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8009220392610957e-05,
+      "loss": 1.3174,
+      "step": 8668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008709753509174e-05,
+      "loss": 1.097,
+      "step": 8669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008199056167167e-05,
+      "loss": 1.2348,
+      "step": 8670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8007688300588643e-05,
+      "loss": 0.884,
+      "step": 8671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8007177486777322e-05,
+      "loss": 0.9373,
+      "step": 8672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8006666614736915e-05,
+      "loss": 1.7109,
+      "step": 8673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800615568447114e-05,
+      "loss": 0.7631,
+      "step": 8674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8005644695983713e-05,
+      "loss": 1.2279,
+      "step": 8675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800513364927835e-05,
+      "loss": 0.9408,
+      "step": 8676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800462254435876e-05,
+      "loss": 1.2316,
+      "step": 8677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8004111381228675e-05,
+      "loss": 1.0094,
+      "step": 8678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80036001598918e-05,
+      "loss": 1.0118,
+      "step": 8679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8003088880351853e-05,
+      "loss": 1.1949,
+      "step": 8680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8002577542612555e-05,
+      "loss": 0.7647,
+      "step": 8681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8002066146677624e-05,
+      "loss": 0.9963,
+      "step": 8682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800155469255078e-05,
+      "loss": 0.9185,
+      "step": 8683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8001043180235747e-05,
+      "loss": 1.0425,
+      "step": 8684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8000531609736236e-05,
+      "loss": 0.8722,
+      "step": 8685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800001998105597e-05,
+      "loss": 1.0267,
+      "step": 8686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799950829419867e-05,
+      "loss": 1.0232,
+      "step": 8687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799899654916806e-05,
+      "loss": 0.9496,
+      "step": 8688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998484745967856e-05,
+      "loss": 1.0571,
+      "step": 8689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7997972884601782e-05,
+      "loss": 0.6548,
+      "step": 8690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7997460965073564e-05,
+      "loss": 0.9921,
+      "step": 8691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996948987386918e-05,
+      "loss": 1.1625,
+      "step": 8692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996436951545572e-05,
+      "loss": 1.0694,
+      "step": 8693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799592485755325e-05,
+      "loss": 1.1216,
+      "step": 8694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7995412705413677e-05,
+      "loss": 0.9201,
+      "step": 8695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799490049513057e-05,
+      "loss": 0.9865,
+      "step": 8696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7994388226707662e-05,
+      "loss": 1.2407,
+      "step": 8697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7993875900148675e-05,
+      "loss": 1.1707,
+      "step": 8698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7993363515457335e-05,
+      "loss": 0.8228,
+      "step": 8699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7992851072637366e-05,
+      "loss": 1.2667,
+      "step": 8700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7992338571692495e-05,
+      "loss": 1.5579,
+      "step": 8701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7991826012626453e-05,
+      "loss": 0.9684,
+      "step": 8702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7991313395442963e-05,
+      "loss": 1.2311,
+      "step": 8703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799080072014576e-05,
+      "loss": 1.221,
+      "step": 8704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799028798673856e-05,
+      "loss": 1.0937,
+      "step": 8705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7989775195225102e-05,
+      "loss": 1.1042,
+      "step": 8706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7989262345609107e-05,
+      "loss": 0.8419,
+      "step": 8707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7988749437894313e-05,
+      "loss": 0.6117,
+      "step": 8708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798823647208445e-05,
+      "loss": 0.8748,
+      "step": 8709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7987723448183235e-05,
+      "loss": 0.9455,
+      "step": 8710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7987210366194414e-05,
+      "loss": 0.9805,
+      "step": 8711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7986697226121714e-05,
+      "loss": 0.8424,
+      "step": 8712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798618402796886e-05,
+      "loss": 1.1359,
+      "step": 8713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7985670771739593e-05,
+      "loss": 1.223,
+      "step": 8714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798515745743764e-05,
+      "loss": 1.9117,
+      "step": 8715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798464408506673e-05,
+      "loss": 1.189,
+      "step": 8716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7984130654630607e-05,
+      "loss": 1.0328,
+      "step": 8717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983617166133e-05,
+      "loss": 0.703,
+      "step": 8718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798310361957764e-05,
+      "loss": 0.8827,
+      "step": 8719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7982590014968264e-05,
+      "loss": 0.848,
+      "step": 8720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798207635230861e-05,
+      "loss": 0.7685,
+      "step": 8721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7981562631602406e-05,
+      "loss": 0.9293,
+      "step": 8722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7981048852853393e-05,
+      "loss": 0.9871,
+      "step": 8723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798053501606531e-05,
+      "loss": 1.0315,
+      "step": 8724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7980021121241885e-05,
+      "loss": 1.5484,
+      "step": 8725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979507168386866e-05,
+      "loss": 1.2653,
+      "step": 8726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797899315750398e-05,
+      "loss": 0.9884,
+      "step": 8727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7978479088596973e-05,
+      "loss": 1.3108,
+      "step": 8728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7977964961669578e-05,
+      "loss": 1.0567,
+      "step": 8729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797745077672554e-05,
+      "loss": 0.9323,
+      "step": 8730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797693653376859e-05,
+      "loss": 0.9155,
+      "step": 8731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7976422232802472e-05,
+      "loss": 1.2183,
+      "step": 8732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7975907873830923e-05,
+      "loss": 0.9582,
+      "step": 8733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797539345685769e-05,
+      "loss": 1.3468,
+      "step": 8734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797487898188651e-05,
+      "loss": 1.0051,
+      "step": 8735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7974364448921123e-05,
+      "loss": 1.0282,
+      "step": 8736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973849857965274e-05,
+      "loss": 1.2759,
+      "step": 8737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973335209022702e-05,
+      "loss": 0.9985,
+      "step": 8738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972820502097153e-05,
+      "loss": 0.8243,
+      "step": 8739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972305737192365e-05,
+      "loss": 0.8938,
+      "step": 8740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971790914312085e-05,
+      "loss": 1.0407,
+      "step": 8741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971276033460057e-05,
+      "loss": 1.1965,
+      "step": 8742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970761094640025e-05,
+      "loss": 1.7051,
+      "step": 8743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970246097855733e-05,
+      "loss": 1.2073,
+      "step": 8744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969731043110928e-05,
+      "loss": 0.8053,
+      "step": 8745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796921593040935e-05,
+      "loss": 1.1752,
+      "step": 8746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796870075975475e-05,
+      "loss": 0.8926,
+      "step": 8747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968185531150876e-05,
+      "loss": 0.8464,
+      "step": 8748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796767024460147e-05,
+      "loss": 0.7493,
+      "step": 8749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796715490011028e-05,
+      "loss": 1.2377,
+      "step": 8750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7966639497681054e-05,
+      "loss": 0.9889,
+      "step": 8751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7966124037317544e-05,
+      "loss": 0.9532,
+      "step": 8752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965608519023496e-05,
+      "loss": 0.9432,
+      "step": 8753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965092942802653e-05,
+      "loss": 0.9072,
+      "step": 8754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7964577308658773e-05,
+      "loss": 0.8351,
+      "step": 8755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7964061616595604e-05,
+      "loss": 0.9611,
+      "step": 8756
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7963545866616893e-05,
+      "loss": 1.3491,
+      "step": 8757
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796303005872639e-05,
+      "loss": 1.2567,
+      "step": 8758
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7962514192927852e-05,
+      "loss": 1.1337,
+      "step": 8759
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7961998269225024e-05,
+      "loss": 1.1744,
+      "step": 8760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796148228762166e-05,
+      "loss": 0.9046,
+      "step": 8761
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7960966248121514e-05,
+      "loss": 0.7455,
+      "step": 8762
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7960450150728336e-05,
+      "loss": 1.2011,
+      "step": 8763
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7959933995445884e-05,
+      "loss": 0.9177,
+      "step": 8764
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7959417782277905e-05,
+      "loss": 0.8383,
+      "step": 8765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958901511228156e-05,
+      "loss": 1.1124,
+      "step": 8766
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958385182300392e-05,
+      "loss": 0.8573,
+      "step": 8767
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957868795498367e-05,
+      "loss": 0.9717,
+      "step": 8768
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957352350825838e-05,
+      "loss": 1.0085,
+      "step": 8769
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795683584828656e-05,
+      "loss": 0.8736,
+      "step": 8770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7956319287884284e-05,
+      "loss": 0.9089,
+      "step": 8771
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795580266962277e-05,
+      "loss": 1.1313,
+      "step": 8772
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795528599350578e-05,
+      "loss": 1.1081,
+      "step": 8773
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954769259537066e-05,
+      "loss": 1.2448,
+      "step": 8774
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954252467720386e-05,
+      "loss": 1.0812,
+      "step": 8775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79537356180595e-05,
+      "loss": 0.9635,
+      "step": 8776
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795321871055816e-05,
+      "loss": 0.9475,
+      "step": 8777
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7952701745220137e-05,
+      "loss": 0.7754,
+      "step": 8778
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795218472204918e-05,
+      "loss": 0.8248,
+      "step": 8779
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951667641049052e-05,
+      "loss": 0.9431,
+      "step": 8780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795115050222352e-05,
+      "loss": 1.048,
+      "step": 8781
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7950633305576332e-05,
+      "loss": 1.1888,
+      "step": 8782
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7950116051111255e-05,
+      "loss": 0.9437,
+      "step": 8783
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7949598738832054e-05,
+      "loss": 1.0489,
+      "step": 8784
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794908136874249e-05,
+      "loss": 1.0578,
+      "step": 8785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794856394084632e-05,
+      "loss": 0.9113,
+      "step": 8786
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794804645514731e-05,
+      "loss": 0.9801,
+      "step": 8787
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794752891164923e-05,
+      "loss": 1.2191,
+      "step": 8788
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794701131035583e-05,
+      "loss": 0.9184,
+      "step": 8789
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7946493651270883e-05,
+      "loss": 0.9644,
+      "step": 8790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945975934398154e-05,
+      "loss": 0.7511,
+      "step": 8791
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945458159741406e-05,
+      "loss": 0.5233,
+      "step": 8792
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79449403273044e-05,
+      "loss": 0.8877,
+      "step": 8793
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794442243709091e-05,
+      "loss": 1.0145,
+      "step": 8794
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794390448910469e-05,
+      "loss": 1.2301,
+      "step": 8795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7943386483349526e-05,
+      "loss": 1.3307,
+      "step": 8796
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7942868419829166e-05,
+      "loss": 0.8148,
+      "step": 8797
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7942350298547387e-05,
+      "loss": 1.1752,
+      "step": 8798
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7941832119507952e-05,
+      "loss": 1.0136,
+      "step": 8799
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7941313882714636e-05,
+      "loss": 1.1875,
+      "step": 8800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7940795588171204e-05,
+      "loss": 0.9386,
+      "step": 8801
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794027723588142e-05,
+      "loss": 0.9541,
+      "step": 8802
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7939758825849062e-05,
+      "loss": 0.8036,
+      "step": 8803
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7939240358077895e-05,
+      "loss": 1.1775,
+      "step": 8804
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7938721832571688e-05,
+      "loss": 1.6711,
+      "step": 8805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7938203249334216e-05,
+      "loss": 1.3021,
+      "step": 8806
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7937684608369247e-05,
+      "loss": 1.1044,
+      "step": 8807
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7937165909680557e-05,
+      "loss": 0.9512,
+      "step": 8808
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936647153271915e-05,
+      "loss": 1.0852,
+      "step": 8809
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936128339147092e-05,
+      "loss": 1.2165,
+      "step": 8810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7935609467309865e-05,
+      "loss": 1.1325,
+      "step": 8811
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7935090537764003e-05,
+      "loss": 1.1842,
+      "step": 8812
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793457155051328e-05,
+      "loss": 0.8101,
+      "step": 8813
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793405250556147e-05,
+      "loss": 0.9257,
+      "step": 8814
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7933533402912354e-05,
+      "loss": 1.1629,
+      "step": 8815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7933014242569698e-05,
+      "loss": 1.0234,
+      "step": 8816
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7932495024537285e-05,
+      "loss": 1.0913,
+      "step": 8817
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7931975748818886e-05,
+      "loss": 0.989,
+      "step": 8818
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793145641541828e-05,
+      "loss": 1.0153,
+      "step": 8819
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793093702433924e-05,
+      "loss": 1.1366,
+      "step": 8820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7930417575585542e-05,
+      "loss": 1.2076,
+      "step": 8821
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7929898069160972e-05,
+      "loss": 0.7364,
+      "step": 8822
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7929378505069302e-05,
+      "loss": 1.0962,
+      "step": 8823
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928858883314308e-05,
+      "loss": 0.9095,
+      "step": 8824
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7928339203899776e-05,
+      "loss": 1.0334,
+      "step": 8825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7927819466829476e-05,
+      "loss": 1.0771,
+      "step": 8826
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7927299672107197e-05,
+      "loss": 1.3631,
+      "step": 8827
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7926779819736713e-05,
+      "loss": 0.9861,
+      "step": 8828
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7926259909721804e-05,
+      "loss": 1.1198,
+      "step": 8829
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7925739942066253e-05,
+      "loss": 1.0071,
+      "step": 8830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7925219916773843e-05,
+      "loss": 1.1334,
+      "step": 8831
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924699833848353e-05,
+      "loss": 1.0351,
+      "step": 8832
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924179693293567e-05,
+      "loss": 0.9796,
+      "step": 8833
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923659495113262e-05,
+      "loss": 1.1815,
+      "step": 8834
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792313923931123e-05,
+      "loss": 1.0259,
+      "step": 8835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792261892589125e-05,
+      "loss": 1.3086,
+      "step": 8836
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79220985548571e-05,
+      "loss": 1.0792,
+      "step": 8837
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921578126212576e-05,
+      "loss": 0.9995,
+      "step": 8838
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792105763996145e-05,
+      "loss": 1.0317,
+      "step": 8839
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7920537096107516e-05,
+      "loss": 1.3383,
+      "step": 8840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7920016494654558e-05,
+      "loss": 1.0237,
+      "step": 8841
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791949583560636e-05,
+      "loss": 1.0444,
+      "step": 8842
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791897511896671e-05,
+      "loss": 1.0157,
+      "step": 8843
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791845434473939e-05,
+      "loss": 0.8763,
+      "step": 8844
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791793351292819e-05,
+      "loss": 1.2524,
+      "step": 8845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7917412623536906e-05,
+      "loss": 1.3461,
+      "step": 8846
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791689167656931e-05,
+      "loss": 0.9402,
+      "step": 8847
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7916370672029203e-05,
+      "loss": 1.0443,
+      "step": 8848
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7915849609920368e-05,
+      "loss": 1.0228,
+      "step": 8849
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7915328490246594e-05,
+      "loss": 1.062,
+      "step": 8850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914807313011674e-05,
+      "loss": 0.8586,
+      "step": 8851
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914286078219397e-05,
+      "loss": 1.1148,
+      "step": 8852
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791376478587355e-05,
+      "loss": 1.0013,
+      "step": 8853
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791324343597793e-05,
+      "loss": 0.9056,
+      "step": 8854
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912722028536325e-05,
+      "loss": 1.1934,
+      "step": 8855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912200563552522e-05,
+      "loss": 1.3508,
+      "step": 8856
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911679041030323e-05,
+      "loss": 0.9656,
+      "step": 8857
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911157460973515e-05,
+      "loss": 1.3161,
+      "step": 8858
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910635823385888e-05,
+      "loss": 1.0949,
+      "step": 8859
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7910114128271242e-05,
+      "loss": 1.0232,
+      "step": 8860
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7909592375633366e-05,
+      "loss": 1.116,
+      "step": 8861
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7909070565476057e-05,
+      "loss": 0.7269,
+      "step": 8862
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7908548697803104e-05,
+      "loss": 1.0024,
+      "step": 8863
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7908026772618314e-05,
+      "loss": 0.9623,
+      "step": 8864
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7907504789925473e-05,
+      "loss": 0.9538,
+      "step": 8865
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7906982749728374e-05,
+      "loss": 1.1733,
+      "step": 8866
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7906460652030824e-05,
+      "loss": 0.952,
+      "step": 8867
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7905938496836613e-05,
+      "loss": 1.3619,
+      "step": 8868
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790541628414954e-05,
+      "loss": 0.9936,
+      "step": 8869
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79048940139734e-05,
+      "loss": 0.7675,
+      "step": 8870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790437168631199e-05,
+      "loss": 1.076,
+      "step": 8871
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903849301169117e-05,
+      "loss": 1.1406,
+      "step": 8872
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903326858548573e-05,
+      "loss": 1.0941,
+      "step": 8873
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7902804358454157e-05,
+      "loss": 0.7822,
+      "step": 8874
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790228180088967e-05,
+      "loss": 0.8845,
+      "step": 8875
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790175918585891e-05,
+      "loss": 0.8057,
+      "step": 8876
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7901236513365686e-05,
+      "loss": 1.3517,
+      "step": 8877
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790071378341379e-05,
+      "loss": 1.0857,
+      "step": 8878
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900190996007026e-05,
+      "loss": 1.2328,
+      "step": 8879
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789966815114919e-05,
+      "loss": 0.7711,
+      "step": 8880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.78991452488441e-05,
+      "loss": 0.793,
+      "step": 8881
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789862228909554e-05,
+      "loss": 1.0493,
+      "step": 8882
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789809927190733e-05,
+      "loss": 0.9528,
+      "step": 8883
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7897576197283254e-05,
+      "loss": 0.983,
+      "step": 8884
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7897053065227134e-05,
+      "loss": 0.7911,
+      "step": 8885
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896529875742766e-05,
+      "loss": 1.0453,
+      "step": 8886
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896006628833956e-05,
+      "loss": 1.1568,
+      "step": 8887
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789548332450451e-05,
+      "loss": 0.7862,
+      "step": 8888
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894959962758233e-05,
+      "loss": 1.1529,
+      "step": 8889
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894436543598926e-05,
+      "loss": 0.8745,
+      "step": 8890
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7893913067030407e-05,
+      "loss": 0.7206,
+      "step": 8891
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789338953305647e-05,
+      "loss": 0.7257,
+      "step": 8892
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7892865941680928e-05,
+      "loss": 1.5071,
+      "step": 8893
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789234229290759e-05,
+      "loss": 0.9478,
+      "step": 8894
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789181858674026e-05,
+      "loss": 1.2838,
+      "step": 8895
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7891294823182752e-05,
+      "loss": 1.6111,
+      "step": 8896
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789077100223887e-05,
+      "loss": 1.1575,
+      "step": 8897
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7890247123912425e-05,
+      "loss": 0.8271,
+      "step": 8898
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788972318820723e-05,
+      "loss": 1.1621,
+      "step": 8899
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7889199195127086e-05,
+      "loss": 0.8395,
+      "step": 8900
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7888675144675815e-05,
+      "loss": 1.0919,
+      "step": 8901
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788815103685722e-05,
+      "loss": 0.928,
+      "step": 8902
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7887626871675113e-05,
+      "loss": 1.3363,
+      "step": 8903
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788710264913331e-05,
+      "loss": 1.2088,
+      "step": 8904
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788657836923562e-05,
+      "loss": 1.1919,
+      "step": 8905
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7886054031985854e-05,
+      "loss": 1.0035,
+      "step": 8906
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788552963738783e-05,
+      "loss": 0.9676,
+      "step": 8907
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7885005185445358e-05,
+      "loss": 1.0752,
+      "step": 8908
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7884480676162254e-05,
+      "loss": 1.0675,
+      "step": 8909
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788395610954233e-05,
+      "loss": 1.025,
+      "step": 8910
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7883431485589403e-05,
+      "loss": 0.8936,
+      "step": 8911
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788290680430728e-05,
+      "loss": 0.9334,
+      "step": 8912
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7882382065699793e-05,
+      "loss": 1.1592,
+      "step": 8913
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7881857269770742e-05,
+      "loss": 1.2544,
+      "step": 8914
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788133241652395e-05,
+      "loss": 1.3914,
+      "step": 8915
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880807505963238e-05,
+      "loss": 1.2849,
+      "step": 8916
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880282538092414e-05,
+      "loss": 0.931,
+      "step": 8917
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7879757512915302e-05,
+      "loss": 0.8735,
+      "step": 8918
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7879232430435717e-05,
+      "loss": 0.989,
+      "step": 8919
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7878707290657483e-05,
+      "loss": 0.8875,
+      "step": 8920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787818209358441e-05,
+      "loss": 0.953,
+      "step": 8921
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7877656839220325e-05,
+      "loss": 1.3986,
+      "step": 8922
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7877131527569044e-05,
+      "loss": 0.9122,
+      "step": 8923
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7876606158634388e-05,
+      "loss": 0.6802,
+      "step": 8924
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7876080732420176e-05,
+      "loss": 0.9646,
+      "step": 8925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7875555248930233e-05,
+      "loss": 1.1425,
+      "step": 8926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7875029708168374e-05,
+      "loss": 1.0795,
+      "step": 8927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7874504110138424e-05,
+      "loss": 1.2847,
+      "step": 8928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7873978454844206e-05,
+      "loss": 1.0297,
+      "step": 8929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7873452742289544e-05,
+      "loss": 1.1549,
+      "step": 8930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7872926972478263e-05,
+      "loss": 1.0353,
+      "step": 8931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787240114541418e-05,
+      "loss": 0.9995,
+      "step": 8932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787187526110112e-05,
+      "loss": 1.0949,
+      "step": 8933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787134931954291e-05,
+      "loss": 1.0907,
+      "step": 8934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7870823320743372e-05,
+      "loss": 1.3265,
+      "step": 8935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7870297264706337e-05,
+      "loss": 1.1117,
+      "step": 8936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869771151435624e-05,
+      "loss": 1.0187,
+      "step": 8937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869244980935055e-05,
+      "loss": 1.0045,
+      "step": 8938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7868718753208473e-05,
+      "loss": 1.0206,
+      "step": 8939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7868192468259686e-05,
+      "loss": 1.2142,
+      "step": 8940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7867666126092535e-05,
+      "loss": 1.0591,
+      "step": 8941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786713972671084e-05,
+      "loss": 1.2432,
+      "step": 8942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7866613270118432e-05,
+      "loss": 1.1128,
+      "step": 8943
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7866086756319137e-05,
+      "loss": 0.9259,
+      "step": 8944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786556018531679e-05,
+      "loss": 1.0121,
+      "step": 8945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7865033557115216e-05,
+      "loss": 1.4446,
+      "step": 8946
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786450687171824e-05,
+      "loss": 1.1015,
+      "step": 8947
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7863980129129697e-05,
+      "loss": 1.1881,
+      "step": 8948
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786345332935342e-05,
+      "loss": 1.2798,
+      "step": 8949
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7862926472393235e-05,
+      "loss": 1.2813,
+      "step": 8950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786239955825298e-05,
+      "loss": 1.4293,
+      "step": 8951
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7861872586936474e-05,
+      "loss": 0.8653,
+      "step": 8952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7861345558447563e-05,
+      "loss": 0.8755,
+      "step": 8953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860818472790074e-05,
+      "loss": 0.898,
+      "step": 8954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860291329967842e-05,
+      "loss": 0.8838,
+      "step": 8955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859764129984693e-05,
+      "loss": 0.9547,
+      "step": 8956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785923687284447e-05,
+      "loss": 1.0358,
+      "step": 8957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858709558551005e-05,
+      "loss": 0.7146,
+      "step": 8958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858182187108127e-05,
+      "loss": 1.0281,
+      "step": 8959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785765475851968e-05,
+      "loss": 1.0176,
+      "step": 8960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7857127272789494e-05,
+      "loss": 0.8571,
+      "step": 8961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78565997299214e-05,
+      "loss": 1.0812,
+      "step": 8962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785607212991925e-05,
+      "loss": 1.2614,
+      "step": 8963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7855544472786865e-05,
+      "loss": 0.8752,
+      "step": 8964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785501675852809e-05,
+      "loss": 1.15,
+      "step": 8965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785448898714676e-05,
+      "loss": 0.9679,
+      "step": 8966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7853961158646715e-05,
+      "loss": 0.9114,
+      "step": 8967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785343327303179e-05,
+      "loss": 1.173,
+      "step": 8968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785290533030583e-05,
+      "loss": 1.1565,
+      "step": 8969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7852377330472668e-05,
+      "loss": 1.2191,
+      "step": 8970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851849273536146e-05,
+      "loss": 1.3685,
+      "step": 8971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851321159500106e-05,
+      "loss": 0.8282,
+      "step": 8972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7850792988368384e-05,
+      "loss": 0.8792,
+      "step": 8973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7850264760144825e-05,
+      "loss": 1.259,
+      "step": 8974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7849736474833268e-05,
+      "loss": 1.0073,
+      "step": 8975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7849208132437556e-05,
+      "loss": 1.0443,
+      "step": 8976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7848679732961533e-05,
+      "loss": 1.1133,
+      "step": 8977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784815127640904e-05,
+      "loss": 1.1064,
+      "step": 8978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7847622762783915e-05,
+      "loss": 1.1827,
+      "step": 8979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7847094192090007e-05,
+      "loss": 1.01,
+      "step": 8980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7846565564331158e-05,
+      "loss": 1.1339,
+      "step": 8981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7846036879511216e-05,
+      "loss": 0.7897,
+      "step": 8982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784550813763402e-05,
+      "loss": 1.1886,
+      "step": 8983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784497933870342e-05,
+      "loss": 0.985,
+      "step": 8984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7844450482723258e-05,
+      "loss": 1.1026,
+      "step": 8985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7843921569697377e-05,
+      "loss": 0.9254,
+      "step": 8986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7843392599629634e-05,
+      "loss": 0.9283,
+      "step": 8987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7842863572523864e-05,
+      "loss": 1.2672,
+      "step": 8988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784233448838392e-05,
+      "loss": 0.9296,
+      "step": 8989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784180534721365e-05,
+      "loss": 1.0874,
+      "step": 8990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78412761490169e-05,
+      "loss": 0.8923,
+      "step": 8991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784074689379752e-05,
+      "loss": 0.9123,
+      "step": 8992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784021758155936e-05,
+      "loss": 1.2081,
+      "step": 8993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839688212306262e-05,
+      "loss": 1.2666,
+      "step": 8994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839158786042084e-05,
+      "loss": 0.7713,
+      "step": 8995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783862930277067e-05,
+      "loss": 0.9876,
+      "step": 8996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7838099762495876e-05,
+      "loss": 0.8476,
+      "step": 8997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783757016522155e-05,
+      "loss": 0.9409,
+      "step": 8998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837040510951543e-05,
+      "loss": 0.8665,
+      "step": 8999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783651079968971e-05,
+      "loss": 1.0362,
+      "step": 9000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835981031439896e-05,
+      "loss": 1.0812,
+      "step": 9001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835451206205954e-05,
+      "loss": 1.3796,
+      "step": 9002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783492132399175e-05,
+      "loss": 0.936,
+      "step": 9003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7834391384801124e-05,
+      "loss": 1.2396,
+      "step": 9004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833861388637932e-05,
+      "loss": 1.3786,
+      "step": 9005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833331335506036e-05,
+      "loss": 1.2415,
+      "step": 9006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832801225409276e-05,
+      "loss": 1.2601,
+      "step": 9007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783227105835152e-05,
+      "loss": 1.2612,
+      "step": 9008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783174083433662e-05,
+      "loss": 0.9272,
+      "step": 9009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783121055336843e-05,
+      "loss": 1.0085,
+      "step": 9010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830680215450806e-05,
+      "loss": 0.8768,
+      "step": 9011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830149820587608e-05,
+      "loss": 0.9185,
+      "step": 9012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782961936878269e-05,
+      "loss": 0.9788,
+      "step": 9013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7829088860039913e-05,
+      "loss": 0.8766,
+      "step": 9014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782855829436313e-05,
+      "loss": 0.8824,
+      "step": 9015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78280276717562e-05,
+      "loss": 1.1415,
+      "step": 9016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7827496992222984e-05,
+      "loss": 0.8826,
+      "step": 9017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7826966255767346e-05,
+      "loss": 1.0396,
+      "step": 9018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7826435462393136e-05,
+      "loss": 1.0799,
+      "step": 9019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7825904612104218e-05,
+      "loss": 1.1461,
+      "step": 9020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782537370490445e-05,
+      "loss": 0.8653,
+      "step": 9021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78248427407977e-05,
+      "loss": 1.301,
+      "step": 9022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7824311719787824e-05,
+      "loss": 0.9308,
+      "step": 9023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7823780641878683e-05,
+      "loss": 1.2345,
+      "step": 9024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782324950707414e-05,
+      "loss": 1.1611,
+      "step": 9025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782271831537806e-05,
+      "loss": 0.9148,
+      "step": 9026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7822187066794302e-05,
+      "loss": 1.0457,
+      "step": 9027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782165576132673e-05,
+      "loss": 0.872,
+      "step": 9028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782112439897921e-05,
+      "loss": 0.7636,
+      "step": 9029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7820592979755605e-05,
+      "loss": 0.765,
+      "step": 9030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782006150365978e-05,
+      "loss": 0.9949,
+      "step": 9031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78195299706956e-05,
+      "loss": 1.1369,
+      "step": 9032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7818998380866926e-05,
+      "loss": 0.8593,
+      "step": 9033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7818466734177634e-05,
+      "loss": 0.8789,
+      "step": 9034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7817935030631578e-05,
+      "loss": 0.9072,
+      "step": 9035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7817403270232633e-05,
+      "loss": 0.8953,
+      "step": 9036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781687145298466e-05,
+      "loss": 0.7954,
+      "step": 9037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7816339578891533e-05,
+      "loss": 0.8867,
+      "step": 9038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7815807647957117e-05,
+      "loss": 1.1114,
+      "step": 9039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7815275660185275e-05,
+      "loss": 0.9359,
+      "step": 9040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814743615579885e-05,
+      "loss": 1.0833,
+      "step": 9041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781421151414481e-05,
+      "loss": 0.9883,
+      "step": 9042
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7813679355883923e-05,
+      "loss": 0.7423,
+      "step": 9043
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781314714080109e-05,
+      "loss": 1.2377,
+      "step": 9044
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7812614868900185e-05,
+      "loss": 1.0911,
+      "step": 9045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7812082540185077e-05,
+      "loss": 1.0421,
+      "step": 9046
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7811550154659635e-05,
+      "loss": 0.905,
+      "step": 9047
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7811017712327735e-05,
+      "loss": 1.1884,
+      "step": 9048
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7810485213193243e-05,
+      "loss": 1.1451,
+      "step": 9049
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780995265726004e-05,
+      "loss": 0.8834,
+      "step": 9050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7809420044531992e-05,
+      "loss": 0.7917,
+      "step": 9051
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7808887375012974e-05,
+      "loss": 0.5363,
+      "step": 9052
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7808354648706862e-05,
+      "loss": 0.9215,
+      "step": 9053
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807821865617528e-05,
+      "loss": 1.0788,
+      "step": 9054
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780728902574884e-05,
+      "loss": 0.8463,
+      "step": 9055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806756129104687e-05,
+      "loss": 1.0659,
+      "step": 9056
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7806223175688933e-05,
+      "loss": 1.3707,
+      "step": 9057
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805690165505458e-05,
+      "loss": 1.0715,
+      "step": 9058
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805157098558137e-05,
+      "loss": 1.2495,
+      "step": 9059
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804623974850844e-05,
+      "loss": 0.935,
+      "step": 9060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780409079438746e-05,
+      "loss": 1.4758,
+      "step": 9061
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7803557557171865e-05,
+      "loss": 1.2023,
+      "step": 9062
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780302426320793e-05,
+      "loss": 0.6529,
+      "step": 9063
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7802490912499538e-05,
+      "loss": 0.792,
+      "step": 9064
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7801957505050563e-05,
+      "loss": 1.0063,
+      "step": 9065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780142404086489e-05,
+      "loss": 0.7212,
+      "step": 9066
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780089051994639e-05,
+      "loss": 0.9377,
+      "step": 9067
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7800356942298952e-05,
+      "loss": 1.1963,
+      "step": 9068
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779982330792645e-05,
+      "loss": 1.0882,
+      "step": 9069
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7799289616832768e-05,
+      "loss": 0.8862,
+      "step": 9070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7798755869021787e-05,
+      "loss": 0.8218,
+      "step": 9071
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7798222064497387e-05,
+      "loss": 1.0057,
+      "step": 9072
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797688203263446e-05,
+      "loss": 1.0479,
+      "step": 9073
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797154285323852e-05,
+      "loss": 1.024,
+      "step": 9074
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779662031068249e-05,
+      "loss": 1.4088,
+      "step": 9075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7796086279343235e-05,
+      "loss": 0.9651,
+      "step": 9076
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779555219130998e-05,
+      "loss": 1.01,
+      "step": 9077
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77950180465866e-05,
+      "loss": 1.3136,
+      "step": 9078
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7794483845176984e-05,
+      "loss": 0.9961,
+      "step": 9079
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7793949587085014e-05,
+      "loss": 1.1443,
+      "step": 9080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779341527231458e-05,
+      "loss": 0.7177,
+      "step": 9081
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779288090086956e-05,
+      "loss": 0.9224,
+      "step": 9082
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779234647275385e-05,
+      "loss": 1.002,
+      "step": 9083
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791811987971327e-05,
+      "loss": 0.942,
+      "step": 9084
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791277446525884e-05,
+      "loss": 1.2646,
+      "step": 9085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790742848421408e-05,
+      "loss": 1.0282,
+      "step": 9086
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779020819366178e-05,
+      "loss": 0.9629,
+      "step": 9087
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77896734822509e-05,
+      "loss": 1.1582,
+      "step": 9088
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7789138714192644e-05,
+      "loss": 1.0627,
+      "step": 9089
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7788603889490907e-05,
+      "loss": 0.8886,
+      "step": 9090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778806900814958e-05,
+      "loss": 0.9562,
+      "step": 9091
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7787534070172548e-05,
+      "loss": 0.7324,
+      "step": 9092
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786999075563706e-05,
+      "loss": 1.2051,
+      "step": 9093
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778646402432694e-05,
+      "loss": 1.0709,
+      "step": 9094
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778592891646614e-05,
+      "loss": 0.7844,
+      "step": 9095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785393751985207e-05,
+      "loss": 1.0782,
+      "step": 9096
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778485853088802e-05,
+      "loss": 0.6994,
+      "step": 9097
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778432325317848e-05,
+      "loss": 0.9309,
+      "step": 9098
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778378791886048e-05,
+      "loss": 0.9759,
+      "step": 9099
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7783252527937904e-05,
+      "loss": 0.9916,
+      "step": 9100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7782717080414653e-05,
+      "loss": 1.1934,
+      "step": 9101
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7782181576294622e-05,
+      "loss": 1.0259,
+      "step": 9102
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7781646015581698e-05,
+      "loss": 0.8294,
+      "step": 9103
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7781110398279783e-05,
+      "loss": 1.1981,
+      "step": 9104
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778057472439277e-05,
+      "loss": 0.8316,
+      "step": 9105
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7780038993924554e-05,
+      "loss": 0.4105,
+      "step": 9106
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777950320687903e-05,
+      "loss": 1.2172,
+      "step": 9107
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777896736326009e-05,
+      "loss": 1.0243,
+      "step": 9108
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777843146307164e-05,
+      "loss": 0.9013,
+      "step": 9109
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777789550631757e-05,
+      "loss": 0.9381,
+      "step": 9110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777359493001786e-05,
+      "loss": 0.6708,
+      "step": 9111
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7776823423128172e-05,
+      "loss": 0.9644,
+      "step": 9112
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777628729670064e-05,
+      "loss": 0.9937,
+      "step": 9113
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777575111372308e-05,
+      "loss": 1.0332,
+      "step": 9114
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77752148741994e-05,
+      "loss": 1.2238,
+      "step": 9115
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7774678578133486e-05,
+      "loss": 0.8761,
+      "step": 9116
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777414222552925e-05,
+      "loss": 1.001,
+      "step": 9117
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777360581639059e-05,
+      "loss": 1.1771,
+      "step": 9118
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77730693507214e-05,
+      "loss": 0.9179,
+      "step": 9119
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7772532828525593e-05,
+      "loss": 0.845,
+      "step": 9120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7771996249807063e-05,
+      "loss": 1.1153,
+      "step": 9121
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777145961456971e-05,
+      "loss": 0.8669,
+      "step": 9122
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770922922817442e-05,
+      "loss": 0.9675,
+      "step": 9123
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770386174554158e-05,
+      "loss": 1.0664,
+      "step": 9124
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7769849369783764e-05,
+      "loss": 1.0901,
+      "step": 9125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7769312508510164e-05,
+      "loss": 0.6647,
+      "step": 9126
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7768775590737257e-05,
+      "loss": 0.8247,
+      "step": 9127
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7768238616468954e-05,
+      "loss": 1.0188,
+      "step": 9128
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7767701585709157e-05,
+      "loss": 1.2207,
+      "step": 9129
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776716449846177e-05,
+      "loss": 1.381,
+      "step": 9130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77666273547307e-05,
+      "loss": 0.6723,
+      "step": 9131
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7766090154519855e-05,
+      "loss": 0.6782,
+      "step": 9132
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7765552897833142e-05,
+      "loss": 1.1422,
+      "step": 9133
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776501558467446e-05,
+      "loss": 1.4822,
+      "step": 9134
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7764478215047725e-05,
+      "loss": 1.0962,
+      "step": 9135
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7763940788956845e-05,
+      "loss": 0.8499,
+      "step": 9136
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7763403306405724e-05,
+      "loss": 0.8418,
+      "step": 9137
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776286576739827e-05,
+      "loss": 0.8377,
+      "step": 9138
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7762328171938396e-05,
+      "loss": 0.8414,
+      "step": 9139
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7761790520030012e-05,
+      "loss": 1.1513,
+      "step": 9140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776125281167702e-05,
+      "loss": 1.0243,
+      "step": 9141
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776071504688334e-05,
+      "loss": 0.6776,
+      "step": 9142
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7760177225652874e-05,
+      "loss": 1.0307,
+      "step": 9143
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7759639347989543e-05,
+      "loss": 0.9581,
+      "step": 9144
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7759101413897246e-05,
+      "loss": 0.8958,
+      "step": 9145
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775856342337991e-05,
+      "loss": 1.2047,
+      "step": 9146
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7758025376441432e-05,
+      "loss": 1.2643,
+      "step": 9147
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757487273085735e-05,
+      "loss": 1.0054,
+      "step": 9148
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7756949113316732e-05,
+      "loss": 1.0211,
+      "step": 9149
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7756410897138326e-05,
+      "loss": 1.2039,
+      "step": 9150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7755872624554445e-05,
+      "loss": 1.2571,
+      "step": 9151
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7755334295568998e-05,
+      "loss": 1.5993,
+      "step": 9152
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754795910185895e-05,
+      "loss": 1.0356,
+      "step": 9153
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754257468409057e-05,
+      "loss": 0.9367,
+      "step": 9154
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77537189702424e-05,
+      "loss": 1.1017,
+      "step": 9155
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775318041568983e-05,
+      "loss": 0.9238,
+      "step": 9156
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7752641804755275e-05,
+      "loss": 1.1016,
+      "step": 9157
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775210313744265e-05,
+      "loss": 0.9267,
+      "step": 9158
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775156441375587e-05,
+      "loss": 1.0064,
+      "step": 9159
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775102563369885e-05,
+      "loss": 1.1326,
+      "step": 9160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750486797275512e-05,
+      "loss": 0.6814,
+      "step": 9161
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7749947904489773e-05,
+      "loss": 1.1293,
+      "step": 9162
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7749408955345552e-05,
+      "loss": 1.192,
+      "step": 9163
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774886994984677e-05,
+      "loss": 0.8568,
+      "step": 9164
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7748330887997344e-05,
+      "loss": 0.8646,
+      "step": 9165
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7747791769801196e-05,
+      "loss": 0.8504,
+      "step": 9166
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7747252595262247e-05,
+      "loss": 1.4229,
+      "step": 9167
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7746713364384412e-05,
+      "loss": 1.2145,
+      "step": 9168
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7746174077171625e-05,
+      "loss": 0.9698,
+      "step": 9169
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745634733627793e-05,
+      "loss": 1.1156,
+      "step": 9170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745095333756846e-05,
+      "loss": 1.0433,
+      "step": 9171
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774455587756271e-05,
+      "loss": 1.0215,
+      "step": 9172
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77440163650493e-05,
+      "loss": 1.2253,
+      "step": 9173
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7743476796220543e-05,
+      "loss": 1.0067,
+      "step": 9174
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7742937171080368e-05,
+      "loss": 1.1579,
+      "step": 9175
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774239748963269e-05,
+      "loss": 0.9482,
+      "step": 9176
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741857751881436e-05,
+      "loss": 0.9802,
+      "step": 9177
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741317957830534e-05,
+      "loss": 0.9384,
+      "step": 9178
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7740778107483908e-05,
+      "loss": 1.2419,
+      "step": 9179
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7740238200845485e-05,
+      "loss": 1.2698,
+      "step": 9180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773969823791919e-05,
+      "loss": 1.4669,
+      "step": 9181
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773915821870895e-05,
+      "loss": 1.0836,
+      "step": 9182
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738618143218695e-05,
+      "loss": 1.0639,
+      "step": 9183
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738078011452345e-05,
+      "loss": 1.0041,
+      "step": 9184
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7737537823413832e-05,
+      "loss": 1.0846,
+      "step": 9185
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7736997579107087e-05,
+      "loss": 1.032,
+      "step": 9186
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7736457278536035e-05,
+      "loss": 0.7031,
+      "step": 9187
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773591692170461e-05,
+      "loss": 1.0307,
+      "step": 9188
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7735376508616733e-05,
+      "loss": 1.0114,
+      "step": 9189
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7734836039276342e-05,
+      "loss": 0.928,
+      "step": 9190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7734295513687363e-05,
+      "loss": 1.0742,
+      "step": 9191
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773375493185373e-05,
+      "loss": 0.8846,
+      "step": 9192
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773321429377937e-05,
+      "loss": 1.3763,
+      "step": 9193
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773267359946822e-05,
+      "loss": 0.8505,
+      "step": 9194
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7732132848924206e-05,
+      "loss": 1.0882,
+      "step": 9195
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7731592042151266e-05,
+      "loss": 0.8688,
+      "step": 9196
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7731051179153326e-05,
+      "loss": 1.3304,
+      "step": 9197
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7730510259934326e-05,
+      "loss": 0.8882,
+      "step": 9198
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7729969284498197e-05,
+      "loss": 0.9286,
+      "step": 9199
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7729428252848872e-05,
+      "loss": 0.6346,
+      "step": 9200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772888716499029e-05,
+      "loss": 1.0462,
+      "step": 9201
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7728346020926376e-05,
+      "loss": 0.973,
+      "step": 9202
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7727804820661075e-05,
+      "loss": 1.3389,
+      "step": 9203
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772726356419832e-05,
+      "loss": 1.322,
+      "step": 9204
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7726722251542047e-05,
+      "loss": 0.8458,
+      "step": 9205
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772618088269619e-05,
+      "loss": 1.2477,
+      "step": 9206
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7725639457664687e-05,
+      "loss": 0.7833,
+      "step": 9207
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7725097976451476e-05,
+      "loss": 1.043,
+      "step": 9208
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77245564390605e-05,
+      "loss": 1.1239,
+      "step": 9209
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7724014845495684e-05,
+      "loss": 0.9089,
+      "step": 9210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772347319576098e-05,
+      "loss": 0.8523,
+      "step": 9211
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7722931489860316e-05,
+      "loss": 0.8735,
+      "step": 9212
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7722389727797642e-05,
+      "loss": 1.1283,
+      "step": 9213
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7721847909576887e-05,
+      "loss": 1.0052,
+      "step": 9214
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7721306035202e-05,
+      "loss": 0.9377,
+      "step": 9215
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7720764104676917e-05,
+      "loss": 1.1236,
+      "step": 9216
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772022211800558e-05,
+      "loss": 1.1416,
+      "step": 9217
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7719680075191932e-05,
+      "loss": 0.7259,
+      "step": 9218
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7719137976239914e-05,
+      "loss": 0.8588,
+      "step": 9219
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7718595821153465e-05,
+      "loss": 0.658,
+      "step": 9220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771805360993653e-05,
+      "loss": 1.3761,
+      "step": 9221
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771751134259305e-05,
+      "loss": 1.1801,
+      "step": 9222
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7716969019126975e-05,
+      "loss": 1.0572,
+      "step": 9223
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7716426639542242e-05,
+      "loss": 1.105,
+      "step": 9224
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77158842038428e-05,
+      "loss": 0.7413,
+      "step": 9225
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7715341712032587e-05,
+      "loss": 0.9944,
+      "step": 9226
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7714799164115553e-05,
+      "loss": 0.8874,
+      "step": 9227
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7714256560095648e-05,
+      "loss": 0.6779,
+      "step": 9228
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7713713899976808e-05,
+      "loss": 1.0511,
+      "step": 9229
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7713171183762984e-05,
+      "loss": 1.6005,
+      "step": 9230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712628411458123e-05,
+      "loss": 0.9589,
+      "step": 9231
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7712085583066173e-05,
+      "loss": 0.8244,
+      "step": 9232
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7711542698591084e-05,
+      "loss": 1.0337,
+      "step": 9233
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710999758036792e-05,
+      "loss": 1.3457,
+      "step": 9234
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710456761407256e-05,
+      "loss": 0.921,
+      "step": 9235
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709913708706426e-05,
+      "loss": 1.1994,
+      "step": 9236
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709370599938242e-05,
+      "loss": 1.124,
+      "step": 9237
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708827435106662e-05,
+      "loss": 0.8394,
+      "step": 9238
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708284214215633e-05,
+      "loss": 0.9994,
+      "step": 9239
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7707740937269108e-05,
+      "loss": 0.9142,
+      "step": 9240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770719760427103e-05,
+      "loss": 0.8013,
+      "step": 9241
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706654215225355e-05,
+      "loss": 1.4431,
+      "step": 9242
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706110770136037e-05,
+      "loss": 1.0509,
+      "step": 9243
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7705567269007024e-05,
+      "loss": 1.0655,
+      "step": 9244
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7705023711842274e-05,
+      "loss": 1.0752,
+      "step": 9245
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704480098645733e-05,
+      "loss": 1.1767,
+      "step": 9246
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7703936429421357e-05,
+      "loss": 1.054,
+      "step": 9247
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77033927041731e-05,
+      "loss": 1.1116,
+      "step": 9248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7702848922904914e-05,
+      "loss": 0.9465,
+      "step": 9249
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7702305085620762e-05,
+      "loss": 1.0268,
+      "step": 9250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7701761192324588e-05,
+      "loss": 0.9262,
+      "step": 9251
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7701217243020348e-05,
+      "loss": 1.1128,
+      "step": 9252
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7700673237712005e-05,
+      "loss": 0.9017,
+      "step": 9253
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7700129176403508e-05,
+      "loss": 1.0453,
+      "step": 9254
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769958505909882e-05,
+      "loss": 0.9307,
+      "step": 9255
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7699040885801893e-05,
+      "loss": 0.8325,
+      "step": 9256
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7698496656516686e-05,
+      "loss": 1.0435,
+      "step": 9257
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7697952371247156e-05,
+      "loss": 1.0413,
+      "step": 9258
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769740802999726e-05,
+      "loss": 1.0444,
+      "step": 9259
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769686363277096e-05,
+      "loss": 1.0166,
+      "step": 9260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696319179572215e-05,
+      "loss": 1.0982,
+      "step": 9261
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7695774670404978e-05,
+      "loss": 1.2213,
+      "step": 9262
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7695230105273217e-05,
+      "loss": 0.8933,
+      "step": 9263
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7694685484180886e-05,
+      "loss": 1.0105,
+      "step": 9264
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769414080713195e-05,
+      "loss": 1.3956,
+      "step": 9265
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769359607413036e-05,
+      "loss": 0.8195,
+      "step": 9266
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769305128518009e-05,
+      "loss": 1.0194,
+      "step": 9267
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7692506440285097e-05,
+      "loss": 1.1553,
+      "step": 9268
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691961539449344e-05,
+      "loss": 0.8923,
+      "step": 9269
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7691416582676792e-05,
+      "loss": 1.0061,
+      "step": 9270
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690871569971405e-05,
+      "loss": 1.1069,
+      "step": 9271
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769032650133714e-05,
+      "loss": 0.9162,
+      "step": 9272
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7689781376777974e-05,
+      "loss": 0.8669,
+      "step": 9273
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7689236196297865e-05,
+      "loss": 1.5206,
+      "step": 9274
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768869095990077e-05,
+      "loss": 0.8756,
+      "step": 9275
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7688145667590665e-05,
+      "loss": 1.2226,
+      "step": 9276
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768760031937151e-05,
+      "loss": 1.0863,
+      "step": 9277
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768705491524727e-05,
+      "loss": 1.1474,
+      "step": 9278
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7686509455221915e-05,
+      "loss": 1.2146,
+      "step": 9279
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7685963939299408e-05,
+      "loss": 0.8831,
+      "step": 9280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768541836748372e-05,
+      "loss": 0.6106,
+      "step": 9281
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7684872739778814e-05,
+      "loss": 1.1212,
+      "step": 9282
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7684327056188662e-05,
+      "loss": 0.7376,
+      "step": 9283
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683781316717226e-05,
+      "loss": 1.0047,
+      "step": 9284
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683235521368484e-05,
+      "loss": 0.9121,
+      "step": 9285
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.76826896701464e-05,
+      "loss": 1.2342,
+      "step": 9286
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768214376305494e-05,
+      "loss": 1.0897,
+      "step": 9287
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768159780009808e-05,
+      "loss": 1.2876,
+      "step": 9288
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7681051781279787e-05,
+      "loss": 0.7553,
+      "step": 9289
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768050570660403e-05,
+      "loss": 0.7747,
+      "step": 9290
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679959576074786e-05,
+      "loss": 0.5474,
+      "step": 9291
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679413389696025e-05,
+      "loss": 1.1551,
+      "step": 9292
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7678867147471713e-05,
+      "loss": 0.7111,
+      "step": 9293
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767832084940583e-05,
+      "loss": 1.1477,
+      "step": 9294
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677774495502342e-05,
+      "loss": 1.1298,
+      "step": 9295
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677228085765224e-05,
+      "loss": 0.9093,
+      "step": 9296
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7676681620198453e-05,
+      "loss": 1.4281,
+      "step": 9297
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7676135098806004e-05,
+      "loss": 0.8811,
+      "step": 9298
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7675588521591845e-05,
+      "loss": 1.0863,
+      "step": 9299
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7675041888559952e-05,
+      "loss": 1.0544,
+      "step": 9300
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7674495199714307e-05,
+      "loss": 0.8131,
+      "step": 9301
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7673948455058874e-05,
+      "loss": 0.8243,
+      "step": 9302
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767340165459764e-05,
+      "loss": 1.3137,
+      "step": 9303
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.767285479833458e-05,
+      "loss": 1.0264,
+      "step": 9304
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7672307886273662e-05,
+      "loss": 1.2399,
+      "step": 9305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767176091841887e-05,
+      "loss": 1.095,
+      "step": 9306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767121389477418e-05,
+      "loss": 0.9789,
+      "step": 9307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7670666815343575e-05,
+      "loss": 1.1486,
+      "step": 9308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7670119680131028e-05,
+      "loss": 0.8706,
+      "step": 9309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766957248914052e-05,
+      "loss": 0.9583,
+      "step": 9310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7669025242376023e-05,
+      "loss": 1.2933,
+      "step": 9311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7668477939841526e-05,
+      "loss": 1.0472,
+      "step": 9312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667930581541006e-05,
+      "loss": 1.1362,
+      "step": 9313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667383167478444e-05,
+      "loss": 1.1078,
+      "step": 9314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666835697657824e-05,
+      "loss": 0.9127,
+      "step": 9315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666288172083116e-05,
+      "loss": 0.8435,
+      "step": 9316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665740590758313e-05,
+      "loss": 1.2475,
+      "step": 9317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665192953687392e-05,
+      "loss": 0.725,
+      "step": 9318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7664645260874337e-05,
+      "loss": 1.0669,
+      "step": 9319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7664097512323132e-05,
+      "loss": 1.2035,
+      "step": 9320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766354970803776e-05,
+      "loss": 1.0839,
+      "step": 9321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76630018480222e-05,
+      "loss": 0.7872,
+      "step": 9322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766245393228044e-05,
+      "loss": 1.1171,
+      "step": 9323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7661905960816463e-05,
+      "loss": 1.3109,
+      "step": 9324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766135793363426e-05,
+      "loss": 1.3329,
+      "step": 9325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7660809850737805e-05,
+      "loss": 1.2786,
+      "step": 9326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7660261712131096e-05,
+      "loss": 0.9862,
+      "step": 9327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765971351781811e-05,
+      "loss": 1.1259,
+      "step": 9328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765916526780284e-05,
+      "loss": 0.8987,
+      "step": 9329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7658616962089262e-05,
+      "loss": 0.6873,
+      "step": 9330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765806860068138e-05,
+      "loss": 0.8189,
+      "step": 9331
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765752018358317e-05,
+      "loss": 0.8585,
+      "step": 9332
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765697171079862e-05,
+      "loss": 0.9726,
+      "step": 9333
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765642318233172e-05,
+      "loss": 0.9624,
+      "step": 9334
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7655874598186465e-05,
+      "loss": 1.0896,
+      "step": 9335
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7655325958366834e-05,
+      "loss": 1.333,
+      "step": 9336
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7654777262876827e-05,
+      "loss": 0.8167,
+      "step": 9337
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765422851172043e-05,
+      "loss": 1.2149,
+      "step": 9338
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7653679704901635e-05,
+      "loss": 0.84,
+      "step": 9339
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7653130842424428e-05,
+      "loss": 1.2106,
+      "step": 9340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7652581924292803e-05,
+      "loss": 1.0992,
+      "step": 9341
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7652032950510755e-05,
+      "loss": 1.4722,
+      "step": 9342
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7651483921082274e-05,
+      "loss": 1.1291,
+      "step": 9343
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7650934836011348e-05,
+      "loss": 1.1743,
+      "step": 9344
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765038569530198e-05,
+      "loss": 0.7334,
+      "step": 9345
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649836498958154e-05,
+      "loss": 1.3036,
+      "step": 9346
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649287246983865e-05,
+      "loss": 0.8924,
+      "step": 9347
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7648737939383114e-05,
+      "loss": 1.0781,
+      "step": 9348
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764818857615989e-05,
+      "loss": 1.1196,
+      "step": 9349
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7647639157318193e-05,
+      "loss": 1.0098,
+      "step": 9350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764708968286201e-05,
+      "loss": 0.8053,
+      "step": 9351
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7646540152795342e-05,
+      "loss": 1.4982,
+      "step": 9352
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645990567122185e-05,
+      "loss": 1.0505,
+      "step": 9353
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764544092584654e-05,
+      "loss": 0.7371,
+      "step": 9354
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7644891228972392e-05,
+      "loss": 0.9901,
+      "step": 9355
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7644341476503753e-05,
+      "loss": 0.9886,
+      "step": 9356
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764379166844461e-05,
+      "loss": 0.7805,
+      "step": 9357
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7643241804798966e-05,
+      "loss": 0.7855,
+      "step": 9358
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7642691885570817e-05,
+      "loss": 0.9233,
+      "step": 9359
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7642141910764164e-05,
+      "loss": 1.2721,
+      "step": 9360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7641591880383012e-05,
+      "loss": 1.0406,
+      "step": 9361
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764104179443135e-05,
+      "loss": 1.1091,
+      "step": 9362
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7640491652913186e-05,
+      "loss": 1.0712,
+      "step": 9363
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7639941455832518e-05,
+      "loss": 1.1422,
+      "step": 9364
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7639391203193346e-05,
+      "loss": 1.0456,
+      "step": 9365
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763884089499967e-05,
+      "loss": 1.0797,
+      "step": 9366
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7638290531255503e-05,
+      "loss": 1.0918,
+      "step": 9367
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763774011196483e-05,
+      "loss": 1.1046,
+      "step": 9368
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763718963713167e-05,
+      "loss": 0.8138,
+      "step": 9369
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7636639106760016e-05,
+      "loss": 1.1393,
+      "step": 9370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7636088520853874e-05,
+      "loss": 1.268,
+      "step": 9371
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635537879417248e-05,
+      "loss": 0.8849,
+      "step": 9372
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7634987182454144e-05,
+      "loss": 0.8312,
+      "step": 9373
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7634436429968564e-05,
+      "loss": 0.9449,
+      "step": 9374
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7633885621964516e-05,
+      "loss": 0.7427,
+      "step": 9375
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7633334758446002e-05,
+      "loss": 1.1726,
+      "step": 9376
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7632783839417032e-05,
+      "loss": 1.1824,
+      "step": 9377
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763223286488161e-05,
+      "loss": 1.068,
+      "step": 9378
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763168183484374e-05,
+      "loss": 1.24,
+      "step": 9379
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631130749307436e-05,
+      "loss": 0.6069,
+      "step": 9380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76305796082767e-05,
+      "loss": 1.2734,
+      "step": 9381
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7630028411755542e-05,
+      "loss": 1.0689,
+      "step": 9382
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762947715974797e-05,
+      "loss": 0.9301,
+      "step": 9383
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7628925852257994e-05,
+      "loss": 1.0758,
+      "step": 9384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762837448928962e-05,
+      "loss": 0.8941,
+      "step": 9385
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7627823070846863e-05,
+      "loss": 1.0224,
+      "step": 9386
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7627271596933724e-05,
+      "loss": 0.8468,
+      "step": 9387
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7626720067554225e-05,
+      "loss": 0.991,
+      "step": 9388
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7626168482712367e-05,
+      "loss": 1.1735,
+      "step": 9389
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7625616842412166e-05,
+      "loss": 1.2323,
+      "step": 9390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7625065146657632e-05,
+      "loss": 1.0775,
+      "step": 9391
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7624513395452777e-05,
+      "loss": 1.1022,
+      "step": 9392
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7623961588801614e-05,
+      "loss": 1.1962,
+      "step": 9393
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762340972670816e-05,
+      "loss": 1.0087,
+      "step": 9394
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7622857809176422e-05,
+      "loss": 1.2161,
+      "step": 9395
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762230583621041e-05,
+      "loss": 0.9477,
+      "step": 9396
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762175380781415e-05,
+      "loss": 1.1121,
+      "step": 9397
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7621201723991646e-05,
+      "loss": 0.9497,
+      "step": 9398
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7620649584746918e-05,
+      "loss": 1.2501,
+      "step": 9399
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762009739008398e-05,
+      "loss": 0.9635,
+      "step": 9400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7619545140006848e-05,
+      "loss": 0.7962,
+      "step": 9401
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618992834519536e-05,
+      "loss": 0.993,
+      "step": 9402
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618440473626065e-05,
+      "loss": 1.0911,
+      "step": 9403
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761788805733045e-05,
+      "loss": 0.8469,
+      "step": 9404
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76173355856367e-05,
+      "loss": 1.7499,
+      "step": 9405
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616783058548846e-05,
+      "loss": 0.989,
+      "step": 9406
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76162304760709e-05,
+      "loss": 1.0935,
+      "step": 9407
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7615677838206876e-05,
+      "loss": 1.2613,
+      "step": 9408
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7615125144960797e-05,
+      "loss": 0.8983,
+      "step": 9409
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7614572396336686e-05,
+      "loss": 1.2583,
+      "step": 9410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761401959233856e-05,
+      "loss": 0.7483,
+      "step": 9411
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7613466732970434e-05,
+      "loss": 1.0306,
+      "step": 9412
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612913818236332e-05,
+      "loss": 0.9343,
+      "step": 9413
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761236084814028e-05,
+      "loss": 0.956,
+      "step": 9414
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7611807822686292e-05,
+      "loss": 1.1556,
+      "step": 9415
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761125474187839e-05,
+      "loss": 1.1013,
+      "step": 9416
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7610701605720603e-05,
+      "loss": 1.2111,
+      "step": 9417
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7610148414216947e-05,
+      "loss": 1.1909,
+      "step": 9418
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760959516737144e-05,
+      "loss": 0.832,
+      "step": 9419
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7609041865188122e-05,
+      "loss": 1.0781,
+      "step": 9420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7608488507671003e-05,
+      "loss": 0.92,
+      "step": 9421
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7607935094824106e-05,
+      "loss": 1.4987,
+      "step": 9422
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7607381626651466e-05,
+      "loss": 0.8817,
+      "step": 9423
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606828103157102e-05,
+      "loss": 1.1564,
+      "step": 9424
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606274524345036e-05,
+      "loss": 0.7915,
+      "step": 9425
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76057208902193e-05,
+      "loss": 1.0467,
+      "step": 9426
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7605167200783914e-05,
+      "loss": 1.3802,
+      "step": 9427
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7604613456042908e-05,
+      "loss": 1.2404,
+      "step": 9428
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7604059656000313e-05,
+      "loss": 0.8752,
+      "step": 9429
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760350580066015e-05,
+      "loss": 1.2674,
+      "step": 9430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7602951890026442e-05,
+      "loss": 0.9631,
+      "step": 9431
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760239792410323e-05,
+      "loss": 0.8538,
+      "step": 9432
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7601843902894535e-05,
+      "loss": 1.2571,
+      "step": 9433
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7601289826404385e-05,
+      "loss": 1.261,
+      "step": 9434
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600735694636814e-05,
+      "loss": 1.2878,
+      "step": 9435
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600181507595847e-05,
+      "loss": 0.5689,
+      "step": 9436
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7599627265285515e-05,
+      "loss": 1.2257,
+      "step": 9437
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7599072967709852e-05,
+      "loss": 1.1189,
+      "step": 9438
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7598518614872885e-05,
+      "loss": 1.18,
+      "step": 9439
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759796420677865e-05,
+      "loss": 1.11,
+      "step": 9440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759740974343117e-05,
+      "loss": 0.9375,
+      "step": 9441
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759685522483449e-05,
+      "loss": 0.9933,
+      "step": 9442
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759630065099263e-05,
+      "loss": 1.0833,
+      "step": 9443
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759574602190963e-05,
+      "loss": 1.0474,
+      "step": 9444
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759519133758952e-05,
+      "loss": 1.0162,
+      "step": 9445
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594636598036338e-05,
+      "loss": 1.05,
+      "step": 9446
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594081803254113e-05,
+      "loss": 1.3608,
+      "step": 9447
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759352695324688e-05,
+      "loss": 1.0067,
+      "step": 9448
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759297204801868e-05,
+      "loss": 1.0839,
+      "step": 9449
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759241708757354e-05,
+      "loss": 0.7463,
+      "step": 9450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7591862071915502e-05,
+      "loss": 1.0402,
+      "step": 9451
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7591307001048598e-05,
+      "loss": 1.1256,
+      "step": 9452
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759075187497687e-05,
+      "loss": 0.9747,
+      "step": 9453
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759019669370435e-05,
+      "loss": 1.3768,
+      "step": 9454
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7589641457235077e-05,
+      "loss": 1.2378,
+      "step": 9455
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7589086165573087e-05,
+      "loss": 0.5615,
+      "step": 9456
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758853081872242e-05,
+      "loss": 1.0977,
+      "step": 9457
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7587975416687115e-05,
+      "loss": 1.2602,
+      "step": 9458
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758741995947121e-05,
+      "loss": 1.0886,
+      "step": 9459
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7586864447078746e-05,
+      "loss": 1.1703,
+      "step": 9460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7586308879513757e-05,
+      "loss": 1.0427,
+      "step": 9461
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7585753256780293e-05,
+      "loss": 1.1601,
+      "step": 9462
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758519757888238e-05,
+      "loss": 0.8154,
+      "step": 9463
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758464184582408e-05,
+      "loss": 1.1174,
+      "step": 9464
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7584086057609413e-05,
+      "loss": 0.7906,
+      "step": 9465
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583530214242436e-05,
+      "loss": 1.1255,
+      "step": 9466
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758297431572718e-05,
+      "loss": 1.1415,
+      "step": 9467
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7582418362067698e-05,
+      "loss": 0.9974,
+      "step": 9468
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758186235326802e-05,
+      "loss": 0.72,
+      "step": 9469
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75813062893322e-05,
+      "loss": 0.8346,
+      "step": 9470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758075017026428e-05,
+      "loss": 1.29,
+      "step": 9471
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758019399606831e-05,
+      "loss": 1.1802,
+      "step": 9472
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757963776674832e-05,
+      "loss": 1.0772,
+      "step": 9473
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7579081482308364e-05,
+      "loss": 1.6058,
+      "step": 9474
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757852514275248e-05,
+      "loss": 0.8049,
+      "step": 9475
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757796874808473e-05,
+      "loss": 1.2798,
+      "step": 9476
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7577412298309142e-05,
+      "loss": 1.1705,
+      "step": 9477
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7576855793429773e-05,
+      "loss": 1.2989,
+      "step": 9478
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7576299233450666e-05,
+      "loss": 0.7721,
+      "step": 9479
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757574261837587e-05,
+      "loss": 0.8109,
+      "step": 9480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7575185948209433e-05,
+      "loss": 1.4211,
+      "step": 9481
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7574629222955405e-05,
+      "loss": 0.9066,
+      "step": 9482
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7574072442617827e-05,
+      "loss": 1.2884,
+      "step": 9483
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573515607200757e-05,
+      "loss": 1.0529,
+      "step": 9484
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7572958716708242e-05,
+      "loss": 1.3097,
+      "step": 9485
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7572401771144325e-05,
+      "loss": 1.3168,
+      "step": 9486
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571844770513065e-05,
+      "loss": 1.2219,
+      "step": 9487
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571287714818507e-05,
+      "loss": 0.9699,
+      "step": 9488
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7570730604064707e-05,
+      "loss": 1.0858,
+      "step": 9489
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757017343825571e-05,
+      "loss": 1.2522,
+      "step": 9490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7569616217395572e-05,
+      "loss": 1.4409,
+      "step": 9491
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7569058941488345e-05,
+      "loss": 0.997,
+      "step": 9492
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756850161053808e-05,
+      "loss": 1.1069,
+      "step": 9493
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7567944224548833e-05,
+      "loss": 0.8656,
+      "step": 9494
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7567386783524655e-05,
+      "loss": 1.0912,
+      "step": 9495
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7566829287469598e-05,
+      "loss": 1.0752,
+      "step": 9496
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756627173638772e-05,
+      "loss": 0.9872,
+      "step": 9497
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7565714130283068e-05,
+      "loss": 1.2809,
+      "step": 9498
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756515646915971e-05,
+      "loss": 1.3123,
+      "step": 9499
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756459875302169e-05,
+      "loss": 1.0469,
+      "step": 9500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756404098187307e-05,
+      "loss": 1.0644,
+      "step": 9501
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7563483155717902e-05,
+      "loss": 0.9155,
+      "step": 9502
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7562925274560247e-05,
+      "loss": 1.2139,
+      "step": 9503
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756236733840416e-05,
+      "loss": 0.9927,
+      "step": 9504
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561809347253696e-05,
+      "loss": 1.1093,
+      "step": 9505
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561251301112915e-05,
+      "loss": 0.8438,
+      "step": 9506
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7560693199985875e-05,
+      "loss": 0.6644,
+      "step": 9507
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7560135043876632e-05,
+      "loss": 1.1037,
+      "step": 9508
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7559576832789253e-05,
+      "loss": 0.7056,
+      "step": 9509
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7559018566727788e-05,
+      "loss": 0.9372,
+      "step": 9510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75584602456963e-05,
+      "loss": 1.0374,
+      "step": 9511
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7557901869698847e-05,
+      "loss": 1.0511,
+      "step": 9512
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75573434387395e-05,
+      "loss": 1.2271,
+      "step": 9513
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7556784952822308e-05,
+      "loss": 1.3278,
+      "step": 9514
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7556226411951337e-05,
+      "loss": 0.7147,
+      "step": 9515
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755566781613065e-05,
+      "loss": 0.8931,
+      "step": 9516
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7555109165364303e-05,
+      "loss": 1.0929,
+      "step": 9517
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554550459656367e-05,
+      "loss": 1.1106,
+      "step": 9518
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75539916990109e-05,
+      "loss": 1.2038,
+      "step": 9519
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755343288343197e-05,
+      "loss": 0.8998,
+      "step": 9520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755287401292363e-05,
+      "loss": 0.8656,
+      "step": 9521
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7552315087489958e-05,
+      "loss": 1.1636,
+      "step": 9522
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755175610713501e-05,
+      "loss": 0.7789,
+      "step": 9523
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7551197071862852e-05,
+      "loss": 0.8979,
+      "step": 9524
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755063798167755e-05,
+      "loss": 1.2497,
+      "step": 9525
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7550078836583173e-05,
+      "loss": 0.8965,
+      "step": 9526
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7549519636583784e-05,
+      "loss": 0.7898,
+      "step": 9527
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7548960381683446e-05,
+      "loss": 0.9133,
+      "step": 9528
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754840107188623e-05,
+      "loss": 1.0603,
+      "step": 9529
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7547841707196207e-05,
+      "loss": 0.9459,
+      "step": 9530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7547282287617438e-05,
+      "loss": 1.0398,
+      "step": 9531
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7546722813153997e-05,
+      "loss": 1.0972,
+      "step": 9532
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7546163283809946e-05,
+      "loss": 1.2723,
+      "step": 9533
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754560369958936e-05,
+      "loss": 0.9784,
+      "step": 9534
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754504406049631e-05,
+      "loss": 1.0032,
+      "step": 9535
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7544484366534854e-05,
+      "loss": 0.741,
+      "step": 9536
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7543924617709072e-05,
+      "loss": 0.7558,
+      "step": 9537
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7543364814023033e-05,
+      "loss": 0.8693,
+      "step": 9538
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754280495548081e-05,
+      "loss": 1.0807,
+      "step": 9539
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754224504208647e-05,
+      "loss": 0.836,
+      "step": 9540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7541685073844086e-05,
+      "loss": 0.8723,
+      "step": 9541
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754112505075773e-05,
+      "loss": 1.2295,
+      "step": 9542
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7540564972831476e-05,
+      "loss": 1.0019,
+      "step": 9543
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7540004840069397e-05,
+      "loss": 1.1001,
+      "step": 9544
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7539444652475565e-05,
+      "loss": 0.8322,
+      "step": 9545
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7538884410054056e-05,
+      "loss": 0.9061,
+      "step": 9546
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753832411280894e-05,
+      "loss": 0.8542,
+      "step": 9547
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7537763760744294e-05,
+      "loss": 0.911,
+      "step": 9548
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7537203353864192e-05,
+      "loss": 1.0184,
+      "step": 9549
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7536642892172714e-05,
+      "loss": 0.8974,
+      "step": 9550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753608237567393e-05,
+      "loss": 0.9885,
+      "step": 9551
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7535521804371915e-05,
+      "loss": 0.9857,
+      "step": 9552
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753496117827075e-05,
+      "loss": 0.6876,
+      "step": 9553
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7534400497374514e-05,
+      "loss": 1.0729,
+      "step": 9554
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533839761687278e-05,
+      "loss": 0.9414,
+      "step": 9555
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533278971213124e-05,
+      "loss": 0.9828,
+      "step": 9556
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753271812595613e-05,
+      "loss": 0.8593,
+      "step": 9557
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753215722592037e-05,
+      "loss": 1.3279,
+      "step": 9558
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7531596271109926e-05,
+      "loss": 0.9917,
+      "step": 9559
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753103526152888e-05,
+      "loss": 1.0851,
+      "step": 9560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7530474197181308e-05,
+      "loss": 0.9601,
+      "step": 9561
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529913078071293e-05,
+      "loss": 0.8481,
+      "step": 9562
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529351904202912e-05,
+      "loss": 1.0865,
+      "step": 9563
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7528790675580248e-05,
+      "loss": 0.9999,
+      "step": 9564
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7528229392207383e-05,
+      "loss": 1.1882,
+      "step": 9565
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527668054088397e-05,
+      "loss": 0.9246,
+      "step": 9566
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527106661227376e-05,
+      "loss": 0.802,
+      "step": 9567
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7526545213628393e-05,
+      "loss": 0.9098,
+      "step": 9568
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7525983711295543e-05,
+      "loss": 1.0603,
+      "step": 9569
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7525422154232906e-05,
+      "loss": 0.7142,
+      "step": 9570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524860542444558e-05,
+      "loss": 1.1069,
+      "step": 9571
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524298875934588e-05,
+      "loss": 1.012,
+      "step": 9572
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7523737154707086e-05,
+      "loss": 1.0776,
+      "step": 9573
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7523175378766127e-05,
+      "loss": 0.6238,
+      "step": 9574
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522613548115806e-05,
+      "loss": 0.8017,
+      "step": 9575
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522051662760197e-05,
+      "loss": 1.0027,
+      "step": 9576
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75214897227034e-05,
+      "loss": 1.1216,
+      "step": 9577
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520927727949494e-05,
+      "loss": 1.2544,
+      "step": 9578
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520365678502563e-05,
+      "loss": 1.0115,
+      "step": 9579
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75198035743667e-05,
+      "loss": 0.6685,
+      "step": 9580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7519241415545987e-05,
+      "loss": 1.3163,
+      "step": 9581
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751867920204452e-05,
+      "loss": 1.073,
+      "step": 9582
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751811693386638e-05,
+      "loss": 1.0108,
+      "step": 9583
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7517554611015657e-05,
+      "loss": 0.9691,
+      "step": 9584
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516992233496443e-05,
+      "loss": 1.1355,
+      "step": 9585
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751642980131283e-05,
+      "loss": 1.1262,
+      "step": 9586
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7515867314468902e-05,
+      "loss": 0.8642,
+      "step": 9587
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7515304772968755e-05,
+      "loss": 1.0426,
+      "step": 9588
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7514742176816476e-05,
+      "loss": 1.0889,
+      "step": 9589
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7514179526016156e-05,
+      "loss": 1.2486,
+      "step": 9590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513616820571887e-05,
+      "loss": 0.7296,
+      "step": 9591
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513054060487765e-05,
+      "loss": 0.9011,
+      "step": 9592
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7512491245767878e-05,
+      "loss": 0.9628,
+      "step": 9593
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751192837641632e-05,
+      "loss": 0.8515,
+      "step": 9594
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7511365452437184e-05,
+      "loss": 1.3243,
+      "step": 9595
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510802473834568e-05,
+      "loss": 1.0899,
+      "step": 9596
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751023944061256e-05,
+      "loss": 1.0518,
+      "step": 9597
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7509676352775257e-05,
+      "loss": 1.0746,
+      "step": 9598
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7509113210326755e-05,
+      "loss": 0.9537,
+      "step": 9599
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7508550013271146e-05,
+      "loss": 0.5897,
+      "step": 9600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507986761612525e-05,
+      "loss": 1.0248,
+      "step": 9601
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507423455354996e-05,
+      "loss": 0.8064,
+      "step": 9602
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7506860094502647e-05,
+      "loss": 0.8661,
+      "step": 9603
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750629667905958e-05,
+      "loss": 1.152,
+      "step": 9604
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7505733209029887e-05,
+      "loss": 0.8785,
+      "step": 9605
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750516968441767e-05,
+      "loss": 1.216,
+      "step": 9606
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504606105227023e-05,
+      "loss": 1.3645,
+      "step": 9607
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750404247146205e-05,
+      "loss": 1.2856,
+      "step": 9608
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7503478783126848e-05,
+      "loss": 0.8632,
+      "step": 9609
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502915040225512e-05,
+      "loss": 0.548,
+      "step": 9610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502351242762146e-05,
+      "loss": 1.1819,
+      "step": 9611
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750178739074085e-05,
+      "loss": 0.8703,
+      "step": 9612
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750122348416572e-05,
+      "loss": 0.9471,
+      "step": 9613
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750065952304086e-05,
+      "loss": 0.8862,
+      "step": 9614
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7500095507370376e-05,
+      "loss": 0.9747,
+      "step": 9615
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749953143715836e-05,
+      "loss": 0.8513,
+      "step": 9616
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7498967312408922e-05,
+      "loss": 0.8901,
+      "step": 9617
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749840313312616e-05,
+      "loss": 1.3188,
+      "step": 9618
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7497838899314173e-05,
+      "loss": 1.0435,
+      "step": 9619
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7497274610977074e-05,
+      "loss": 0.942,
+      "step": 9620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749671026811896e-05,
+      "loss": 1.0157,
+      "step": 9621
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7496145870743934e-05,
+      "loss": 1.4633,
+      "step": 9622
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7495581418856105e-05,
+      "loss": 1.1606,
+      "step": 9623
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7495016912459576e-05,
+      "loss": 0.8873,
+      "step": 9624
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7494452351558452e-05,
+      "loss": 1.0318,
+      "step": 9625
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749388773615684e-05,
+      "loss": 0.8536,
+      "step": 9626
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7493323066258837e-05,
+      "loss": 0.9978,
+      "step": 9627
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492758341868564e-05,
+      "loss": 1.4753,
+      "step": 9628
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492193562990117e-05,
+      "loss": 0.6454,
+      "step": 9629
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749162872962761e-05,
+      "loss": 1.1118,
+      "step": 9630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7491063841785143e-05,
+      "loss": 1.1509,
+      "step": 9631
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7490498899466827e-05,
+      "loss": 1.4628,
+      "step": 9632
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489933902676775e-05,
+      "loss": 0.7086,
+      "step": 9633
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748936885141909e-05,
+      "loss": 0.776,
+      "step": 9634
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7488803745697882e-05,
+      "loss": 1.081,
+      "step": 9635
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7488238585517265e-05,
+      "loss": 0.7902,
+      "step": 9636
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748767337088134e-05,
+      "loss": 0.9638,
+      "step": 9637
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487108101794227e-05,
+      "loss": 0.8443,
+      "step": 9638
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7486542778260033e-05,
+      "loss": 0.9222,
+      "step": 9639
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7485977400282867e-05,
+      "loss": 0.8083,
+      "step": 9640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748541196786684e-05,
+      "loss": 0.5327,
+      "step": 9641
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7484846481016067e-05,
+      "loss": 0.9581,
+      "step": 9642
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7484280939734662e-05,
+      "loss": 1.0279,
+      "step": 9643
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7483715344026732e-05,
+      "loss": 0.985,
+      "step": 9644
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7483149693896396e-05,
+      "loss": 1.0225,
+      "step": 9645
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482583989347758e-05,
+      "loss": 1.3562,
+      "step": 9646
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482018230384944e-05,
+      "loss": 0.7798,
+      "step": 9647
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7481452417012064e-05,
+      "loss": 0.961,
+      "step": 9648
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7480886549233228e-05,
+      "loss": 0.9928,
+      "step": 9649
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7480320627052552e-05,
+      "loss": 1.0546,
+      "step": 9650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479754650474157e-05,
+      "loss": 1.1626,
+      "step": 9651
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479188619502158e-05,
+      "loss": 0.9447,
+      "step": 9652
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478622534140663e-05,
+      "loss": 0.7847,
+      "step": 9653
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478056394393798e-05,
+      "loss": 0.9001,
+      "step": 9654
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7477490200265676e-05,
+      "loss": 1.3105,
+      "step": 9655
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7476923951760418e-05,
+      "loss": 1.0631,
+      "step": 9656
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7476357648882132e-05,
+      "loss": 1.1382,
+      "step": 9657
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475791291634946e-05,
+      "loss": 1.1717,
+      "step": 9658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475224880022977e-05,
+      "loss": 1.1034,
+      "step": 9659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474658414050344e-05,
+      "loss": 0.8758,
+      "step": 9660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474091893721157e-05,
+      "loss": 0.6351,
+      "step": 9661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7473525319039554e-05,
+      "loss": 0.9864,
+      "step": 9662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747295869000964e-05,
+      "loss": 0.7782,
+      "step": 9663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472392006635538e-05,
+      "loss": 1.1161,
+      "step": 9664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7471825268921376e-05,
+      "loss": 1.0573,
+      "step": 9665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747125847687127e-05,
+      "loss": 0.8823,
+      "step": 9666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747069163048934e-05,
+      "loss": 0.8542,
+      "step": 9667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7470124729779714e-05,
+      "loss": 1.1047,
+      "step": 9668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7469557774746507e-05,
+      "loss": 1.284,
+      "step": 9669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7468990765393853e-05,
+      "loss": 0.7139,
+      "step": 9670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7468423701725865e-05,
+      "loss": 1.0181,
+      "step": 9671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746785658374667e-05,
+      "loss": 1.2454,
+      "step": 9672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7467289411460394e-05,
+      "loss": 1.0573,
+      "step": 9673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7466722184871158e-05,
+      "loss": 1.0095,
+      "step": 9674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7466154903983092e-05,
+      "loss": 0.9665,
+      "step": 9675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746558756880032e-05,
+      "loss": 0.588,
+      "step": 9676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7465020179326962e-05,
+      "loss": 1.1902,
+      "step": 9677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746445273556715e-05,
+      "loss": 1.0516,
+      "step": 9678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463885237525006e-05,
+      "loss": 1.1258,
+      "step": 9679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463317685204662e-05,
+      "loss": 1.0486,
+      "step": 9680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7462750078610246e-05,
+      "loss": 1.1981,
+      "step": 9681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746218241774588e-05,
+      "loss": 0.8733,
+      "step": 9682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7461614702615694e-05,
+      "loss": 0.9355,
+      "step": 9683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746104693322382e-05,
+      "loss": 1.155,
+      "step": 9684
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.746047910957438e-05,
+      "loss": 1.3119,
+      "step": 9685
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745991123167151e-05,
+      "loss": 1.0779,
+      "step": 9686
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7459343299519336e-05,
+      "loss": 1.0669,
+      "step": 9687
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745877531312199e-05,
+      "loss": 1.353,
+      "step": 9688
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7458207272483602e-05,
+      "loss": 0.8899,
+      "step": 9689
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74576391776083e-05,
+      "loss": 1.1077,
+      "step": 9690
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7457071028500223e-05,
+      "loss": 0.6786,
+      "step": 9691
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7456502825163492e-05,
+      "loss": 1.0691,
+      "step": 9692
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7455934567602246e-05,
+      "loss": 1.1532,
+      "step": 9693
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745536625582062e-05,
+      "loss": 1.1681,
+      "step": 9694
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454797889822737e-05,
+      "loss": 1.1391,
+      "step": 9695
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454229469612734e-05,
+      "loss": 1.0768,
+      "step": 9696
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7453660995194754e-05,
+      "loss": 1.2018,
+      "step": 9697
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7453092466572922e-05,
+      "loss": 0.8992,
+      "step": 9698
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745252388375137e-05,
+      "loss": 0.7986,
+      "step": 9699
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745195524673424e-05,
+      "loss": 0.7651,
+      "step": 9700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7451386555525663e-05,
+      "loss": 0.9199,
+      "step": 9701
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450817810129774e-05,
+      "loss": 1.1802,
+      "step": 9702
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450249010550712e-05,
+      "loss": 0.688,
+      "step": 9703
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744968015679261e-05,
+      "loss": 1.0641,
+      "step": 9704
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744911124885961e-05,
+      "loss": 1.0826,
+      "step": 9705
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7448542286755844e-05,
+      "loss": 0.9021,
+      "step": 9706
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744797327048545e-05,
+      "loss": 1.0848,
+      "step": 9707
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7447404200052573e-05,
+      "loss": 1.074,
+      "step": 9708
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7446835075461338e-05,
+      "loss": 1.0204,
+      "step": 9709
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7446265896715895e-05,
+      "loss": 0.7889,
+      "step": 9710
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744569666382038e-05,
+      "loss": 0.7631,
+      "step": 9711
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744512737677893e-05,
+      "loss": 1.0475,
+      "step": 9712
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744455803559569e-05,
+      "loss": 1.1953,
+      "step": 9713
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443988640274795e-05,
+      "loss": 0.9667,
+      "step": 9714
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7443419190820387e-05,
+      "loss": 1.1088,
+      "step": 9715
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744284968723661e-05,
+      "loss": 1.0535,
+      "step": 9716
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7442280129527603e-05,
+      "loss": 0.9775,
+      "step": 9717
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744171051769751e-05,
+      "loss": 0.9744,
+      "step": 9718
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441140851750468e-05,
+      "loss": 1.034,
+      "step": 9719
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7440571131690626e-05,
+      "loss": 0.9656,
+      "step": 9720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7440001357522126e-05,
+      "loss": 0.9996,
+      "step": 9721
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439431529249105e-05,
+      "loss": 0.8531,
+      "step": 9722
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7438861646875718e-05,
+      "loss": 1.5115,
+      "step": 9723
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74382917104061e-05,
+      "loss": 0.9672,
+      "step": 9724
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7437721719844396e-05,
+      "loss": 0.9858,
+      "step": 9725
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7437151675194756e-05,
+      "loss": 0.8823,
+      "step": 9726
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7436581576461326e-05,
+      "loss": 0.9628,
+      "step": 9727
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7436011423648245e-05,
+      "loss": 1.1996,
+      "step": 9728
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7435441216759666e-05,
+      "loss": 1.2325,
+      "step": 9729
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743487095579973e-05,
+      "loss": 0.9508,
+      "step": 9730
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434300640772588e-05,
+      "loss": 0.7302,
+      "step": 9731
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7433730271682387e-05,
+      "loss": 0.9237,
+      "step": 9732
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7433159848533275e-05,
+      "loss": 1.3157,
+      "step": 9733
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432589371329394e-05,
+      "loss": 1.3755,
+      "step": 9734
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432018840074905e-05,
+      "loss": 1.0589,
+      "step": 9735
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7431448254773943e-05,
+      "loss": 0.9203,
+      "step": 9736
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7430877615430666e-05,
+      "loss": 1.0251,
+      "step": 9737
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7430306922049225e-05,
+      "loss": 0.7826,
+      "step": 9738
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7429736174633764e-05,
+      "loss": 0.7482,
+      "step": 9739
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742916537318844e-05,
+      "loss": 1.1669,
+      "step": 9740
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7428594517717394e-05,
+      "loss": 0.9804,
+      "step": 9741
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7428023608224787e-05,
+      "loss": 0.8932,
+      "step": 9742
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742745264471477e-05,
+      "loss": 0.8794,
+      "step": 9743
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7426881627191487e-05,
+      "loss": 1.136,
+      "step": 9744
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74263105556591e-05,
+      "loss": 1.0236,
+      "step": 9745
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742573943012176e-05,
+      "loss": 0.9198,
+      "step": 9746
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7425168250583614e-05,
+      "loss": 1.2622,
+      "step": 9747
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7424597017048823e-05,
+      "loss": 0.862,
+      "step": 9748
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7424025729521535e-05,
+      "loss": 0.8837,
+      "step": 9749
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742345438800591e-05,
+      "loss": 0.685,
+      "step": 9750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7422882992506102e-05,
+      "loss": 1.2554,
+      "step": 9751
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742231154302626e-05,
+      "loss": 0.9628,
+      "step": 9752
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7421740039570546e-05,
+      "loss": 0.8541,
+      "step": 9753
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7421168482143115e-05,
+      "loss": 0.6071,
+      "step": 9754
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742059687074812e-05,
+      "loss": 1.0476,
+      "step": 9755
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7420025205389722e-05,
+      "loss": 1.133,
+      "step": 9756
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741945348607208e-05,
+      "loss": 0.9134,
+      "step": 9757
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741888171279935e-05,
+      "loss": 1.0362,
+      "step": 9758
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741830988557568e-05,
+      "loss": 0.9729,
+      "step": 9759
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741773800440524e-05,
+      "loss": 1.3694,
+      "step": 9760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7417166069292188e-05,
+      "loss": 1.0846,
+      "step": 9761
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741659408024068e-05,
+      "loss": 0.7383,
+      "step": 9762
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741602203725487e-05,
+      "loss": 0.8478,
+      "step": 9763
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7415449940338932e-05,
+      "loss": 0.5913,
+      "step": 9764
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414877789497017e-05,
+      "loss": 1.0869,
+      "step": 9765
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414305584733285e-05,
+      "loss": 0.9441,
+      "step": 9766
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413733326051902e-05,
+      "loss": 0.796,
+      "step": 9767
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413161013457024e-05,
+      "loss": 0.8297,
+      "step": 9768
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741258864695282e-05,
+      "loss": 1.3032,
+      "step": 9769
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7412016226543443e-05,
+      "loss": 1.0613,
+      "step": 9770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411443752233066e-05,
+      "loss": 0.966,
+      "step": 9771
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7410871224025846e-05,
+      "loss": 1.6225,
+      "step": 9772
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7410298641925945e-05,
+      "loss": 0.9686,
+      "step": 9773
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740972600593753e-05,
+      "loss": 0.9795,
+      "step": 9774
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7409153316064765e-05,
+      "loss": 0.8576,
+      "step": 9775
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408580572311814e-05,
+      "loss": 1.0021,
+      "step": 9776
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408007774682845e-05,
+      "loss": 0.9901,
+      "step": 9777
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740743492318202e-05,
+      "loss": 1.0917,
+      "step": 9778
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7406862017813503e-05,
+      "loss": 0.8908,
+      "step": 9779
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7406289058581466e-05,
+      "loss": 1.2698,
+      "step": 9780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740571604549007e-05,
+      "loss": 1.2509,
+      "step": 9781
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7405142978543485e-05,
+      "loss": 1.01,
+      "step": 9782
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740456985774588e-05,
+      "loss": 0.9126,
+      "step": 9783
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7403996683101417e-05,
+      "loss": 0.7731,
+      "step": 9784
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.740342345461427e-05,
+      "loss": 0.896,
+      "step": 9785
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7402850172288604e-05,
+      "loss": 1.375,
+      "step": 9786
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7402276836128593e-05,
+      "loss": 1.2337,
+      "step": 9787
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7401703446138404e-05,
+      "loss": 1.4511,
+      "step": 9788
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74011300023222e-05,
+      "loss": 0.946,
+      "step": 9789
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7400556504684162e-05,
+      "loss": 1.0396,
+      "step": 9790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739998295322845e-05,
+      "loss": 0.9047,
+      "step": 9791
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739940934795924e-05,
+      "loss": 1.5239,
+      "step": 9792
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7398835688880708e-05,
+      "loss": 1.1102,
+      "step": 9793
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739826197599702e-05,
+      "loss": 0.9743,
+      "step": 9794
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739768820931235e-05,
+      "loss": 1.1263,
+      "step": 9795
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7397114388830866e-05,
+      "loss": 1.0732,
+      "step": 9796
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739654051455675e-05,
+      "loss": 1.3139,
+      "step": 9797
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395966586494165e-05,
+      "loss": 1.2181,
+      "step": 9798
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7395392604647294e-05,
+      "loss": 1.0616,
+      "step": 9799
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7394818569020306e-05,
+      "loss": 1.1817,
+      "step": 9800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7394244479617374e-05,
+      "loss": 1.2616,
+      "step": 9801
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7393670336442676e-05,
+      "loss": 1.3022,
+      "step": 9802
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7393096139500387e-05,
+      "loss": 0.8802,
+      "step": 9803
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739252188879468e-05,
+      "loss": 0.7331,
+      "step": 9804
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391947584329735e-05,
+      "loss": 1.0308,
+      "step": 9805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391373226109726e-05,
+      "loss": 1.032,
+      "step": 9806
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739079881413883e-05,
+      "loss": 0.9609,
+      "step": 9807
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7390224348421223e-05,
+      "loss": 0.9414,
+      "step": 9808
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7389649828961084e-05,
+      "loss": 0.99,
+      "step": 9809
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7389075255762592e-05,
+      "loss": 0.8496,
+      "step": 9810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7388500628829926e-05,
+      "loss": 1.1262,
+      "step": 9811
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738792594816726e-05,
+      "loss": 1.1075,
+      "step": 9812
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7387351213778778e-05,
+      "loss": 1.058,
+      "step": 9813
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7386776425668654e-05,
+      "loss": 1.0415,
+      "step": 9814
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7386201583841073e-05,
+      "loss": 0.8789,
+      "step": 9815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385626688300216e-05,
+      "loss": 0.8881,
+      "step": 9816
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738505173905026e-05,
+      "loss": 0.7743,
+      "step": 9817
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738447673609539e-05,
+      "loss": 0.7489,
+      "step": 9818
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738390167943978e-05,
+      "loss": 1.1519,
+      "step": 9819
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7383326569087625e-05,
+      "loss": 0.6995,
+      "step": 9820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382751405043093e-05,
+      "loss": 1.2385,
+      "step": 9821
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382176187310374e-05,
+      "loss": 1.0715,
+      "step": 9822
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7381600915893653e-05,
+      "loss": 0.8807,
+      "step": 9823
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7381025590797105e-05,
+      "loss": 0.6748,
+      "step": 9824
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7380450212024924e-05,
+      "loss": 1.1969,
+      "step": 9825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7379874779581286e-05,
+      "loss": 0.9724,
+      "step": 9826
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7379299293470382e-05,
+      "loss": 0.6916,
+      "step": 9827
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378723753696394e-05,
+      "loss": 1.2364,
+      "step": 9828
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378148160263503e-05,
+      "loss": 0.792,
+      "step": 9829
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7377572513175905e-05,
+      "loss": 0.8057,
+      "step": 9830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376996812437776e-05,
+      "loss": 1.0547,
+      "step": 9831
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737642105805331e-05,
+      "loss": 0.8585,
+      "step": 9832
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737584525002669e-05,
+      "loss": 0.87,
+      "step": 9833
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7375269388362107e-05,
+      "loss": 1.2468,
+      "step": 9834
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7374693473063743e-05,
+      "loss": 0.9793,
+      "step": 9835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737411750413579e-05,
+      "loss": 0.7092,
+      "step": 9836
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373541481582433e-05,
+      "loss": 1.0177,
+      "step": 9837
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7372965405407867e-05,
+      "loss": 1.1992,
+      "step": 9838
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7372389275616274e-05,
+      "loss": 0.6626,
+      "step": 9839
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737181309221185e-05,
+      "loss": 0.8386,
+      "step": 9840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7371236855198784e-05,
+      "loss": 1.1141,
+      "step": 9841
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370660564581263e-05,
+      "loss": 1.0777,
+      "step": 9842
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370084220363483e-05,
+      "loss": 0.8863,
+      "step": 9843
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7369507822549625e-05,
+      "loss": 0.8988,
+      "step": 9844
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7368931371143895e-05,
+      "loss": 1.4476,
+      "step": 9845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7368354866150473e-05,
+      "loss": 0.769,
+      "step": 9846
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736777830757356e-05,
+      "loss": 0.8905,
+      "step": 9847
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736720169541734e-05,
+      "loss": 0.7461,
+      "step": 9848
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7366625029686014e-05,
+      "loss": 1.2738,
+      "step": 9849
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7366048310383776e-05,
+      "loss": 1.0972,
+      "step": 9850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7365471537514812e-05,
+      "loss": 0.979,
+      "step": 9851
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364894711083323e-05,
+      "loss": 0.8538,
+      "step": 9852
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364317831093504e-05,
+      "loss": 1.1504,
+      "step": 9853
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7363740897549542e-05,
+      "loss": 0.976,
+      "step": 9854
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7363163910455646e-05,
+      "loss": 1.2069,
+      "step": 9855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7362586869816e-05,
+      "loss": 0.8186,
+      "step": 9856
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7362009775634804e-05,
+      "loss": 1.0416,
+      "step": 9857
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7361432627916258e-05,
+      "loss": 1.0312,
+      "step": 9858
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7360855426664555e-05,
+      "loss": 1.2809,
+      "step": 9859
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736027817188389e-05,
+      "loss": 0.9439,
+      "step": 9860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735970086357847e-05,
+      "loss": 1.1844,
+      "step": 9861
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735912350175249e-05,
+      "loss": 1.0777,
+      "step": 9862
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735854608641014e-05,
+      "loss": 1.0527,
+      "step": 9863
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735796861755563e-05,
+      "loss": 1.2993,
+      "step": 9864
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357391095193154e-05,
+      "loss": 1.0235,
+      "step": 9865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7356813519326913e-05,
+      "loss": 0.9865,
+      "step": 9866
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735623588996111e-05,
+      "loss": 0.9739,
+      "step": 9867
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7355658207099937e-05,
+      "loss": 0.8502,
+      "step": 9868
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7355080470747605e-05,
+      "loss": 0.901,
+      "step": 9869
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735450268090831e-05,
+      "loss": 0.9705,
+      "step": 9870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7353924837586252e-05,
+      "loss": 0.8042,
+      "step": 9871
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735334694078564e-05,
+      "loss": 0.7162,
+      "step": 9872
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7352768990510672e-05,
+      "loss": 0.9828,
+      "step": 9873
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735219098676555e-05,
+      "loss": 0.8449,
+      "step": 9874
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735161292955448e-05,
+      "loss": 0.9719,
+      "step": 9875
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735103481888166e-05,
+      "loss": 1.1109,
+      "step": 9876
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7350456654751307e-05,
+      "loss": 0.8688,
+      "step": 9877
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734987843716761e-05,
+      "loss": 1.0735,
+      "step": 9878
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7349300166134782e-05,
+      "loss": 0.9388,
+      "step": 9879
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734872184165703e-05,
+      "loss": 0.7917,
+      "step": 9880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7348143463738553e-05,
+      "loss": 1.0024,
+      "step": 9881
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734756503238356e-05,
+      "loss": 1.0617,
+      "step": 9882
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7346986547596263e-05,
+      "loss": 1.144,
+      "step": 9883
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734640800938086e-05,
+      "loss": 1.2146,
+      "step": 9884
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7345829417741564e-05,
+      "loss": 0.801,
+      "step": 9885
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7345250772682582e-05,
+      "loss": 1.1272,
+      "step": 9886
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7344672074208116e-05,
+      "loss": 1.1552,
+      "step": 9887
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7344093322322386e-05,
+      "loss": 0.9043,
+      "step": 9888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7343514517029588e-05,
+      "loss": 0.9096,
+      "step": 9889
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7342935658333938e-05,
+      "loss": 0.9662,
+      "step": 9890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7342356746239647e-05,
+      "loss": 0.9028,
+      "step": 9891
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734177778075092e-05,
+      "loss": 0.9114,
+      "step": 9892
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734119876187197e-05,
+      "loss": 0.9664,
+      "step": 9893
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734061968960701e-05,
+      "loss": 0.8,
+      "step": 9894
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7340040563960245e-05,
+      "loss": 1.391,
+      "step": 9895
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733946138493589e-05,
+      "loss": 0.9759,
+      "step": 9896
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733888215253816e-05,
+      "loss": 0.7815,
+      "step": 9897
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733830286677126e-05,
+      "loss": 0.8357,
+      "step": 9898
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733772352763941e-05,
+      "loss": 0.748,
+      "step": 9899
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7337144135146818e-05,
+      "loss": 1.0772,
+      "step": 9900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73365646892977e-05,
+      "loss": 0.861,
+      "step": 9901
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7335985190096267e-05,
+      "loss": 0.7832,
+      "step": 9902
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7335405637546738e-05,
+      "loss": 0.8561,
+      "step": 9903
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733482603165332e-05,
+      "loss": 0.7581,
+      "step": 9904
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7334246372420234e-05,
+      "loss": 1.1034,
+      "step": 9905
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73336666598517e-05,
+      "loss": 1.0779,
+      "step": 9906
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7333086893951922e-05,
+      "loss": 0.9335,
+      "step": 9907
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733250707472512e-05,
+      "loss": 1.074,
+      "step": 9908
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733192720217552e-05,
+      "loss": 0.8497,
+      "step": 9909
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7331347276307324e-05,
+      "loss": 1.078,
+      "step": 9910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7330767297124758e-05,
+      "loss": 0.8884,
+      "step": 9911
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7330187264632043e-05,
+      "loss": 1.0611,
+      "step": 9912
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7329607178833385e-05,
+      "loss": 0.938,
+      "step": 9913
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732902703973301e-05,
+      "loss": 1.1807,
+      "step": 9914
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328446847335142e-05,
+      "loss": 1.0644,
+      "step": 9915
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7327866601643993e-05,
+      "loss": 0.8987,
+      "step": 9916
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7327286302663782e-05,
+      "loss": 1.093,
+      "step": 9917
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7326705950398734e-05,
+      "loss": 1.0479,
+      "step": 9918
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7326125544853064e-05,
+      "loss": 0.7624,
+      "step": 9919
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7325545086031e-05,
+      "loss": 1.1634,
+      "step": 9920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7324964573936754e-05,
+      "loss": 0.9259,
+      "step": 9921
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7324384008574554e-05,
+      "loss": 1.0276,
+      "step": 9922
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7323803389948622e-05,
+      "loss": 0.9224,
+      "step": 9923
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7323222718063176e-05,
+      "loss": 0.6139,
+      "step": 9924
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732264199292244e-05,
+      "loss": 0.8478,
+      "step": 9925
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7322061214530638e-05,
+      "loss": 0.8098,
+      "step": 9926
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7321480382892e-05,
+      "loss": 0.8329,
+      "step": 9927
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7320899498010737e-05,
+      "loss": 1.1964,
+      "step": 9928
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7320318559891082e-05,
+      "loss": 0.9801,
+      "step": 9929
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731973756853726e-05,
+      "loss": 1.0253,
+      "step": 9930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7319156523953488e-05,
+      "loss": 0.9703,
+      "step": 9931
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7318575426144002e-05,
+      "loss": 1.4831,
+      "step": 9932
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731799427511302e-05,
+      "loss": 0.8656,
+      "step": 9933
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7317413070864772e-05,
+      "loss": 0.9512,
+      "step": 9934
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7316831813403484e-05,
+      "loss": 0.7822,
+      "step": 9935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7316250502733382e-05,
+      "loss": 1.036,
+      "step": 9936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315669138858695e-05,
+      "loss": 0.8324,
+      "step": 9937
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315087721783646e-05,
+      "loss": 0.7978,
+      "step": 9938
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731450625151247e-05,
+      "loss": 1.043,
+      "step": 9939
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313924728049395e-05,
+      "loss": 0.659,
+      "step": 9940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313343151398642e-05,
+      "loss": 0.8111,
+      "step": 9941
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731276152156445e-05,
+      "loss": 1.0097,
+      "step": 9942
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7312179838551036e-05,
+      "loss": 0.877,
+      "step": 9943
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7311598102362648e-05,
+      "loss": 0.8465,
+      "step": 9944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73110163130035e-05,
+      "loss": 0.8954,
+      "step": 9945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7310434470477833e-05,
+      "loss": 0.5655,
+      "step": 9946
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7309852574789877e-05,
+      "loss": 0.9355,
+      "step": 9947
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7309270625943855e-05,
+      "loss": 0.9074,
+      "step": 9948
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308688623944008e-05,
+      "loss": 1.2301,
+      "step": 9949
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308106568794568e-05,
+      "loss": 0.812,
+      "step": 9950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730752446049976e-05,
+      "loss": 1.4007,
+      "step": 9951
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7306942299063827e-05,
+      "loss": 0.8462,
+      "step": 9952
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7306360084491e-05,
+      "loss": 0.8974,
+      "step": 9953
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305777816785506e-05,
+      "loss": 0.8318,
+      "step": 9954
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730519549595159e-05,
+      "loss": 0.489,
+      "step": 9955
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7304613121993476e-05,
+      "loss": 0.8745,
+      "step": 9956
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7304030694915406e-05,
+      "loss": 1.0271,
+      "step": 9957
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730344821472161e-05,
+      "loss": 1.7197,
+      "step": 9958
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302865681416333e-05,
+      "loss": 1.109,
+      "step": 9959
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302283095003807e-05,
+      "loss": 0.7655,
+      "step": 9960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7301700455488265e-05,
+      "loss": 1.3062,
+      "step": 9961
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7301117762873945e-05,
+      "loss": 1.1584,
+      "step": 9962
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730053501716509e-05,
+      "loss": 0.8074,
+      "step": 9963
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729995221836593e-05,
+      "loss": 0.8088,
+      "step": 9964
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7299369366480708e-05,
+      "loss": 0.6109,
+      "step": 9965
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298786461513656e-05,
+      "loss": 1.1766,
+      "step": 9966
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298203503469023e-05,
+      "loss": 1.1262,
+      "step": 9967
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7297620492351046e-05,
+      "loss": 0.8609,
+      "step": 9968
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7297037428163962e-05,
+      "loss": 0.6185,
+      "step": 9969
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729645431091201e-05,
+      "loss": 1.0853,
+      "step": 9970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295871140599428e-05,
+      "loss": 0.9138,
+      "step": 9971
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295287917230465e-05,
+      "loss": 1.1538,
+      "step": 9972
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7294704640809363e-05,
+      "loss": 1.7391,
+      "step": 9973
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729412131134035e-05,
+      "loss": 0.8731,
+      "step": 9974
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729353792882768e-05,
+      "loss": 1.3545,
+      "step": 9975
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7292954493275593e-05,
+      "loss": 1.0731,
+      "step": 9976
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7292371004688332e-05,
+      "loss": 0.8533,
+      "step": 9977
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7291787463070138e-05,
+      "loss": 0.8051,
+      "step": 9978
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7291203868425255e-05,
+      "loss": 1.231,
+      "step": 9979
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290620220757926e-05,
+      "loss": 0.9594,
+      "step": 9980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290036520072403e-05,
+      "loss": 0.9831,
+      "step": 9981
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7289452766372923e-05,
+      "loss": 0.5216,
+      "step": 9982
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728886895966373e-05,
+      "loss": 0.8439,
+      "step": 9983
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288285099949072e-05,
+      "loss": 0.9521,
+      "step": 9984
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72877011872332e-05,
+      "loss": 1.0731,
+      "step": 9985
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728711722152035e-05,
+      "loss": 0.6167,
+      "step": 9986
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728653320281478e-05,
+      "loss": 0.906,
+      "step": 9987
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7285949131120728e-05,
+      "loss": 0.9228,
+      "step": 9988
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7285365006442448e-05,
+      "loss": 0.813,
+      "step": 9989
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728478082878418e-05,
+      "loss": 1.1181,
+      "step": 9990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728419659815018e-05,
+      "loss": 0.8927,
+      "step": 9991
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7283612314544697e-05,
+      "loss": 1.0571,
+      "step": 9992
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728302797797197e-05,
+      "loss": 1.3251,
+      "step": 9993
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728244358843626e-05,
+      "loss": 0.8822,
+      "step": 9994
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7281859145941804e-05,
+      "loss": 1.3364,
+      "step": 9995
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7281274650492866e-05,
+      "loss": 0.8158,
+      "step": 9996
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728069010209369e-05,
+      "loss": 0.9537,
+      "step": 9997
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7280105500748522e-05,
+      "loss": 0.9001,
+      "step": 9998
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7279520846461624e-05,
+      "loss": 1.1757,
+      "step": 9999
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278936139237238e-05,
+      "loss": 0.8084,
+      "step": 10000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278351379079624e-05,
+      "loss": 1.1823,
+      "step": 10001
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727776656599303e-05,
+      "loss": 0.8036,
+      "step": 10002
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277181699981708e-05,
+      "loss": 0.9571,
+      "step": 10003
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727659678104991e-05,
+      "loss": 1.0302,
+      "step": 10004
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7276011809201896e-05,
+      "loss": 1.0194,
+      "step": 10005
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7275426784441914e-05,
+      "loss": 1.2954,
+      "step": 10006
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727484170677422e-05,
+      "loss": 1.0219,
+      "step": 10007
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727425657620307e-05,
+      "loss": 0.8681,
+      "step": 10008
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727367139273272e-05,
+      "loss": 0.9045,
+      "step": 10009
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7273086156367423e-05,
+      "loss": 1.068,
+      "step": 10010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7272500867111434e-05,
+      "loss": 1.0692,
+      "step": 10011
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271915524969016e-05,
+      "loss": 1.0241,
+      "step": 10012
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271330129944416e-05,
+      "loss": 1.2327,
+      "step": 10013
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7270744682041897e-05,
+      "loss": 1.022,
+      "step": 10014
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7270159181265715e-05,
+      "loss": 1.0888,
+      "step": 10015
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726957362762013e-05,
+      "loss": 0.9951,
+      "step": 10016
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268988021109397e-05,
+      "loss": 0.7804,
+      "step": 10017
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268402361737774e-05,
+      "loss": 1.1563,
+      "step": 10018
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7267816649509524e-05,
+      "loss": 1.1452,
+      "step": 10019
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7267230884428905e-05,
+      "loss": 1.2065,
+      "step": 10020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7266645066500175e-05,
+      "loss": 1.0048,
+      "step": 10021
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7266059195727595e-05,
+      "loss": 0.7625,
+      "step": 10022
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7265473272115425e-05,
+      "loss": 1.1933,
+      "step": 10023
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726488729566793e-05,
+      "loss": 1.1425,
+      "step": 10024
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7264301266389364e-05,
+      "loss": 0.8262,
+      "step": 10025
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7263715184283994e-05,
+      "loss": 0.9867,
+      "step": 10026
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726312904935608e-05,
+      "loss": 1.191,
+      "step": 10027
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7262542861609884e-05,
+      "loss": 0.8822,
+      "step": 10028
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726195662104967e-05,
+      "loss": 0.9334,
+      "step": 10029
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72613703276797e-05,
+      "loss": 0.9433,
+      "step": 10030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7260783981504245e-05,
+      "loss": 0.8011,
+      "step": 10031
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726019758252756e-05,
+      "loss": 1.2413,
+      "step": 10032
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7259611130753908e-05,
+      "loss": 1.2839,
+      "step": 10033
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725902462618756e-05,
+      "loss": 1.3658,
+      "step": 10034
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725843806883278e-05,
+      "loss": 0.961,
+      "step": 10035
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257851458693832e-05,
+      "loss": 1.1943,
+      "step": 10036
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257264795774982e-05,
+      "loss": 0.9096,
+      "step": 10037
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7256678080080494e-05,
+      "loss": 1.1082,
+      "step": 10038
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725609131161464e-05,
+      "loss": 1.0961,
+      "step": 10039
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7255504490381683e-05,
+      "loss": 0.785,
+      "step": 10040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7254917616385892e-05,
+      "loss": 0.7164,
+      "step": 10041
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7254330689631532e-05,
+      "loss": 0.6286,
+      "step": 10042
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7253743710122877e-05,
+      "loss": 1.0358,
+      "step": 10043
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7253156677864194e-05,
+      "loss": 1.2473,
+      "step": 10044
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7252569592859746e-05,
+      "loss": 1.3077,
+      "step": 10045
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725198245511381e-05,
+      "loss": 1.0723,
+      "step": 10046
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725139526463065e-05,
+      "loss": 0.8171,
+      "step": 10047
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7250808021414537e-05,
+      "loss": 0.9395,
+      "step": 10048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7250220725469744e-05,
+      "loss": 1.0227,
+      "step": 10049
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7249633376800542e-05,
+      "loss": 1.0997,
+      "step": 10050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72490459754112e-05,
+      "loss": 1.283,
+      "step": 10051
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724845852130599e-05,
+      "loss": 1.1705,
+      "step": 10052
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724787101448919e-05,
+      "loss": 1.2324,
+      "step": 10053
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724728345496506e-05,
+      "loss": 0.9573,
+      "step": 10054
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7246695842737882e-05,
+      "loss": 1.1328,
+      "step": 10055
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724610817781193e-05,
+      "loss": 1.2376,
+      "step": 10056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724552046019147e-05,
+      "loss": 1.0586,
+      "step": 10057
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244932689880786e-05,
+      "loss": 1.0462,
+      "step": 10058
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724434486688414e-05,
+      "loss": 1.0885,
+      "step": 10059
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724375699120582e-05,
+      "loss": 1.0854,
+      "step": 10060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7243169062850095e-05,
+      "loss": 1.0028,
+      "step": 10061
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7242581081821236e-05,
+      "loss": 0.8562,
+      "step": 10062
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241993048123526e-05,
+      "loss": 0.9807,
+      "step": 10063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241404961761237e-05,
+      "loss": 1.0481,
+      "step": 10064
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7240816822738646e-05,
+      "loss": 1.1868,
+      "step": 10065
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.724022863106003e-05,
+      "loss": 1.1902,
+      "step": 10066
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7239640386729673e-05,
+      "loss": 0.9337,
+      "step": 10067
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7239052089751843e-05,
+      "loss": 0.7034,
+      "step": 10068
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7238463740130823e-05,
+      "loss": 0.7345,
+      "step": 10069
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723787533787089e-05,
+      "loss": 0.9192,
+      "step": 10070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7237286882976326e-05,
+      "loss": 1.2399,
+      "step": 10071
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236698375451408e-05,
+      "loss": 0.987,
+      "step": 10072
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236109815300414e-05,
+      "loss": 1.379,
+      "step": 10073
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7235521202527623e-05,
+      "loss": 0.7758,
+      "step": 10074
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7234932537137325e-05,
+      "loss": 0.9945,
+      "step": 10075
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723434381913379e-05,
+      "loss": 1.0014,
+      "step": 10076
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7233755048521304e-05,
+      "loss": 0.6997,
+      "step": 10077
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723316622530415e-05,
+      "loss": 1.4483,
+      "step": 10078
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7232577349486606e-05,
+      "loss": 1.3299,
+      "step": 10079
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231988421072957e-05,
+      "loss": 1.0864,
+      "step": 10080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231399440067484e-05,
+      "loss": 0.7703,
+      "step": 10081
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723081040647447e-05,
+      "loss": 0.9026,
+      "step": 10082
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7230221320298205e-05,
+      "loss": 1.0907,
+      "step": 10083
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7229632181542965e-05,
+      "loss": 1.1107,
+      "step": 10084
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7229042990213037e-05,
+      "loss": 1.232,
+      "step": 10085
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72284537463127e-05,
+      "loss": 0.9596,
+      "step": 10086
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7227864449846255e-05,
+      "loss": 0.996,
+      "step": 10087
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722727510081797e-05,
+      "loss": 1.1234,
+      "step": 10088
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7226685699232137e-05,
+      "loss": 1.0983,
+      "step": 10089
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7226096245093046e-05,
+      "loss": 0.9887,
+      "step": 10090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7225506738404983e-05,
+      "loss": 1.0538,
+      "step": 10091
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722491717917223e-05,
+      "loss": 0.8582,
+      "step": 10092
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7224327567399078e-05,
+      "loss": 1.0692,
+      "step": 10093
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722373790308981e-05,
+      "loss": 0.7598,
+      "step": 10094
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722314818624872e-05,
+      "loss": 1.6104,
+      "step": 10095
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222558416880098e-05,
+      "loss": 0.9895,
+      "step": 10096
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7221968594988226e-05,
+      "loss": 1.1617,
+      "step": 10097
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7221378720577397e-05,
+      "loss": 0.7955,
+      "step": 10098
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7220788793651898e-05,
+      "loss": 0.6457,
+      "step": 10099
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722019881421602e-05,
+      "loss": 0.7945,
+      "step": 10100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721960878227406e-05,
+      "loss": 0.9922,
+      "step": 10101
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72190186978303e-05,
+      "loss": 1.5568,
+      "step": 10102
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218428560889034e-05,
+      "loss": 1.1189,
+      "step": 10103
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7217838371454553e-05,
+      "loss": 1.062,
+      "step": 10104
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721724812953115e-05,
+      "loss": 1.122,
+      "step": 10105
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721665783512312e-05,
+      "loss": 0.9438,
+      "step": 10106
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7216067488234754e-05,
+      "loss": 0.8643,
+      "step": 10107
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721547708887034e-05,
+      "loss": 0.8576,
+      "step": 10108
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214886637034176e-05,
+      "loss": 1.1921,
+      "step": 10109
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7214296132730555e-05,
+      "loss": 0.7095,
+      "step": 10110
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721370557596377e-05,
+      "loss": 0.9296,
+      "step": 10111
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721311496673812e-05,
+      "loss": 1.108,
+      "step": 10112
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7212524305057897e-05,
+      "loss": 0.7724,
+      "step": 10113
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211933590927396e-05,
+      "loss": 0.7874,
+      "step": 10114
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211342824350913e-05,
+      "loss": 0.7574,
+      "step": 10115
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210752005332743e-05,
+      "loss": 1.3508,
+      "step": 10116
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210161133877183e-05,
+      "loss": 0.9274,
+      "step": 10117
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7209570209988534e-05,
+      "loss": 0.8427,
+      "step": 10118
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7208979233671088e-05,
+      "loss": 1.4061,
+      "step": 10119
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7208388204929142e-05,
+      "loss": 0.9271,
+      "step": 10120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7207797123766997e-05,
+      "loss": 0.8859,
+      "step": 10121
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7207205990188956e-05,
+      "loss": 0.9128,
+      "step": 10122
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7206614804199308e-05,
+      "loss": 0.8336,
+      "step": 10123
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7206023565802356e-05,
+      "loss": 1.2481,
+      "step": 10124
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7205432275002403e-05,
+      "loss": 1.1175,
+      "step": 10125
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7204840931803745e-05,
+      "loss": 1.0885,
+      "step": 10126
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7204249536210688e-05,
+      "loss": 0.8559,
+      "step": 10127
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7203658088227522e-05,
+      "loss": 0.9412,
+      "step": 10128
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720306658785856e-05,
+      "loss": 1.4724,
+      "step": 10129
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7202475035108094e-05,
+      "loss": 0.9152,
+      "step": 10130
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720188342998043e-05,
+      "loss": 1.0703,
+      "step": 10131
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720129177247987e-05,
+      "loss": 1.2797,
+      "step": 10132
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720070006261072e-05,
+      "loss": 1.0737,
+      "step": 10133
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7200108300377277e-05,
+      "loss": 0.8758,
+      "step": 10134
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7199516485783845e-05,
+      "loss": 1.0306,
+      "step": 10135
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719892461883473e-05,
+      "loss": 1.4806,
+      "step": 10136
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719833269953424e-05,
+      "loss": 1.0559,
+      "step": 10137
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7197740727886673e-05,
+      "loss": 0.8766,
+      "step": 10138
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7197148703896336e-05,
+      "loss": 1.2825,
+      "step": 10139
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719655662756753e-05,
+      "loss": 0.9335,
+      "step": 10140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7195964498904572e-05,
+      "loss": 1.208,
+      "step": 10141
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7195372317911758e-05,
+      "loss": 0.8914,
+      "step": 10142
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7194780084593396e-05,
+      "loss": 1.2046,
+      "step": 10143
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7194187798953795e-05,
+      "loss": 1.1304,
+      "step": 10144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193595460997262e-05,
+      "loss": 0.8478,
+      "step": 10145
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193003070728106e-05,
+      "loss": 0.5481,
+      "step": 10146
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7192410628150634e-05,
+      "loss": 0.9664,
+      "step": 10147
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7191818133269148e-05,
+      "loss": 0.9767,
+      "step": 10148
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7191225586087965e-05,
+      "loss": 1.4579,
+      "step": 10149
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719063298661139e-05,
+      "loss": 0.6974,
+      "step": 10150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7190040334843734e-05,
+      "loss": 1.0311,
+      "step": 10151
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7189447630789305e-05,
+      "loss": 0.803,
+      "step": 10152
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7188854874452415e-05,
+      "loss": 1.1056,
+      "step": 10153
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7188262065837376e-05,
+      "loss": 0.662,
+      "step": 10154
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187669204948495e-05,
+      "loss": 1.0438,
+      "step": 10155
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187076291790085e-05,
+      "loss": 1.1933,
+      "step": 10156
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718648332636646e-05,
+      "loss": 0.9696,
+      "step": 10157
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185890308681927e-05,
+      "loss": 1.1688,
+      "step": 10158
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185297238740805e-05,
+      "loss": 1.2515,
+      "step": 10159
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7184704116547403e-05,
+      "loss": 0.8341,
+      "step": 10160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7184110942106034e-05,
+      "loss": 0.7942,
+      "step": 10161
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718351771542101e-05,
+      "loss": 0.8478,
+      "step": 10162
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7182924436496654e-05,
+      "loss": 1.1761,
+      "step": 10163
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718233110533727e-05,
+      "loss": 0.7899,
+      "step": 10164
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7181737721947174e-05,
+      "loss": 1.143,
+      "step": 10165
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7181144286330686e-05,
+      "loss": 1.0174,
+      "step": 10166
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718055079849212e-05,
+      "loss": 0.7852,
+      "step": 10167
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7179957258435794e-05,
+      "loss": 0.6359,
+      "step": 10168
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7179363666166018e-05,
+      "loss": 0.9188,
+      "step": 10169
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7178770021687113e-05,
+      "loss": 1.2481,
+      "step": 10170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7178176325003396e-05,
+      "loss": 0.9086,
+      "step": 10171
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717758257611918e-05,
+      "loss": 1.0055,
+      "step": 10172
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717698877503879e-05,
+      "loss": 0.8714,
+      "step": 10173
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717639492176654e-05,
+      "loss": 1.1649,
+      "step": 10174
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7175801016306745e-05,
+      "loss": 1.5267,
+      "step": 10175
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7175207058663734e-05,
+      "loss": 1.0681,
+      "step": 10176
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717461304884182e-05,
+      "loss": 1.0715,
+      "step": 10177
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717401898684532e-05,
+      "loss": 0.8553,
+      "step": 10178
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717342487267856e-05,
+      "loss": 0.7396,
+      "step": 10179
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7172830706345855e-05,
+      "loss": 1.057,
+      "step": 10180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717223648785153e-05,
+      "loss": 0.9217,
+      "step": 10181
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7171642217199903e-05,
+      "loss": 1.0621,
+      "step": 10182
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7171047894395303e-05,
+      "loss": 0.8641,
+      "step": 10183
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7170453519442043e-05,
+      "loss": 0.9567,
+      "step": 10184
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7169859092344448e-05,
+      "loss": 0.9806,
+      "step": 10185
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716926461310684e-05,
+      "loss": 0.6374,
+      "step": 10186
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168670081733548e-05,
+      "loss": 1.0983,
+      "step": 10187
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168075498228887e-05,
+      "loss": 1.1194,
+      "step": 10188
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7167480862597188e-05,
+      "loss": 1.1212,
+      "step": 10189
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716688617484277e-05,
+      "loss": 1.0512,
+      "step": 10190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716629143496996e-05,
+      "loss": 1.0633,
+      "step": 10191
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7165696642983086e-05,
+      "loss": 0.9614,
+      "step": 10192
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716510179888647e-05,
+      "loss": 0.9027,
+      "step": 10193
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7164506902684433e-05,
+      "loss": 1.1222,
+      "step": 10194
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716391195438131e-05,
+      "loss": 0.7789,
+      "step": 10195
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716331695398142e-05,
+      "loss": 1.3066,
+      "step": 10196
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7162721901489097e-05,
+      "loss": 0.9152,
+      "step": 10197
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7162126796908665e-05,
+      "loss": 0.9365,
+      "step": 10198
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7161531640244452e-05,
+      "loss": 0.8865,
+      "step": 10199
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7160936431500785e-05,
+      "loss": 0.9804,
+      "step": 10200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7160341170681992e-05,
+      "loss": 0.9097,
+      "step": 10201
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7159745857792404e-05,
+      "loss": 1.035,
+      "step": 10202
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715915049283635e-05,
+      "loss": 1.14,
+      "step": 10203
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715855507581816e-05,
+      "loss": 0.6007,
+      "step": 10204
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157959606742157e-05,
+      "loss": 1.0952,
+      "step": 10205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157364085612677e-05,
+      "loss": 0.9397,
+      "step": 10206
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7156768512434056e-05,
+      "loss": 1.0955,
+      "step": 10207
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7156172887210617e-05,
+      "loss": 1.2686,
+      "step": 10208
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7155577209946694e-05,
+      "loss": 1.0898,
+      "step": 10209
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715498148064662e-05,
+      "loss": 1.3074,
+      "step": 10210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7154385699314723e-05,
+      "loss": 0.9637,
+      "step": 10211
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7153789865955343e-05,
+      "loss": 0.8156,
+      "step": 10212
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7153193980572805e-05,
+      "loss": 1.1682,
+      "step": 10213
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7152598043171447e-05,
+      "loss": 0.6533,
+      "step": 10214
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7152002053755604e-05,
+      "loss": 1.0149,
+      "step": 10215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7151406012329604e-05,
+      "loss": 0.7031,
+      "step": 10216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7150809918897787e-05,
+      "loss": 0.9885,
+      "step": 10217
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715021377346449e-05,
+      "loss": 1.0382,
+      "step": 10218
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714961757603404e-05,
+      "loss": 1.2266,
+      "step": 10219
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7149021326610777e-05,
+      "loss": 0.9376,
+      "step": 10220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714842502519904e-05,
+      "loss": 1.3457,
+      "step": 10221
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714782867180316e-05,
+      "loss": 1.0998,
+      "step": 10222
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7147232266427476e-05,
+      "loss": 1.0487,
+      "step": 10223
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7146635809076325e-05,
+      "loss": 1.115,
+      "step": 10224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714603929975405e-05,
+      "loss": 1.3636,
+      "step": 10225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7145442738464975e-05,
+      "loss": 0.9542,
+      "step": 10226
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7144846125213453e-05,
+      "loss": 0.7552,
+      "step": 10227
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7144249460003815e-05,
+      "loss": 0.9655,
+      "step": 10228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71436527428404e-05,
+      "loss": 1.0334,
+      "step": 10229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714305597372755e-05,
+      "loss": 1.179,
+      "step": 10230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7142459152669606e-05,
+      "loss": 0.9491,
+      "step": 10231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7141862279670905e-05,
+      "loss": 1.2834,
+      "step": 10232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7141265354735788e-05,
+      "loss": 1.1183,
+      "step": 10233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7140668377868593e-05,
+      "loss": 1.5443,
+      "step": 10234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7140071349073668e-05,
+      "loss": 0.7709,
+      "step": 10235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139474268355352e-05,
+      "loss": 0.9375,
+      "step": 10236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7138877135717985e-05,
+      "loss": 0.9533,
+      "step": 10237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713827995116591e-05,
+      "loss": 0.8986,
+      "step": 10238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713768271470347e-05,
+      "loss": 0.7774,
+      "step": 10239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713708542633501e-05,
+      "loss": 0.7204,
+      "step": 10240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7136488086064875e-05,
+      "loss": 1.2561,
+      "step": 10241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135890693897402e-05,
+      "loss": 1.037,
+      "step": 10242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713529324983694e-05,
+      "loss": 0.9355,
+      "step": 10243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7134695753887834e-05,
+      "loss": 0.8656,
+      "step": 10244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713409820605443e-05,
+      "loss": 1.324,
+      "step": 10245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713350060634107e-05,
+      "loss": 0.8605,
+      "step": 10246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71329029547521e-05,
+      "loss": 1.0401,
+      "step": 10247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713230525129187e-05,
+      "loss": 0.9327,
+      "step": 10248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713170749596472e-05,
+      "loss": 1.094,
+      "step": 10249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7131109688775006e-05,
+      "loss": 0.8312,
+      "step": 10250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7130511829727066e-05,
+      "loss": 1.1272,
+      "step": 10251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7129913918825257e-05,
+      "loss": 1.2676,
+      "step": 10252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712931595607392e-05,
+      "loss": 1.2002,
+      "step": 10253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7128717941477403e-05,
+      "loss": 1.0812,
+      "step": 10254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7128119875040063e-05,
+      "loss": 0.8455,
+      "step": 10255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712752175676624e-05,
+      "loss": 1.2003,
+      "step": 10256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7126923586660288e-05,
+      "loss": 0.7737,
+      "step": 10257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7126325364726553e-05,
+      "loss": 0.8592,
+      "step": 10258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71257270909694e-05,
+      "loss": 1.0263,
+      "step": 10259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7125128765393157e-05,
+      "loss": 1.2061,
+      "step": 10260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712453038800219e-05,
+      "loss": 1.1198,
+      "step": 10261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7123931958800844e-05,
+      "loss": 1.0934,
+      "step": 10262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7123333477793477e-05,
+      "loss": 0.6695,
+      "step": 10263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7122734944984438e-05,
+      "loss": 1.0072,
+      "step": 10264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7122136360378077e-05,
+      "loss": 0.7639,
+      "step": 10265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712153772397875e-05,
+      "loss": 1.1078,
+      "step": 10266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7120939035790812e-05,
+      "loss": 1.1425,
+      "step": 10267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7120340295818615e-05,
+      "loss": 1.1151,
+      "step": 10268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711974150406651e-05,
+      "loss": 0.8933,
+      "step": 10269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119142660538856e-05,
+      "loss": 0.7696,
+      "step": 10270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7118543765240003e-05,
+      "loss": 1.1251,
+      "step": 10271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7117944818174312e-05,
+      "loss": 0.9591,
+      "step": 10272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711734581934614e-05,
+      "loss": 1.1769,
+      "step": 10273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116746768759833e-05,
+      "loss": 1.0016,
+      "step": 10274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116147666419755e-05,
+      "loss": 0.6186,
+      "step": 10275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7115548512330257e-05,
+      "loss": 0.9119,
+      "step": 10276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7114949306495706e-05,
+      "loss": 0.8101,
+      "step": 10277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711435004892045e-05,
+      "loss": 1.0607,
+      "step": 10278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7113750739608853e-05,
+      "loss": 1.3126,
+      "step": 10279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711315137856527e-05,
+      "loss": 0.9619,
+      "step": 10280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7112551965794057e-05,
+      "loss": 1.1697,
+      "step": 10281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7111952501299577e-05,
+      "loss": 1.257,
+      "step": 10282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711135298508619e-05,
+      "loss": 1.0007,
+      "step": 10283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711075341715826e-05,
+      "loss": 0.9646,
+      "step": 10284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7110153797520133e-05,
+      "loss": 1.0871,
+      "step": 10285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109554126176183e-05,
+      "loss": 0.8872,
+      "step": 10286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710895440313076e-05,
+      "loss": 1.044,
+      "step": 10287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710835462838824e-05,
+      "loss": 1.1469,
+      "step": 10288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710775480195297e-05,
+      "loss": 1.2274,
+      "step": 10289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7107154923829317e-05,
+      "loss": 0.5225,
+      "step": 10290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7106554994021648e-05,
+      "loss": 1.2725,
+      "step": 10291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7105955012534318e-05,
+      "loss": 1.0953,
+      "step": 10292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7105354979371696e-05,
+      "loss": 1.0612,
+      "step": 10293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7104754894538143e-05,
+      "loss": 1.0245,
+      "step": 10294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7104154758038027e-05,
+      "loss": 0.7574,
+      "step": 10295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7103554569875704e-05,
+      "loss": 1.0084,
+      "step": 10296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7102954330055546e-05,
+      "loss": 0.8705,
+      "step": 10297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7102354038581915e-05,
+      "loss": 0.8314,
+      "step": 10298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101753695459177e-05,
+      "loss": 0.9637,
+      "step": 10299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101153300691695e-05,
+      "loss": 1.0191,
+      "step": 10300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710055285428384e-05,
+      "loss": 0.7799,
+      "step": 10301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709995235623997e-05,
+      "loss": 1.0417,
+      "step": 10302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7099351806564464e-05,
+      "loss": 1.2513,
+      "step": 10303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7098751205261686e-05,
+      "loss": 0.7742,
+      "step": 10304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7098150552335997e-05,
+      "loss": 1.0207,
+      "step": 10305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709754984779177e-05,
+      "loss": 1.0841,
+      "step": 10306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709694909163337e-05,
+      "loss": 0.8446,
+      "step": 10307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709634828386517e-05,
+      "loss": 1.1243,
+      "step": 10308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709574742449154e-05,
+      "loss": 1.2067,
+      "step": 10309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7095146513516842e-05,
+      "loss": 0.8582,
+      "step": 10310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7094545550945455e-05,
+      "loss": 1.1004,
+      "step": 10311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093944536781743e-05,
+      "loss": 1.0922,
+      "step": 10312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093343471030074e-05,
+      "loss": 1.0055,
+      "step": 10313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709274235369483e-05,
+      "loss": 1.1569,
+      "step": 10314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7092141184780374e-05,
+      "loss": 0.7579,
+      "step": 10315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709153996429108e-05,
+      "loss": 0.9759,
+      "step": 10316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709093869223132e-05,
+      "loss": 0.885,
+      "step": 10317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7090337368605465e-05,
+      "loss": 1.1259,
+      "step": 10318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7089735993417892e-05,
+      "loss": 0.973,
+      "step": 10319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708913456667297e-05,
+      "loss": 1.2001,
+      "step": 10320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7088533088375072e-05,
+      "loss": 1.2085,
+      "step": 10321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7087931558528573e-05,
+      "loss": 1.124,
+      "step": 10322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7087329977137853e-05,
+      "loss": 0.9417,
+      "step": 10323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708672834420728e-05,
+      "loss": 1.1634,
+      "step": 10324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7086126659741237e-05,
+      "loss": 0.927,
+      "step": 10325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7085524923744086e-05,
+      "loss": 1.1068,
+      "step": 10326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7084923136220214e-05,
+      "loss": 1.2938,
+      "step": 10327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7084321297173997e-05,
+      "loss": 0.7797,
+      "step": 10328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7083719406609807e-05,
+      "loss": 1.2319,
+      "step": 10329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708311746453202e-05,
+      "loss": 1.0529,
+      "step": 10330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082515470945022e-05,
+      "loss": 0.9925,
+      "step": 10331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708191342585318e-05,
+      "loss": 0.7392,
+      "step": 10332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708131132926088e-05,
+      "loss": 1.2216,
+      "step": 10333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7080709181172496e-05,
+      "loss": 0.9265,
+      "step": 10334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7080106981592407e-05,
+      "loss": 1.056,
+      "step": 10335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7079504730524998e-05,
+      "loss": 1.0764,
+      "step": 10336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707890242797464e-05,
+      "loss": 1.1237,
+      "step": 10337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707830007394572e-05,
+      "loss": 1.2347,
+      "step": 10338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7077697668442616e-05,
+      "loss": 1.094,
+      "step": 10339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707709521146971e-05,
+      "loss": 1.144,
+      "step": 10340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7076492703031376e-05,
+      "loss": 0.7518,
+      "step": 10341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7075890143132006e-05,
+      "loss": 1.0731,
+      "step": 10342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707528753177598e-05,
+      "loss": 1.2137,
+      "step": 10343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707468486896767e-05,
+      "loss": 1.4186,
+      "step": 10344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707408215471147e-05,
+      "loss": 1.0121,
+      "step": 10345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707347938901176e-05,
+      "loss": 0.7329,
+      "step": 10346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707287657187292e-05,
+      "loss": 0.9393,
+      "step": 10347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7072273703299336e-05,
+      "loss": 1.077,
+      "step": 10348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7071670783295394e-05,
+      "loss": 0.9749,
+      "step": 10349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7071067811865477e-05,
+      "loss": 0.9069,
+      "step": 10350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707046478901397e-05,
+      "loss": 0.9247,
+      "step": 10351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7069861714745256e-05,
+      "loss": 1.1489,
+      "step": 10352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7069258589063725e-05,
+      "loss": 1.2852,
+      "step": 10353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7068655411973757e-05,
+      "loss": 0.9023,
+      "step": 10354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7068052183479743e-05,
+      "loss": 1.1627,
+      "step": 10355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706744890358607e-05,
+      "loss": 1.0738,
+      "step": 10356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706684557229712e-05,
+      "loss": 1.0602,
+      "step": 10357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706624218961729e-05,
+      "loss": 1.0595,
+      "step": 10358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7065638755550958e-05,
+      "loss": 1.0129,
+      "step": 10359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7065035270102516e-05,
+      "loss": 1.0174,
+      "step": 10360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7064431733276355e-05,
+      "loss": 1.1628,
+      "step": 10361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7063828145076863e-05,
+      "loss": 0.958,
+      "step": 10362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7063224505508428e-05,
+      "loss": 0.5772,
+      "step": 10363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706262081457544e-05,
+      "loss": 0.8749,
+      "step": 10364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7062017072282285e-05,
+      "loss": 1.336,
+      "step": 10365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7061413278633363e-05,
+      "loss": 1.4296,
+      "step": 10366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7060809433633058e-05,
+      "loss": 0.8497,
+      "step": 10367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706020553728576e-05,
+      "loss": 0.7926,
+      "step": 10368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7059601589595868e-05,
+      "loss": 1.0595,
+      "step": 10369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7058997590567765e-05,
+      "loss": 1.113,
+      "step": 10370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705839354020585e-05,
+      "loss": 1.1312,
+      "step": 10371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705778943851451e-05,
+      "loss": 0.8927,
+      "step": 10372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705718528549815e-05,
+      "loss": 0.9381,
+      "step": 10373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7056581081161145e-05,
+      "loss": 0.7896,
+      "step": 10374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7055976825507904e-05,
+      "loss": 0.8123,
+      "step": 10375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7055372518542812e-05,
+      "loss": 1.0952,
+      "step": 10376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705476816027027e-05,
+      "loss": 1.3523,
+      "step": 10377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705416375069467e-05,
+      "loss": 1.0441,
+      "step": 10378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705355928982041e-05,
+      "loss": 0.8664,
+      "step": 10379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705295477765188e-05,
+      "loss": 0.9052,
+      "step": 10380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7052350214193486e-05,
+      "loss": 0.8438,
+      "step": 10381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7051745599449612e-05,
+      "loss": 1.0753,
+      "step": 10382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7051140933424663e-05,
+      "loss": 1.124,
+      "step": 10383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7050536216123035e-05,
+      "loss": 1.3461,
+      "step": 10384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704993144754912e-05,
+      "loss": 1.0422,
+      "step": 10385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7049326627707326e-05,
+      "loss": 0.8324,
+      "step": 10386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7048721756602045e-05,
+      "loss": 0.9467,
+      "step": 10387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704811683423768e-05,
+      "loss": 0.9693,
+      "step": 10388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704751186061862e-05,
+      "loss": 1.232,
+      "step": 10389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046906835749273e-05,
+      "loss": 0.7943,
+      "step": 10390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046301759634037e-05,
+      "loss": 1.11,
+      "step": 10391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7045696632277312e-05,
+      "loss": 0.9248,
+      "step": 10392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7045091453683504e-05,
+      "loss": 1.0223,
+      "step": 10393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7044486223857002e-05,
+      "loss": 1.2396,
+      "step": 10394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043880942802212e-05,
+      "loss": 0.868,
+      "step": 10395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043275610523545e-05,
+      "loss": 1.2906,
+      "step": 10396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7042670227025392e-05,
+      "loss": 0.7859,
+      "step": 10397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7042064792312158e-05,
+      "loss": 0.9991,
+      "step": 10398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7041459306388247e-05,
+      "loss": 0.8342,
+      "step": 10399
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7040853769258066e-05,
+      "loss": 1.0045,
+      "step": 10400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7040248180926012e-05,
+      "loss": 0.8537,
+      "step": 10401
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703964254139649e-05,
+      "loss": 1.2471,
+      "step": 10402
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7039036850673907e-05,
+      "loss": 0.9206,
+      "step": 10403
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7038431108762667e-05,
+      "loss": 0.7003,
+      "step": 10404
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7037825315667173e-05,
+      "loss": 0.9742,
+      "step": 10405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703721947139183e-05,
+      "loss": 1.4144,
+      "step": 10406
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703661357594105e-05,
+      "loss": 1.0736,
+      "step": 10407
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7036007629319232e-05,
+      "loss": 1.2497,
+      "step": 10408
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7035401631530786e-05,
+      "loss": 1.4413,
+      "step": 10409
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7034795582580118e-05,
+      "loss": 1.0086,
+      "step": 10410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7034189482471634e-05,
+      "loss": 0.7332,
+      "step": 10411
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7033583331209743e-05,
+      "loss": 1.1268,
+      "step": 10412
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7032977128798856e-05,
+      "loss": 0.8533,
+      "step": 10413
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7032370875243377e-05,
+      "loss": 1.3022,
+      "step": 10414
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7031764570547715e-05,
+      "loss": 1.1093,
+      "step": 10415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703115821471628e-05,
+      "loss": 1.5096,
+      "step": 10416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703055180775348e-05,
+      "loss": 0.7636,
+      "step": 10417
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702994534966373e-05,
+      "loss": 0.9669,
+      "step": 10418
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7029338840451437e-05,
+      "loss": 0.6892,
+      "step": 10419
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7028732280121008e-05,
+      "loss": 0.7568,
+      "step": 10420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702812566867686e-05,
+      "loss": 0.8737,
+      "step": 10421
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7027519006123402e-05,
+      "loss": 0.7712,
+      "step": 10422
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026912292465044e-05,
+      "loss": 1.0249,
+      "step": 10423
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70263055277062e-05,
+      "loss": 0.6069,
+      "step": 10424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7025698711851283e-05,
+      "loss": 0.8061,
+      "step": 10425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7025091844904702e-05,
+      "loss": 0.9833,
+      "step": 10426
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7024484926870874e-05,
+      "loss": 0.8397,
+      "step": 10427
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7023877957754213e-05,
+      "loss": 1.1694,
+      "step": 10428
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7023270937559127e-05,
+      "loss": 0.9891,
+      "step": 10429
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7022663866290042e-05,
+      "loss": 1.0065,
+      "step": 10430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702205674395136e-05,
+      "loss": 0.8559,
+      "step": 10431
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7021449570547502e-05,
+      "loss": 1.2839,
+      "step": 10432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7020842346082886e-05,
+      "loss": 1.2997,
+      "step": 10433
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7020235070561923e-05,
+      "loss": 0.6736,
+      "step": 10434
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7019627743989034e-05,
+      "loss": 1.3041,
+      "step": 10435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7019020366368627e-05,
+      "loss": 1.2158,
+      "step": 10436
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701841293770513e-05,
+      "loss": 1.0649,
+      "step": 10437
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701780545800295e-05,
+      "loss": 1.2429,
+      "step": 10438
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7017197927266512e-05,
+      "loss": 1.0839,
+      "step": 10439
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701659034550023e-05,
+      "loss": 1.0498,
+      "step": 10440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7015982712708526e-05,
+      "loss": 0.9185,
+      "step": 10441
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7015375028895815e-05,
+      "loss": 0.7689,
+      "step": 10442
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701476729406652e-05,
+      "loss": 1.0389,
+      "step": 10443
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7014159508225056e-05,
+      "loss": 0.9643,
+      "step": 10444
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7013551671375843e-05,
+      "loss": 0.8843,
+      "step": 10445
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7012943783523307e-05,
+      "loss": 0.8381,
+      "step": 10446
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7012335844671862e-05,
+      "loss": 0.6371,
+      "step": 10447
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7011727854825934e-05,
+      "loss": 1.1403,
+      "step": 10448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7011119813989944e-05,
+      "loss": 1.0351,
+      "step": 10449
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701051172216831e-05,
+      "loss": 1.1913,
+      "step": 10450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700990357936546e-05,
+      "loss": 1.513,
+      "step": 10451
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7009295385585806e-05,
+      "loss": 0.9592,
+      "step": 10452
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7008687140833783e-05,
+      "loss": 1.0747,
+      "step": 10453
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7008078845113808e-05,
+      "loss": 0.7659,
+      "step": 10454
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.70074704984303e-05,
+      "loss": 0.9637,
+      "step": 10455
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7006862100787692e-05,
+      "loss": 1.0662,
+      "step": 10456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7006253652190407e-05,
+      "loss": 0.9319,
+      "step": 10457
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700564515264287e-05,
+      "loss": 0.9957,
+      "step": 10458
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7005036602149497e-05,
+      "loss": 1.2227,
+      "step": 10459
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7004428000714722e-05,
+      "loss": 1.1341,
+      "step": 10460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700381934834297e-05,
+      "loss": 1.2927,
+      "step": 10461
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7003210645038663e-05,
+      "loss": 1.102,
+      "step": 10462
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7002601890806233e-05,
+      "loss": 1.0614,
+      "step": 10463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7001993085650103e-05,
+      "loss": 0.7701,
+      "step": 10464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7001384229574703e-05,
+      "loss": 0.904,
+      "step": 10465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700077532258446e-05,
+      "loss": 0.9173,
+      "step": 10466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.70001663646838e-05,
+      "loss": 1.0652,
+      "step": 10467
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6999557355877154e-05,
+      "loss": 1.0739,
+      "step": 10468
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6998948296168948e-05,
+      "loss": 0.8661,
+      "step": 10469
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6998339185563614e-05,
+      "loss": 0.8302,
+      "step": 10470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699773002406558e-05,
+      "loss": 1.0544,
+      "step": 10471
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6997120811679274e-05,
+      "loss": 0.8192,
+      "step": 10472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6996511548409133e-05,
+      "loss": 1.024,
+      "step": 10473
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699590223425958e-05,
+      "loss": 0.9868,
+      "step": 10474
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699529286923505e-05,
+      "loss": 1.2087,
+      "step": 10475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6994683453339974e-05,
+      "loss": 0.9313,
+      "step": 10476
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699407398657878e-05,
+      "loss": 0.9114,
+      "step": 10477
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6993464468955906e-05,
+      "loss": 1.299,
+      "step": 10478
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992854900475785e-05,
+      "loss": 0.8454,
+      "step": 10479
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992245281142842e-05,
+      "loss": 1.0465,
+      "step": 10480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6991635610961515e-05,
+      "loss": 0.7017,
+      "step": 10481
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6991025889936236e-05,
+      "loss": 0.96,
+      "step": 10482
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6990416118071446e-05,
+      "loss": 1.2414,
+      "step": 10483
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698980629537157e-05,
+      "loss": 0.8831,
+      "step": 10484
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989196421841045e-05,
+      "loss": 0.7241,
+      "step": 10485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698858649748431e-05,
+      "loss": 0.8819,
+      "step": 10486
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69879765223058e-05,
+      "loss": 1.1375,
+      "step": 10487
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6987366496309943e-05,
+      "loss": 0.9611,
+      "step": 10488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6986756419501186e-05,
+      "loss": 1.0497,
+      "step": 10489
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6986146291883958e-05,
+      "loss": 1.0175,
+      "step": 10490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6985536113462696e-05,
+      "loss": 0.8524,
+      "step": 10491
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6984925884241847e-05,
+      "loss": 1.0722,
+      "step": 10492
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6984315604225834e-05,
+      "loss": 0.987,
+      "step": 10493
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6983705273419103e-05,
+      "loss": 0.7888,
+      "step": 10494
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6983094891826096e-05,
+      "loss": 1.0116,
+      "step": 10495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982484459451245e-05,
+      "loss": 0.9066,
+      "step": 10496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698187397629899e-05,
+      "loss": 0.6276,
+      "step": 10497
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6981263442373777e-05,
+      "loss": 1.1467,
+      "step": 10498
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6980652857680036e-05,
+      "loss": 0.9759,
+      "step": 10499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6980042222222216e-05,
+      "loss": 0.8262,
+      "step": 10500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697943153600475e-05,
+      "loss": 1.0104,
+      "step": 10501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978820799032084e-05,
+      "loss": 1.3432,
+      "step": 10502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978210011308663e-05,
+      "loss": 1.0123,
+      "step": 10503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697759917283892e-05,
+      "loss": 0.812,
+      "step": 10504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69769882836273e-05,
+      "loss": 1.0874,
+      "step": 10505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6976377343678248e-05,
+      "loss": 1.3473,
+      "step": 10506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6975766352996205e-05,
+      "loss": 1.3236,
+      "step": 10507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6975155311585613e-05,
+      "loss": 1.159,
+      "step": 10508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697454421945092e-05,
+      "loss": 1.1653,
+      "step": 10509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6973933076596567e-05,
+      "loss": 1.0485,
+      "step": 10510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6973321883026992e-05,
+      "loss": 0.9426,
+      "step": 10511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697271063874665e-05,
+      "loss": 1.0298,
+      "step": 10512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6972099343759984e-05,
+      "loss": 0.7088,
+      "step": 10513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6971487998071434e-05,
+      "loss": 0.8036,
+      "step": 10514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697087660168545e-05,
+      "loss": 1.1104,
+      "step": 10515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6970265154606478e-05,
+      "loss": 1.3454,
+      "step": 10516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6969653656838966e-05,
+      "loss": 1.1699,
+      "step": 10517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6969042108387354e-05,
+      "loss": 1.0132,
+      "step": 10518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6968430509256097e-05,
+      "loss": 1.1392,
+      "step": 10519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696781885944964e-05,
+      "loss": 0.8197,
+      "step": 10520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6967207158972426e-05,
+      "loss": 1.5476,
+      "step": 10521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6966595407828913e-05,
+      "loss": 0.7083,
+      "step": 10522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965983606023542e-05,
+      "loss": 1.1858,
+      "step": 10523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6965371753560767e-05,
+      "loss": 1.4537,
+      "step": 10524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696475985044503e-05,
+      "loss": 1.0112,
+      "step": 10525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6964147896680792e-05,
+      "loss": 1.0469,
+      "step": 10526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6963535892272493e-05,
+      "loss": 0.5676,
+      "step": 10527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696292383722459e-05,
+      "loss": 1.1873,
+      "step": 10528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6962311731541528e-05,
+      "loss": 0.9672,
+      "step": 10529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6961699575227767e-05,
+      "loss": 0.9472,
+      "step": 10530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696108736828775e-05,
+      "loss": 0.7776,
+      "step": 10531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6960475110725932e-05,
+      "loss": 0.9409,
+      "step": 10532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6959862802546773e-05,
+      "loss": 0.9316,
+      "step": 10533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6959250443754712e-05,
+      "loss": 1.3222,
+      "step": 10534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695863803435421e-05,
+      "loss": 0.7521,
+      "step": 10535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6958025574349722e-05,
+      "loss": 0.8449,
+      "step": 10536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6957413063745697e-05,
+      "loss": 0.9037,
+      "step": 10537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6956800502546592e-05,
+      "loss": 1.0497,
+      "step": 10538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6956187890756862e-05,
+      "loss": 1.4065,
+      "step": 10539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695557522838096e-05,
+      "loss": 1.1926,
+      "step": 10540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6954962515423342e-05,
+      "loss": 1.3216,
+      "step": 10541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6954349751888466e-05,
+      "loss": 1.3061,
+      "step": 10542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6953736937780786e-05,
+      "loss": 0.8644,
+      "step": 10543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6953124073104762e-05,
+      "loss": 0.8852,
+      "step": 10544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695251115786484e-05,
+      "loss": 0.8099,
+      "step": 10545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695189819206549e-05,
+      "loss": 1.2762,
+      "step": 10546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6951285175711165e-05,
+      "loss": 0.9531,
+      "step": 10547
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6950672108806322e-05,
+      "loss": 0.9524,
+      "step": 10548
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6950058991355416e-05,
+      "loss": 0.654,
+      "step": 10549
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694944582336291e-05,
+      "loss": 0.937,
+      "step": 10550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948832604833265e-05,
+      "loss": 0.7655,
+      "step": 10551
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948219335770938e-05,
+      "loss": 1.0608,
+      "step": 10552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6947606016180386e-05,
+      "loss": 0.9154,
+      "step": 10553
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6946992646066073e-05,
+      "loss": 1.0668,
+      "step": 10554
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6946379225432456e-05,
+      "loss": 0.7013,
+      "step": 10555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6945765754284e-05,
+      "loss": 0.8458,
+      "step": 10556
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694515223262516e-05,
+      "loss": 1.0109,
+      "step": 10557
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694453866046041e-05,
+      "loss": 1.1141,
+      "step": 10558
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6943925037794197e-05,
+      "loss": 1.2016,
+      "step": 10559
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694331136463099e-05,
+      "loss": 1.1294,
+      "step": 10560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6942697640975254e-05,
+      "loss": 0.8532,
+      "step": 10561
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694208386683145e-05,
+      "loss": 1.3908,
+      "step": 10562
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694147004220404e-05,
+      "loss": 1.0153,
+      "step": 10563
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694085616709749e-05,
+      "loss": 0.8006,
+      "step": 10564
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6940242241516262e-05,
+      "loss": 1.4026,
+      "step": 10565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693962826546482e-05,
+      "loss": 0.7727,
+      "step": 10566
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6939014238947635e-05,
+      "loss": 1.0943,
+      "step": 10567
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6938400161969166e-05,
+      "loss": 1.1387,
+      "step": 10568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693778603453388e-05,
+      "loss": 1.3651,
+      "step": 10569
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6937171856646248e-05,
+      "loss": 1.0033,
+      "step": 10570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6936557628310723e-05,
+      "loss": 0.8224,
+      "step": 10571
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6935943349531787e-05,
+      "loss": 0.8258,
+      "step": 10572
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69353290203139e-05,
+      "loss": 1.3259,
+      "step": 10573
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693471464066153e-05,
+      "loss": 0.7038,
+      "step": 10574
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6934100210579144e-05,
+      "loss": 1.1384,
+      "step": 10575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6933485730071215e-05,
+      "loss": 1.0041,
+      "step": 10576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69328711991422e-05,
+      "loss": 1.1635,
+      "step": 10577
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6932256617796582e-05,
+      "loss": 1.1835,
+      "step": 10578
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6931641986038824e-05,
+      "loss": 0.9857,
+      "step": 10579
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6931027303873393e-05,
+      "loss": 1.0769,
+      "step": 10580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6930412571304762e-05,
+      "loss": 1.319,
+      "step": 10581
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69297977883374e-05,
+      "loss": 0.9578,
+      "step": 10582
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6929182954975785e-05,
+      "loss": 0.7956,
+      "step": 10583
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692856807122438e-05,
+      "loss": 1.1395,
+      "step": 10584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6927953137087654e-05,
+      "loss": 0.7321,
+      "step": 10585
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6927338152570085e-05,
+      "loss": 0.9323,
+      "step": 10586
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6926723117676144e-05,
+      "loss": 1.1317,
+      "step": 10587
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6926108032410305e-05,
+      "loss": 0.885,
+      "step": 10588
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6925492896777038e-05,
+      "loss": 0.9937,
+      "step": 10589
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6924877710780818e-05,
+      "loss": 0.8353,
+      "step": 10590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692426247442612e-05,
+      "loss": 1.2348,
+      "step": 10591
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6923647187717413e-05,
+      "loss": 1.2226,
+      "step": 10592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692303185065918e-05,
+      "loss": 1.0895,
+      "step": 10593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6922416463255888e-05,
+      "loss": 0.8415,
+      "step": 10594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692180102551201e-05,
+      "loss": 1.3186,
+      "step": 10595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6921185537432033e-05,
+      "loss": 0.9018,
+      "step": 10596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6920569999020423e-05,
+      "loss": 1.099,
+      "step": 10597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691995441028166e-05,
+      "loss": 0.9215,
+      "step": 10598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691933877122022e-05,
+      "loss": 0.7961,
+      "step": 10599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691872308184058e-05,
+      "loss": 1.098,
+      "step": 10600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6918107342147215e-05,
+      "loss": 1.287,
+      "step": 10601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6917491552144607e-05,
+      "loss": 0.6489,
+      "step": 10602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6916875711837235e-05,
+      "loss": 1.0411,
+      "step": 10603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6916259821229573e-05,
+      "loss": 0.9458,
+      "step": 10604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69156438803261e-05,
+      "loss": 1.0269,
+      "step": 10605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6915027889131298e-05,
+      "loss": 0.8541,
+      "step": 10606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6914411847649647e-05,
+      "loss": 0.6873,
+      "step": 10607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6913795755885623e-05,
+      "loss": 1.3015,
+      "step": 10608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6913179613843708e-05,
+      "loss": 1.2427,
+      "step": 10609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6912563421528382e-05,
+      "loss": 1.1415,
+      "step": 10610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691194717894413e-05,
+      "loss": 0.8193,
+      "step": 10611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691133088609543e-05,
+      "loss": 0.9744,
+      "step": 10612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6910714542986766e-05,
+      "loss": 0.8528,
+      "step": 10613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6910098149622614e-05,
+      "loss": 0.8658,
+      "step": 10614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6909481706007463e-05,
+      "loss": 1.2913,
+      "step": 10615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6908865212145796e-05,
+      "loss": 1.1816,
+      "step": 10616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6908248668042092e-05,
+      "loss": 0.9158,
+      "step": 10617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690763207370084e-05,
+      "loss": 0.6986,
+      "step": 10618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6907015429126515e-05,
+      "loss": 0.9768,
+      "step": 10619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690639873432361e-05,
+      "loss": 1.0621,
+      "step": 10620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6905781989296604e-05,
+      "loss": 0.9245,
+      "step": 10621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6905165194049983e-05,
+      "loss": 0.9986,
+      "step": 10622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690454834858824e-05,
+      "loss": 1.0832,
+      "step": 10623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6903931452915853e-05,
+      "loss": 1.0799,
+      "step": 10624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6903314507037307e-05,
+      "loss": 1.065,
+      "step": 10625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6902697510957093e-05,
+      "loss": 0.7475,
+      "step": 10626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6902080464679694e-05,
+      "loss": 0.7353,
+      "step": 10627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6901463368209603e-05,
+      "loss": 1.1061,
+      "step": 10628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6900846221551305e-05,
+      "loss": 1.1622,
+      "step": 10629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690022902470928e-05,
+      "loss": 1.0069,
+      "step": 10630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899611777688027e-05,
+      "loss": 0.9162,
+      "step": 10631
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6898994480492032e-05,
+      "loss": 1.1215,
+      "step": 10632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6898377133125784e-05,
+      "loss": 1.0551,
+      "step": 10633
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689775973559377e-05,
+      "loss": 1.0745,
+      "step": 10634
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6897142287900477e-05,
+      "loss": 1.0967,
+      "step": 10635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6896524790050407e-05,
+      "loss": 0.9129,
+      "step": 10636
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6895907242048038e-05,
+      "loss": 1.0597,
+      "step": 10637
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689528964389787e-05,
+      "loss": 0.9536,
+      "step": 10638
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6894671995604386e-05,
+      "loss": 1.1733,
+      "step": 10639
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6894054297172083e-05,
+      "loss": 0.9262,
+      "step": 10640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689343654860545e-05,
+      "loss": 0.7489,
+      "step": 10641
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6892818749908983e-05,
+      "loss": 1.0831,
+      "step": 10642
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6892200901087172e-05,
+      "loss": 0.7463,
+      "step": 10643
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689158300214451e-05,
+      "loss": 1.0533,
+      "step": 10644
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890965053085492e-05,
+      "loss": 1.1536,
+      "step": 10645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689034705391461e-05,
+      "loss": 0.7759,
+      "step": 10646
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6889729004636358e-05,
+      "loss": 0.9563,
+      "step": 10647
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6889110905255235e-05,
+      "loss": 0.9493,
+      "step": 10648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6888492755775733e-05,
+      "loss": 1.0137,
+      "step": 10649
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6887874556202342e-05,
+      "loss": 0.6404,
+      "step": 10650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6887256306539567e-05,
+      "loss": 0.9431,
+      "step": 10651
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6886638006791896e-05,
+      "loss": 0.8846,
+      "step": 10652
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688601965696383e-05,
+      "loss": 0.8062,
+      "step": 10653
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6885401257059865e-05,
+      "loss": 0.615,
+      "step": 10654
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6884782807084494e-05,
+      "loss": 0.8818,
+      "step": 10655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688416430704222e-05,
+      "loss": 1.0751,
+      "step": 10656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688354575693754e-05,
+      "loss": 0.8518,
+      "step": 10657
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6882927156774954e-05,
+      "loss": 1.3676,
+      "step": 10658
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688230850655895e-05,
+      "loss": 1.3154,
+      "step": 10659
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688168980629404e-05,
+      "loss": 0.8249,
+      "step": 10660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6881071055984715e-05,
+      "loss": 0.7995,
+      "step": 10661
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6880452255635476e-05,
+      "loss": 0.9967,
+      "step": 10662
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6879833405250826e-05,
+      "loss": 0.9391,
+      "step": 10663
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6879214504835263e-05,
+      "loss": 0.9622,
+      "step": 10664
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687859555439329e-05,
+      "loss": 1.0076,
+      "step": 10665
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6877976553929404e-05,
+      "loss": 0.5182,
+      "step": 10666
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6877357503448108e-05,
+      "loss": 1.0858,
+      "step": 10667
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6876738402953906e-05,
+      "loss": 1.0481,
+      "step": 10668
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68761192524513e-05,
+      "loss": 1.0165,
+      "step": 10669
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6875500051944788e-05,
+      "loss": 1.143,
+      "step": 10670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6874880801438877e-05,
+      "loss": 0.9196,
+      "step": 10671
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6874261500938072e-05,
+      "loss": 0.9645,
+      "step": 10672
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687364215044687e-05,
+      "loss": 0.8431,
+      "step": 10673
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6873022749969784e-05,
+      "loss": 1.3453,
+      "step": 10674
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687240329951131e-05,
+      "loss": 0.7303,
+      "step": 10675
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6871783799075954e-05,
+      "loss": 1.0734,
+      "step": 10676
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687116424866823e-05,
+      "loss": 1.0686,
+      "step": 10677
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6870544648292632e-05,
+      "loss": 1.0665,
+      "step": 10678
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6869924997953668e-05,
+      "loss": 1.0104,
+      "step": 10679
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686930529765585e-05,
+      "loss": 1.3115,
+      "step": 10680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686868554740368e-05,
+      "loss": 1.0632,
+      "step": 10681
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6868065747201664e-05,
+      "loss": 1.0271,
+      "step": 10682
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6867445897054316e-05,
+      "loss": 0.9936,
+      "step": 10683
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6866825996966135e-05,
+      "loss": 1.1593,
+      "step": 10684
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6866206046941632e-05,
+      "loss": 1.0392,
+      "step": 10685
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686558604698532e-05,
+      "loss": 0.8218,
+      "step": 10686
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68649659971017e-05,
+      "loss": 0.7067,
+      "step": 10687
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686434589729529e-05,
+      "loss": 0.8621,
+      "step": 10688
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686372574757059e-05,
+      "loss": 0.8397,
+      "step": 10689
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6863105547932117e-05,
+      "loss": 1.3642,
+      "step": 10690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6862485298384377e-05,
+      "loss": 0.8795,
+      "step": 10691
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686186499893188e-05,
+      "loss": 1.1434,
+      "step": 10692
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686124464957914e-05,
+      "loss": 1.1524,
+      "step": 10693
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686062425033067e-05,
+      "loss": 1.0968,
+      "step": 10694
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6860003801190975e-05,
+      "loss": 0.7575,
+      "step": 10695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685938330216457e-05,
+      "loss": 1.1163,
+      "step": 10696
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685876275325597e-05,
+      "loss": 1.0208,
+      "step": 10697
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6858142154469684e-05,
+      "loss": 1.0962,
+      "step": 10698
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6857521505810227e-05,
+      "loss": 1.3109,
+      "step": 10699
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6856900807282116e-05,
+      "loss": 1.3352,
+      "step": 10700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685628005888986e-05,
+      "loss": 0.9295,
+      "step": 10701
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685565926063797e-05,
+      "loss": 0.8542,
+      "step": 10702
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6855038412530966e-05,
+      "loss": 0.8732,
+      "step": 10703
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685441751457336e-05,
+      "loss": 0.6746,
+      "step": 10704
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6853796566769667e-05,
+      "loss": 0.9931,
+      "step": 10705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685317556912441e-05,
+      "loss": 0.9519,
+      "step": 10706
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68525545216421e-05,
+      "loss": 0.7939,
+      "step": 10707
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685193342432725e-05,
+      "loss": 0.7308,
+      "step": 10708
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6851312277184378e-05,
+      "loss": 0.9119,
+      "step": 10709
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6850691080218e-05,
+      "loss": 0.9739,
+      "step": 10710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685006983343264e-05,
+      "loss": 1.1965,
+      "step": 10711
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684944853683281e-05,
+      "loss": 0.8639,
+      "step": 10712
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848827190423033e-05,
+      "loss": 1.3434,
+      "step": 10713
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848205794207818e-05,
+      "loss": 1.2423,
+      "step": 10714
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684758434819169e-05,
+      "loss": 0.8984,
+      "step": 10715
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6846962852379174e-05,
+      "loss": 1.0626,
+      "step": 10716
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684634130677478e-05,
+      "loss": 0.9181,
+      "step": 10717
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845719711383032e-05,
+      "loss": 0.978,
+      "step": 10718
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845098066208453e-05,
+      "loss": 1.1362,
+      "step": 10719
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684447637125556e-05,
+      "loss": 1.1427,
+      "step": 10720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684385462652887e-05,
+      "loss": 0.8908,
+      "step": 10721
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684323283203291e-05,
+      "loss": 0.7961,
+      "step": 10722
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684261098777221e-05,
+      "loss": 0.921,
+      "step": 10723
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6841989093751272e-05,
+      "loss": 1.0009,
+      "step": 10724
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6841367149974638e-05,
+      "loss": 0.9128,
+      "step": 10725
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6840745156446815e-05,
+      "loss": 0.7672,
+      "step": 10726
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6840123113172338e-05,
+      "loss": 1.0765,
+      "step": 10727
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6839501020155725e-05,
+      "loss": 1.3049,
+      "step": 10728
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68388788774015e-05,
+      "loss": 1.0382,
+      "step": 10729
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683825668491419e-05,
+      "loss": 0.6517,
+      "step": 10730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6837634442698317e-05,
+      "loss": 0.8671,
+      "step": 10731
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6837012150758408e-05,
+      "loss": 1.1047,
+      "step": 10732
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6836389809098987e-05,
+      "loss": 1.0588,
+      "step": 10733
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6835767417724583e-05,
+      "loss": 0.9001,
+      "step": 10734
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6835144976639713e-05,
+      "loss": 0.9021,
+      "step": 10735
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6834522485848913e-05,
+      "loss": 1.1638,
+      "step": 10736
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6833899945356707e-05,
+      "loss": 0.8849,
+      "step": 10737
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683327735516762e-05,
+      "loss": 0.8147,
+      "step": 10738
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6832654715286184e-05,
+      "loss": 0.8314,
+      "step": 10739
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683203202571692e-05,
+      "loss": 1.0834,
+      "step": 10740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6831409286464365e-05,
+      "loss": 1.0419,
+      "step": 10741
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683078649753304e-05,
+      "loss": 0.9353,
+      "step": 10742
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683016365892748e-05,
+      "loss": 0.9347,
+      "step": 10743
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682954077065221e-05,
+      "loss": 1.2064,
+      "step": 10744
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6828917832711762e-05,
+      "loss": 0.8495,
+      "step": 10745
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6828294845110663e-05,
+      "loss": 1.1497,
+      "step": 10746
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6827671807853444e-05,
+      "loss": 1.1015,
+      "step": 10747
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682704872094464e-05,
+      "loss": 0.8124,
+      "step": 10748
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6826425584388783e-05,
+      "loss": 0.968,
+      "step": 10749
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6825802398190395e-05,
+      "loss": 1.4174,
+      "step": 10750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6825179162354016e-05,
+      "loss": 0.9557,
+      "step": 10751
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6824555876884176e-05,
+      "loss": 0.9082,
+      "step": 10752
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682393254178541e-05,
+      "loss": 0.9344,
+      "step": 10753
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682330915706225e-05,
+      "loss": 0.9829,
+      "step": 10754
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822685722719224e-05,
+      "loss": 0.9289,
+      "step": 10755
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682206223876087e-05,
+      "loss": 0.856,
+      "step": 10756
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6821438705191722e-05,
+      "loss": 1.0121,
+      "step": 10757
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6820815122016317e-05,
+      "loss": 1.4556,
+      "step": 10758
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6820191489239187e-05,
+      "loss": 1.2732,
+      "step": 10759
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819567806864863e-05,
+      "loss": 0.9595,
+      "step": 10760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681894407489789e-05,
+      "loss": 1.1763,
+      "step": 10761
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818320293342793e-05,
+      "loss": 0.823,
+      "step": 10762
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6817696462204116e-05,
+      "loss": 1.2726,
+      "step": 10763
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6817072581486392e-05,
+      "loss": 1.0667,
+      "step": 10764
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681644865119416e-05,
+      "loss": 0.8457,
+      "step": 10765
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681582467133196e-05,
+      "loss": 0.7024,
+      "step": 10766
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681520064190432e-05,
+      "loss": 0.6218,
+      "step": 10767
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681457656291579e-05,
+      "loss": 0.8802,
+      "step": 10768
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68139524343709e-05,
+      "loss": 1.1888,
+      "step": 10769
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681332825627419e-05,
+      "loss": 0.7568,
+      "step": 10770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68127040286302e-05,
+      "loss": 1.0313,
+      "step": 10771
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6812079751443474e-05,
+      "loss": 1.1665,
+      "step": 10772
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6811455424718546e-05,
+      "loss": 0.8715,
+      "step": 10773
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6810831048459958e-05,
+      "loss": 0.8469,
+      "step": 10774
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6810206622672246e-05,
+      "loss": 0.846,
+      "step": 10775
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6809582147359958e-05,
+      "loss": 0.9831,
+      "step": 10776
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680895762252763e-05,
+      "loss": 1.0169,
+      "step": 10777
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680833304817981e-05,
+      "loss": 1.0255,
+      "step": 10778
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6807708424321033e-05,
+      "loss": 0.9434,
+      "step": 10779
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6807083750955847e-05,
+      "loss": 1.0262,
+      "step": 10780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6806459028088792e-05,
+      "loss": 0.7186,
+      "step": 10781
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805834255724408e-05,
+      "loss": 0.9786,
+      "step": 10782
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805209433867244e-05,
+      "loss": 0.8849,
+      "step": 10783
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6804584562521842e-05,
+      "loss": 1.1819,
+      "step": 10784
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680395964169274e-05,
+      "loss": 1.232,
+      "step": 10785
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6803334671384496e-05,
+      "loss": 0.8729,
+      "step": 10786
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6802709651601642e-05,
+      "loss": 1.2211,
+      "step": 10787
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.680208458234873e-05,
+      "loss": 1.2285,
+      "step": 10788
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6801459463630303e-05,
+      "loss": 0.9232,
+      "step": 10789
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6800834295450906e-05,
+      "loss": 0.9945,
+      "step": 10790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6800209077815087e-05,
+      "loss": 0.9804,
+      "step": 10791
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6799583810727395e-05,
+      "loss": 1.0693,
+      "step": 10792
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6798958494192372e-05,
+      "loss": 0.7285,
+      "step": 10793
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679833312821457e-05,
+      "loss": 0.9686,
+      "step": 10794
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679770771279853e-05,
+      "loss": 0.8941,
+      "step": 10795
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6797082247948807e-05,
+      "loss": 0.8777,
+      "step": 10796
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679645673366995e-05,
+      "loss": 1.0023,
+      "step": 10797
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6795831169966503e-05,
+      "loss": 1.012,
+      "step": 10798
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6795205556843015e-05,
+      "loss": 1.0645,
+      "step": 10799
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794579894304043e-05,
+      "loss": 0.7122,
+      "step": 10800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6793954182354125e-05,
+      "loss": 1.291,
+      "step": 10801
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6793328420997822e-05,
+      "loss": 1.0115,
+      "step": 10802
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6792702610239677e-05,
+      "loss": 1.218,
+      "step": 10803
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679207675008425e-05,
+      "loss": 0.8835,
+      "step": 10804
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679145084053608e-05,
+      "loss": 0.8155,
+      "step": 10805
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679082488159973e-05,
+      "loss": 1.2012,
+      "step": 10806
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6790198873279742e-05,
+      "loss": 1.074,
+      "step": 10807
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678957281558068e-05,
+      "loss": 0.8639,
+      "step": 10808
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6788946708507086e-05,
+      "loss": 0.7784,
+      "step": 10809
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678832055206352e-05,
+      "loss": 0.7853,
+      "step": 10810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678769434625453e-05,
+      "loss": 0.904,
+      "step": 10811
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6787068091084678e-05,
+      "loss": 0.7225,
+      "step": 10812
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678644178655851e-05,
+      "loss": 0.8368,
+      "step": 10813
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6785815432680586e-05,
+      "loss": 1.2324,
+      "step": 10814
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6785189029455455e-05,
+      "loss": 1.0574,
+      "step": 10815
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678456257688768e-05,
+      "loss": 0.9258,
+      "step": 10816
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6783936074981812e-05,
+      "loss": 0.8655,
+      "step": 10817
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6783309523742403e-05,
+      "loss": 0.7411,
+      "step": 10818
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6782682923174015e-05,
+      "loss": 0.9385,
+      "step": 10819
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6782056273281207e-05,
+      "loss": 0.9029,
+      "step": 10820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678142957406853e-05,
+      "loss": 0.8995,
+      "step": 10821
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6780802825540546e-05,
+      "loss": 1.017,
+      "step": 10822
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678017602770181e-05,
+      "loss": 1.0579,
+      "step": 10823
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6779549180556882e-05,
+      "loss": 0.9309,
+      "step": 10824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677892228411032e-05,
+      "loss": 0.9464,
+      "step": 10825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677829533836668e-05,
+      "loss": 0.98,
+      "step": 10826
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6777668343330524e-05,
+      "loss": 0.9899,
+      "step": 10827
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6777041299006413e-05,
+      "loss": 0.9259,
+      "step": 10828
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6776414205398902e-05,
+      "loss": 0.9695,
+      "step": 10829
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6775787062512557e-05,
+      "loss": 0.9473,
+      "step": 10830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6775159870351933e-05,
+      "loss": 1.2697,
+      "step": 10831
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67745326289216e-05,
+      "loss": 0.8254,
+      "step": 10832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6773905338226106e-05,
+      "loss": 1.142,
+      "step": 10833
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6773277998270024e-05,
+      "loss": 0.7776,
+      "step": 10834
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6772650609057915e-05,
+      "loss": 1.162,
+      "step": 10835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6772023170594334e-05,
+      "loss": 0.9971,
+      "step": 10836
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677139568288385e-05,
+      "loss": 0.5523,
+      "step": 10837
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770768145931025e-05,
+      "loss": 1.256,
+      "step": 10838
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770140559740422e-05,
+      "loss": 1.1128,
+      "step": 10839
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6769512924316605e-05,
+      "loss": 1.1523,
+      "step": 10840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6768885239664137e-05,
+      "loss": 1.1641,
+      "step": 10841
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6768257505787587e-05,
+      "loss": 0.929,
+      "step": 10842
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6767629722691517e-05,
+      "loss": 1.3215,
+      "step": 10843
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676700189038049e-05,
+      "loss": 0.8923,
+      "step": 10844
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676637400885907e-05,
+      "loss": 1.0153,
+      "step": 10845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6765746078131834e-05,
+      "loss": 1.0435,
+      "step": 10846
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6765118098203335e-05,
+      "loss": 0.835,
+      "step": 10847
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6764490069078154e-05,
+      "loss": 0.7368,
+      "step": 10848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6763861990760842e-05,
+      "loss": 1.0372,
+      "step": 10849
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676323386325598e-05,
+      "loss": 1.1061,
+      "step": 10850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6762605686568127e-05,
+      "loss": 0.9892,
+      "step": 10851
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6761977460701854e-05,
+      "loss": 1.0952,
+      "step": 10852
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6761349185661733e-05,
+      "loss": 1.0278,
+      "step": 10853
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760720861452326e-05,
+      "loss": 1.0624,
+      "step": 10854
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760092488078206e-05,
+      "loss": 0.7559,
+      "step": 10855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6759464065543947e-05,
+      "loss": 0.847,
+      "step": 10856
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6758835593854114e-05,
+      "loss": 1.1379,
+      "step": 10857
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675820707301328e-05,
+      "loss": 1.1112,
+      "step": 10858
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6757578503026006e-05,
+      "loss": 1.3149,
+      "step": 10859
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6756949883896874e-05,
+      "loss": 1.453,
+      "step": 10860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6756321215630453e-05,
+      "loss": 0.778,
+      "step": 10861
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6755692498231316e-05,
+      "loss": 1.0344,
+      "step": 10862
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675506373170403e-05,
+      "loss": 1.0548,
+      "step": 10863
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675443491605317e-05,
+      "loss": 1.0624,
+      "step": 10864
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6753806051283313e-05,
+      "loss": 0.6073,
+      "step": 10865
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6753177137399025e-05,
+      "loss": 0.8864,
+      "step": 10866
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6752548174404884e-05,
+      "loss": 0.8432,
+      "step": 10867
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6751919162305465e-05,
+      "loss": 0.8275,
+      "step": 10868
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6751290101105335e-05,
+      "loss": 1.0543,
+      "step": 10869
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750660990809078e-05,
+      "loss": 0.9672,
+      "step": 10870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675003183142126e-05,
+      "loss": 0.893,
+      "step": 10871
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674940262294647e-05,
+      "loss": 0.8394,
+      "step": 10872
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6748773365389265e-05,
+      "loss": 1.2048,
+      "step": 10873
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6748144058754237e-05,
+      "loss": 0.948,
+      "step": 10874
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6747514703045952e-05,
+      "loss": 1.0428,
+      "step": 10875
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674688529826899e-05,
+      "loss": 0.8242,
+      "step": 10876
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6746255844427934e-05,
+      "loss": 1.1054,
+      "step": 10877
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6745626341527353e-05,
+      "loss": 0.5114,
+      "step": 10878
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674499678957183e-05,
+      "loss": 0.8967,
+      "step": 10879
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674436718856594e-05,
+      "loss": 0.6769,
+      "step": 10880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674373753851426e-05,
+      "loss": 0.808,
+      "step": 10881
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674310783942138e-05,
+      "loss": 0.9765,
+      "step": 10882
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6742478091291863e-05,
+      "loss": 1.1338,
+      "step": 10883
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6741848294130298e-05,
+      "loss": 1.0367,
+      "step": 10884
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6741218447941266e-05,
+      "loss": 0.9625,
+      "step": 10885
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674058855272934e-05,
+      "loss": 1.0124,
+      "step": 10886
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6739958608499112e-05,
+      "loss": 0.9375,
+      "step": 10887
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6739328615255152e-05,
+      "loss": 0.9437,
+      "step": 10888
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673869857300205e-05,
+      "loss": 0.8964,
+      "step": 10889
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673806848174438e-05,
+      "loss": 0.9196,
+      "step": 10890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673743834148673e-05,
+      "loss": 1.1604,
+      "step": 10891
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736808152233678e-05,
+      "loss": 1.2067,
+      "step": 10892
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736177913989812e-05,
+      "loss": 0.9609,
+      "step": 10893
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673554762675971e-05,
+      "loss": 0.8725,
+      "step": 10894
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6734917290547955e-05,
+      "loss": 1.0221,
+      "step": 10895
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673428690535914e-05,
+      "loss": 1.1776,
+      "step": 10896
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673365647119784e-05,
+      "loss": 0.9493,
+      "step": 10897
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673302598806864e-05,
+      "loss": 1.1967,
+      "step": 10898
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673239545597613e-05,
+      "loss": 1.1041,
+      "step": 10899
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6731764874924888e-05,
+      "loss": 0.887,
+      "step": 10900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673113424491951e-05,
+      "loss": 0.9074,
+      "step": 10901
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6730503565964576e-05,
+      "loss": 1.2069,
+      "step": 10902
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672987283806467e-05,
+      "loss": 0.9929,
+      "step": 10903
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6729242061224382e-05,
+      "loss": 0.9988,
+      "step": 10904
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67286112354483e-05,
+      "loss": 0.6919,
+      "step": 10905
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6727980360741007e-05,
+      "loss": 1.0235,
+      "step": 10906
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67273494371071e-05,
+      "loss": 0.8399,
+      "step": 10907
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726718464551154e-05,
+      "loss": 1.1702,
+      "step": 10908
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726087443077768e-05,
+      "loss": 1.0966,
+      "step": 10909
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6725456372691523e-05,
+      "loss": 1.3273,
+      "step": 10910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6724825253397016e-05,
+      "loss": 1.3702,
+      "step": 10911
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672419408519883e-05,
+      "loss": 0.828,
+      "step": 10912
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6723562868101563e-05,
+      "loss": 1.1291,
+      "step": 10913
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722931602109796e-05,
+      "loss": 0.769,
+      "step": 10914
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6722300287228124e-05,
+      "loss": 1.0522,
+      "step": 10915
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672166892346114e-05,
+      "loss": 0.8985,
+      "step": 10916
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6721037510813433e-05,
+      "loss": 1.0625,
+      "step": 10917
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6720406049289596e-05,
+      "loss": 0.843,
+      "step": 10918
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6719774538894217e-05,
+      "loss": 0.851,
+      "step": 10919
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671914297963189e-05,
+      "loss": 0.9538,
+      "step": 10920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6718511371507213e-05,
+      "loss": 0.8146,
+      "step": 10921
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6717879714524774e-05,
+      "loss": 1.5103,
+      "step": 10922
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671724800868917e-05,
+      "loss": 0.9428,
+      "step": 10923
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6716616254004983e-05,
+      "loss": 1.0425,
+      "step": 10924
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6715984450476828e-05,
+      "loss": 1.1488,
+      "step": 10925
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671535259810928e-05,
+      "loss": 1.014,
+      "step": 10926
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671472069690695e-05,
+      "loss": 0.9262,
+      "step": 10927
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6714088746874422e-05,
+      "loss": 0.8221,
+      "step": 10928
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6713456748016295e-05,
+      "loss": 0.8675,
+      "step": 10929
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6712824700337164e-05,
+      "loss": 1.0324,
+      "step": 10930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6712192603841627e-05,
+      "loss": 1.0117,
+      "step": 10931
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671156045853428e-05,
+      "loss": 1.2389,
+      "step": 10932
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6710928264419718e-05,
+      "loss": 1.3597,
+      "step": 10933
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6710296021502542e-05,
+      "loss": 0.8836,
+      "step": 10934
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670966372978735e-05,
+      "loss": 0.6408,
+      "step": 10935
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6709031389278735e-05,
+      "loss": 1.0576,
+      "step": 10936
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6708398999981302e-05,
+      "loss": 0.9214,
+      "step": 10937
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6707766561899644e-05,
+      "loss": 1.1926,
+      "step": 10938
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6707134075038365e-05,
+      "loss": 0.7642,
+      "step": 10939
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6706501539402063e-05,
+      "loss": 1.146,
+      "step": 10940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670586895499533e-05,
+      "loss": 1.3808,
+      "step": 10941
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670523632182278e-05,
+      "loss": 0.885,
+      "step": 10942
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6704603639889008e-05,
+      "loss": 1.1492,
+      "step": 10943
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670397090919861e-05,
+      "loss": 1.0328,
+      "step": 10944
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670333812975619e-05,
+      "loss": 0.7545,
+      "step": 10945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6702705301566354e-05,
+      "loss": 1.1046,
+      "step": 10946
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67020724246337e-05,
+      "loss": 1.1676,
+      "step": 10947
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670143949896283e-05,
+      "loss": 0.9129,
+      "step": 10948
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6700806524558347e-05,
+      "loss": 0.8264,
+      "step": 10949
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670017350142486e-05,
+      "loss": 1.164,
+      "step": 10950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669954042956696e-05,
+      "loss": 1.0514,
+      "step": 10951
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6698907308989263e-05,
+      "loss": 0.8072,
+      "step": 10952
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6698274139696368e-05,
+      "loss": 0.9889,
+      "step": 10953
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6697640921692878e-05,
+      "loss": 1.4093,
+      "step": 10954
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66970076549834e-05,
+      "loss": 0.9282,
+      "step": 10955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6696374339572538e-05,
+      "loss": 0.9808,
+      "step": 10956
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66957409754649e-05,
+      "loss": 1.214,
+      "step": 10957
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6695107562665086e-05,
+      "loss": 1.0237,
+      "step": 10958
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669447410117771e-05,
+      "loss": 1.0386,
+      "step": 10959
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669384059100737e-05,
+      "loss": 1.011,
+      "step": 10960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6693207032158687e-05,
+      "loss": 0.8939,
+      "step": 10961
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6692573424636254e-05,
+      "loss": 1.0865,
+      "step": 10962
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6691939768444683e-05,
+      "loss": 1.2411,
+      "step": 10963
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6691306063588583e-05,
+      "loss": 0.8842,
+      "step": 10964
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669067231007256e-05,
+      "loss": 0.99,
+      "step": 10965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669003850790123e-05,
+      "loss": 1.021,
+      "step": 10966
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66894046570792e-05,
+      "loss": 0.7157,
+      "step": 10967
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668877075761107e-05,
+      "loss": 0.8951,
+      "step": 10968
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6688136809501455e-05,
+      "loss": 1.204,
+      "step": 10969
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668750281275497e-05,
+      "loss": 0.9364,
+      "step": 10970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668686876737622e-05,
+      "loss": 0.9488,
+      "step": 10971
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668623467336982e-05,
+      "loss": 1.1971,
+      "step": 10972
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668560053074038e-05,
+      "loss": 1.0285,
+      "step": 10973
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6684966339492507e-05,
+      "loss": 0.8852,
+      "step": 10974
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668433209963082e-05,
+      "loss": 1.0333,
+      "step": 10975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683697811159922e-05,
+      "loss": 1.0848,
+      "step": 10976
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683063474084437e-05,
+      "loss": 0.51,
+      "step": 10977
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668242908840897e-05,
+      "loss": 1.0933,
+      "step": 10978
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6681794654138138e-05,
+      "loss": 0.9701,
+      "step": 10979
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668116017127655e-05,
+      "loss": 1.0165,
+      "step": 10980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6680525639828826e-05,
+      "loss": 1.0589,
+      "step": 10981
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667989105979958e-05,
+      "loss": 0.8837,
+      "step": 10982
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667925643119342e-05,
+      "loss": 0.8357,
+      "step": 10983
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6678621754014964e-05,
+      "loss": 0.8758,
+      "step": 10984
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6677987028268833e-05,
+      "loss": 1.5327,
+      "step": 10985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6677352253959637e-05,
+      "loss": 0.9969,
+      "step": 10986
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676717431091996e-05,
+      "loss": 1.1726,
+      "step": 10987
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676082559670523e-05,
+      "loss": 0.9565,
+      "step": 10988
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667544763969984e-05,
+      "loss": 1.1293,
+      "step": 10989
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6674812671184555e-05,
+      "loss": 0.7961,
+      "step": 10990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6674177654129292e-05,
+      "loss": 0.8562,
+      "step": 10991
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667354258853867e-05,
+      "loss": 1.0191,
+      "step": 10992
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6672907474417303e-05,
+      "loss": 0.9803,
+      "step": 10993
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6672272311769812e-05,
+      "loss": 0.5805,
+      "step": 10994
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667163710060082e-05,
+      "loss": 1.075,
+      "step": 10995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6671001840914933e-05,
+      "loss": 0.926,
+      "step": 10996
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6670366532716787e-05,
+      "loss": 0.8568,
+      "step": 10997
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6669731176010996e-05,
+      "loss": 1.197,
+      "step": 10998
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6669095770802177e-05,
+      "loss": 0.9057,
+      "step": 10999
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666846031709495e-05,
+      "loss": 1.3907,
+      "step": 11000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666782481489394e-05,
+      "loss": 0.8307,
+      "step": 11001
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6667189264203774e-05,
+      "loss": 0.9995,
+      "step": 11002
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666655366502906e-05,
+      "loss": 1.0737,
+      "step": 11003
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665918017374433e-05,
+      "loss": 1.0677,
+      "step": 11004
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665282321244505e-05,
+      "loss": 0.9763,
+      "step": 11005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6664646576643905e-05,
+      "loss": 1.1965,
+      "step": 11006
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6664010783577256e-05,
+      "loss": 0.8925,
+      "step": 11007
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666337494204918e-05,
+      "loss": 1.0214,
+      "step": 11008
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6662739052064302e-05,
+      "loss": 0.8594,
+      "step": 11009
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6662103113627246e-05,
+      "loss": 0.9082,
+      "step": 11010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6661467126742634e-05,
+      "loss": 0.9545,
+      "step": 11011
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.66608310914151e-05,
+      "loss": 0.6875,
+      "step": 11012
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6660195007649253e-05,
+      "loss": 1.1521,
+      "step": 11013
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6659558875449734e-05,
+      "loss": 1.274,
+      "step": 11014
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665892269482116e-05,
+      "loss": 0.6777,
+      "step": 11015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6658286465768165e-05,
+      "loss": 1.2941,
+      "step": 11016
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665765018829537e-05,
+      "loss": 1.225,
+      "step": 11017
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6657013862407404e-05,
+      "loss": 0.8422,
+      "step": 11018
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665637748810889e-05,
+      "loss": 0.8599,
+      "step": 11019
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665574106540446e-05,
+      "loss": 0.9231,
+      "step": 11020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665510459429875e-05,
+      "loss": 1.1226,
+      "step": 11021
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6654468074796374e-05,
+      "loss": 1.0702,
+      "step": 11022
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6653831506901966e-05,
+      "loss": 0.9858,
+      "step": 11023
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665319489062016e-05,
+      "loss": 0.8469,
+      "step": 11024
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6652558225955582e-05,
+      "loss": 1.3014,
+      "step": 11025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665192151291286e-05,
+      "loss": 0.9401,
+      "step": 11026
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6651284751496627e-05,
+      "loss": 1.0053,
+      "step": 11027
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665064794171151e-05,
+      "loss": 1.0125,
+      "step": 11028
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6650011083562146e-05,
+      "loss": 0.8023,
+      "step": 11029
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664937417705316e-05,
+      "loss": 0.9453,
+      "step": 11030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664873722218919e-05,
+      "loss": 0.6149,
+      "step": 11031
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6648100218974863e-05,
+      "loss": 1.0185,
+      "step": 11032
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6647463167414813e-05,
+      "loss": 0.7665,
+      "step": 11033
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664682606751367e-05,
+      "loss": 1.0444,
+      "step": 11034
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6646188919276073e-05,
+      "loss": 0.5744,
+      "step": 11035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6645551722706652e-05,
+      "loss": 1.0479,
+      "step": 11036
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664491447781004e-05,
+      "loss": 1.1766,
+      "step": 11037
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6644277184590874e-05,
+      "loss": 1.1355,
+      "step": 11038
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643639843053784e-05,
+      "loss": 1.0851,
+      "step": 11039
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643002453203405e-05,
+      "loss": 0.7255,
+      "step": 11040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664236501504438e-05,
+      "loss": 1.088,
+      "step": 11041
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641727528581337e-05,
+      "loss": 1.1034,
+      "step": 11042
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641089993818914e-05,
+      "loss": 0.7462,
+      "step": 11043
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6640452410761746e-05,
+      "loss": 1.218,
+      "step": 11044
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663981477941447e-05,
+      "loss": 1.3053,
+      "step": 11045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6639177099781723e-05,
+      "loss": 0.8486,
+      "step": 11046
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6638539371868145e-05,
+      "loss": 1.0416,
+      "step": 11047
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6637901595678375e-05,
+      "loss": 0.9966,
+      "step": 11048
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663726377121704e-05,
+      "loss": 0.9206,
+      "step": 11049
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663662589848879e-05,
+      "loss": 0.8907,
+      "step": 11050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6635987977498256e-05,
+      "loss": 0.7219,
+      "step": 11051
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6635350008250087e-05,
+      "loss": 0.8913,
+      "step": 11052
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6634711990748914e-05,
+      "loss": 1.2142,
+      "step": 11053
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6634073924999376e-05,
+      "loss": 0.9158,
+      "step": 11054
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6633435811006117e-05,
+      "loss": 0.8023,
+      "step": 11055
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6632797648773772e-05,
+      "loss": 1.152,
+      "step": 11056
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663215943830699e-05,
+      "loss": 1.0039,
+      "step": 11057
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631521179610408e-05,
+      "loss": 0.8857,
+      "step": 11058
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6630882872688668e-05,
+      "loss": 1.1859,
+      "step": 11059
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663024451754641e-05,
+      "loss": 1.0451,
+      "step": 11060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662960611418828e-05,
+      "loss": 0.7877,
+      "step": 11061
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6628967662618915e-05,
+      "loss": 0.8355,
+      "step": 11062
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6628329162842966e-05,
+      "loss": 0.9528,
+      "step": 11063
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627690614865065e-05,
+      "loss": 0.8307,
+      "step": 11064
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627052018689864e-05,
+      "loss": 1.1519,
+      "step": 11065
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6626413374322006e-05,
+      "loss": 0.9153,
+      "step": 11066
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6625774681766135e-05,
+      "loss": 1.0184,
+      "step": 11067
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6625135941026893e-05,
+      "loss": 1.1972,
+      "step": 11068
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6624497152108925e-05,
+      "loss": 0.9197,
+      "step": 11069
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662385831501688e-05,
+      "loss": 0.995,
+      "step": 11070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6623219429755406e-05,
+      "loss": 1.0408,
+      "step": 11071
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6622580496329138e-05,
+      "loss": 1.0208,
+      "step": 11072
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662194151474273e-05,
+      "loss": 0.9587,
+      "step": 11073
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6621302485000832e-05,
+      "loss": 1.0758,
+      "step": 11074
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6620663407108084e-05,
+      "loss": 1.0356,
+      "step": 11075
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662002428106914e-05,
+      "loss": 0.6043,
+      "step": 11076
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661938510688864e-05,
+      "loss": 1.0655,
+      "step": 11077
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6618745884571235e-05,
+      "loss": 1.0337,
+      "step": 11078
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661810661412158e-05,
+      "loss": 0.7865,
+      "step": 11079
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6617467295544315e-05,
+      "loss": 0.851,
+      "step": 11080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6616827928844096e-05,
+      "loss": 1.0605,
+      "step": 11081
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661618851402557e-05,
+      "loss": 0.8221,
+      "step": 11082
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615549051093384e-05,
+      "loss": 0.7231,
+      "step": 11083
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6614909540052192e-05,
+      "loss": 1.2302,
+      "step": 11084
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661426998090664e-05,
+      "loss": 0.9032,
+      "step": 11085
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6613630373661387e-05,
+      "loss": 0.8876,
+      "step": 11086
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6612990718321078e-05,
+      "loss": 0.9859,
+      "step": 11087
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6612351014890365e-05,
+      "loss": 0.7389,
+      "step": 11088
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6611711263373903e-05,
+      "loss": 1.0459,
+      "step": 11089
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661107146377634e-05,
+      "loss": 1.1977,
+      "step": 11090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661043161610233e-05,
+      "loss": 0.984,
+      "step": 11091
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660979172035653e-05,
+      "loss": 1.0783,
+      "step": 11092
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660915177654359e-05,
+      "loss": 1.3084,
+      "step": 11093
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6608511784668167e-05,
+      "loss": 0.9302,
+      "step": 11094
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660787174473491e-05,
+      "loss": 0.9246,
+      "step": 11095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6607231656748474e-05,
+      "loss": 0.9906,
+      "step": 11096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660659152071352e-05,
+      "loss": 0.7906,
+      "step": 11097
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605951336634697e-05,
+      "loss": 1.3502,
+      "step": 11098
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605311104516658e-05,
+      "loss": 0.9849,
+      "step": 11099
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6604670824364067e-05,
+      "loss": 0.9998,
+      "step": 11100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6604030496181578e-05,
+      "loss": 0.9724,
+      "step": 11101
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6603390119973842e-05,
+      "loss": 0.9862,
+      "step": 11102
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660274969574552e-05,
+      "loss": 1.0491,
+      "step": 11103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6602109223501274e-05,
+      "loss": 1.0077,
+      "step": 11104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6601468703245753e-05,
+      "loss": 1.3549,
+      "step": 11105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6600828134983616e-05,
+      "loss": 0.7492,
+      "step": 11106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660018751871953e-05,
+      "loss": 1.128,
+      "step": 11107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599546854458145e-05,
+      "loss": 0.8808,
+      "step": 11108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6598906142204117e-05,
+      "loss": 1.1292,
+      "step": 11109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659826538196212e-05,
+      "loss": 0.6781,
+      "step": 11110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65976245737368e-05,
+      "loss": 1.0414,
+      "step": 11111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6596983717532818e-05,
+      "loss": 0.8202,
+      "step": 11112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659634281335484e-05,
+      "loss": 1.0519,
+      "step": 11113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6595701861207526e-05,
+      "loss": 0.9449,
+      "step": 11114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6595060861095534e-05,
+      "loss": 1.0994,
+      "step": 11115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659441981302353e-05,
+      "loss": 1.0724,
+      "step": 11116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659377871699617e-05,
+      "loss": 1.1401,
+      "step": 11117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659313757301812e-05,
+      "loss": 0.7737,
+      "step": 11118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659249638109404e-05,
+      "loss": 0.8138,
+      "step": 11119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6591855141228598e-05,
+      "loss": 0.8077,
+      "step": 11120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659121385342645e-05,
+      "loss": 0.9628,
+      "step": 11121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6590572517692267e-05,
+      "loss": 0.9231,
+      "step": 11122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589931134030702e-05,
+      "loss": 0.8624,
+      "step": 11123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658928970244643e-05,
+      "loss": 1.039,
+      "step": 11124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6588648222944114e-05,
+      "loss": 0.9011,
+      "step": 11125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658800669552841e-05,
+      "loss": 1.204,
+      "step": 11126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6587365120203995e-05,
+      "loss": 1.3829,
+      "step": 11127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658672349697553e-05,
+      "loss": 0.9504,
+      "step": 11128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6586081825847676e-05,
+      "loss": 1.2516,
+      "step": 11129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6585440106825107e-05,
+      "loss": 0.9861,
+      "step": 11130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658479833991248e-05,
+      "loss": 1.0354,
+      "step": 11131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6584156525114477e-05,
+      "loss": 1.0066,
+      "step": 11132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658351466243575e-05,
+      "loss": 0.9758,
+      "step": 11133
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658287275188098e-05,
+      "loss": 1.0068,
+      "step": 11134
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6582230793454824e-05,
+      "loss": 0.7978,
+      "step": 11135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6581588787161953e-05,
+      "loss": 1.0547,
+      "step": 11136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658094673300704e-05,
+      "loss": 1.319,
+      "step": 11137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658030463099475e-05,
+      "loss": 1.2214,
+      "step": 11138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6579662481129757e-05,
+      "loss": 1.131,
+      "step": 11139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6579020283416726e-05,
+      "loss": 1.0656,
+      "step": 11140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6578378037860327e-05,
+      "loss": 0.8638,
+      "step": 11141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6577735744465234e-05,
+      "loss": 0.7279,
+      "step": 11142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6577093403236118e-05,
+      "loss": 1.0209,
+      "step": 11143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6576451014177647e-05,
+      "loss": 0.7953,
+      "step": 11144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6575808577294492e-05,
+      "loss": 0.9341,
+      "step": 11145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6575166092591327e-05,
+      "loss": 0.912,
+      "step": 11146
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657452356007283e-05,
+      "loss": 1.1346,
+      "step": 11147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573880979743663e-05,
+      "loss": 0.8129,
+      "step": 11148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573238351608502e-05,
+      "loss": 0.9604,
+      "step": 11149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6572595675672023e-05,
+      "loss": 1.3122,
+      "step": 11150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6571952951938898e-05,
+      "loss": 0.963,
+      "step": 11151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6571310180413802e-05,
+      "loss": 1.0398,
+      "step": 11152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6570667361101408e-05,
+      "loss": 0.691,
+      "step": 11153
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6570024494006394e-05,
+      "loss": 0.7747,
+      "step": 11154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656938157913343e-05,
+      "loss": 0.6547,
+      "step": 11155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6568738616487194e-05,
+      "loss": 0.9745,
+      "step": 11156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6568095606072363e-05,
+      "loss": 0.9501,
+      "step": 11157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6567452547893608e-05,
+      "loss": 0.818,
+      "step": 11158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6566809441955612e-05,
+      "loss": 0.9951,
+      "step": 11159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6566166288263046e-05,
+      "loss": 1.0043,
+      "step": 11160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656552308682059e-05,
+      "loss": 1.0116,
+      "step": 11161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6564879837632925e-05,
+      "loss": 0.9114,
+      "step": 11162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656423654070472e-05,
+      "loss": 0.9281,
+      "step": 11163
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6563593196040655e-05,
+      "loss": 0.9415,
+      "step": 11164
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656294980364542e-05,
+      "loss": 1.0388,
+      "step": 11165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656230636352368e-05,
+      "loss": 0.958,
+      "step": 11166
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656166287568012e-05,
+      "loss": 1.152,
+      "step": 11167
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656101934011942e-05,
+      "loss": 0.9069,
+      "step": 11168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656037575684626e-05,
+      "loss": 1.0474,
+      "step": 11169
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6559732125865316e-05,
+      "loss": 1.0373,
+      "step": 11170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6559088447181272e-05,
+      "loss": 1.2106,
+      "step": 11171
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6558444720798808e-05,
+      "loss": 0.7239,
+      "step": 11172
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655780094672261e-05,
+      "loss": 0.8627,
+      "step": 11173
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655715712495735e-05,
+      "loss": 0.7555,
+      "step": 11174
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6556513255507714e-05,
+      "loss": 1.0493,
+      "step": 11175
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655586933837839e-05,
+      "loss": 0.9096,
+      "step": 11176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6555225373574058e-05,
+      "loss": 0.9383,
+      "step": 11177
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6554581361099395e-05,
+      "loss": 0.8767,
+      "step": 11178
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655393730095909e-05,
+      "loss": 1.4828,
+      "step": 11179
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6553293193157824e-05,
+      "loss": 1.3588,
+      "step": 11180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655264903770028e-05,
+      "loss": 1.0864,
+      "step": 11181
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6552004834591152e-05,
+      "loss": 1.0396,
+      "step": 11182
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6551360583835114e-05,
+      "loss": 1.3238,
+      "step": 11183
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550716285436853e-05,
+      "loss": 0.8386,
+      "step": 11184
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550071939401053e-05,
+      "loss": 1.0223,
+      "step": 11185
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6549427545732406e-05,
+      "loss": 0.9445,
+      "step": 11186
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6548783104435595e-05,
+      "loss": 0.8465,
+      "step": 11187
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6548138615515304e-05,
+      "loss": 0.8835,
+      "step": 11188
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6547494078976224e-05,
+      "loss": 0.9504,
+      "step": 11189
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546849494823037e-05,
+      "loss": 1.0756,
+      "step": 11190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546204863060435e-05,
+      "loss": 1.0465,
+      "step": 11191
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6545560183693105e-05,
+      "loss": 0.928,
+      "step": 11192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654491545672573e-05,
+      "loss": 1.0658,
+      "step": 11193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654427068216301e-05,
+      "loss": 1.1503,
+      "step": 11194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6543625860009625e-05,
+      "loss": 1.0013,
+      "step": 11195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6542980990270267e-05,
+      "loss": 0.8775,
+      "step": 11196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654233607294962e-05,
+      "loss": 0.7596,
+      "step": 11197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654169110805238e-05,
+      "loss": 0.9214,
+      "step": 11198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654104609558324e-05,
+      "loss": 1.1094,
+      "step": 11199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6540401035546884e-05,
+      "loss": 1.0373,
+      "step": 11200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6539755927948e-05,
+      "loss": 1.1246,
+      "step": 11201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6539110772791292e-05,
+      "loss": 1.2572,
+      "step": 11202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6538465570081443e-05,
+      "loss": 1.0645,
+      "step": 11203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6537820319823145e-05,
+      "loss": 1.0685,
+      "step": 11204
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653717502202109e-05,
+      "loss": 1.0046,
+      "step": 11205
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6536529676679974e-05,
+      "loss": 1.1284,
+      "step": 11206
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653588428380449e-05,
+      "loss": 0.9609,
+      "step": 11207
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535238843399326e-05,
+      "loss": 0.9703,
+      "step": 11208
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653459335546918e-05,
+      "loss": 0.8924,
+      "step": 11209
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653394782001875e-05,
+      "loss": 0.968,
+      "step": 11210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6533302237052723e-05,
+      "loss": 1.0578,
+      "step": 11211
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6532656606575796e-05,
+      "loss": 1.0033,
+      "step": 11212
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6532010928592663e-05,
+      "loss": 0.8015,
+      "step": 11213
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6531365203108025e-05,
+      "loss": 1.3092,
+      "step": 11214
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6530719430126572e-05,
+      "loss": 1.0249,
+      "step": 11215
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6530073609653003e-05,
+      "loss": 1.075,
+      "step": 11216
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6529427741692012e-05,
+      "loss": 0.955,
+      "step": 11217
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65287818262483e-05,
+      "loss": 1.0883,
+      "step": 11218
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652813586332656e-05,
+      "loss": 0.8835,
+      "step": 11219
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652748985293149e-05,
+      "loss": 1.4461,
+      "step": 11220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6526843795067788e-05,
+      "loss": 1.0064,
+      "step": 11221
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6526197689740155e-05,
+      "loss": 0.8753,
+      "step": 11222
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6525551536953287e-05,
+      "loss": 1.0821,
+      "step": 11223
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6524905336711885e-05,
+      "loss": 0.7478,
+      "step": 11224
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6524259089020644e-05,
+      "loss": 1.0229,
+      "step": 11225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652361279388427e-05,
+      "loss": 1.1364,
+      "step": 11226
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652296645130746e-05,
+      "loss": 0.8849,
+      "step": 11227
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652232006129491e-05,
+      "loss": 0.8846,
+      "step": 11228
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6521673623851327e-05,
+      "loss": 1.1584,
+      "step": 11229
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6521027138981404e-05,
+      "loss": 1.1882,
+      "step": 11230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6520380606689853e-05,
+      "loss": 0.6986,
+      "step": 11231
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651973402698137e-05,
+      "loss": 1.1907,
+      "step": 11232
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6519087399860656e-05,
+      "loss": 1.1008,
+      "step": 11233
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6518440725332412e-05,
+      "loss": 0.9535,
+      "step": 11234
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6517794003401345e-05,
+      "loss": 1.1672,
+      "step": 11235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6517147234072158e-05,
+      "loss": 0.9218,
+      "step": 11236
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651650041734955e-05,
+      "loss": 0.9282,
+      "step": 11237
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6515853553238228e-05,
+      "loss": 1.5166,
+      "step": 11238
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6515206641742895e-05,
+      "loss": 0.9491,
+      "step": 11239
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651455968286826e-05,
+      "loss": 1.2717,
+      "step": 11240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6513912676619016e-05,
+      "loss": 0.9506,
+      "step": 11241
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6513265622999883e-05,
+      "loss": 1.2148,
+      "step": 11242
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6512618522015554e-05,
+      "loss": 1.0669,
+      "step": 11243
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6511971373670744e-05,
+      "loss": 0.8348,
+      "step": 11244
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651132417797015e-05,
+      "loss": 0.5822,
+      "step": 11245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651067693491849e-05,
+      "loss": 0.9047,
+      "step": 11246
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651002964452046e-05,
+      "loss": 0.8278,
+      "step": 11247
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6509382306780774e-05,
+      "loss": 0.8497,
+      "step": 11248
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6508734921704133e-05,
+      "loss": 0.8984,
+      "step": 11249
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650808748929525e-05,
+      "loss": 0.8079,
+      "step": 11250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6507440009558835e-05,
+      "loss": 0.9628,
+      "step": 11251
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6506792482499593e-05,
+      "loss": 1.1862,
+      "step": 11252
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650614490812223e-05,
+      "loss": 0.9175,
+      "step": 11253
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6505497286431464e-05,
+      "loss": 0.7803,
+      "step": 11254
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6504849617431994e-05,
+      "loss": 0.9501,
+      "step": 11255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6504201901128537e-05,
+      "loss": 0.9651,
+      "step": 11256
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65035541375258e-05,
+      "loss": 1.0586,
+      "step": 11257
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65029063266285e-05,
+      "loss": 1.0037,
+      "step": 11258
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6502258468441337e-05,
+      "loss": 0.9162,
+      "step": 11259
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6501610562969036e-05,
+      "loss": 1.1838,
+      "step": 11260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6500962610216294e-05,
+      "loss": 0.9136,
+      "step": 11261
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6500314610187834e-05,
+      "loss": 1.144,
+      "step": 11262
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6499666562888363e-05,
+      "loss": 0.7558,
+      "step": 11263
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6499018468322594e-05,
+      "loss": 1.0771,
+      "step": 11264
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6498370326495242e-05,
+      "loss": 1.2924,
+      "step": 11265
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6497722137411017e-05,
+      "loss": 0.9772,
+      "step": 11266
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649707390107464e-05,
+      "loss": 1.1371,
+      "step": 11267
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496425617490815e-05,
+      "loss": 1.0792,
+      "step": 11268
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6495777286664267e-05,
+      "loss": 1.3139,
+      "step": 11269
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64951289085997e-05,
+      "loss": 0.9219,
+      "step": 11270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6494480483301836e-05,
+      "loss": 0.9568,
+      "step": 11271
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649383201077539e-05,
+      "loss": 0.764,
+      "step": 11272
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6493183491025078e-05,
+      "loss": 0.8701,
+      "step": 11273
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6492534924055613e-05,
+      "loss": 0.7837,
+      "step": 11274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6491886309871712e-05,
+      "loss": 1.167,
+      "step": 11275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649123764847809e-05,
+      "loss": 0.8839,
+      "step": 11276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6490588939879473e-05,
+      "loss": 0.8444,
+      "step": 11277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648994018408057e-05,
+      "loss": 0.915,
+      "step": 11278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64892913810861e-05,
+      "loss": 1.2709,
+      "step": 11279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648864253090078e-05,
+      "loss": 0.7659,
+      "step": 11280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6487993633529338e-05,
+      "loss": 1.1443,
+      "step": 11281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648734468897648e-05,
+      "loss": 1.0576,
+      "step": 11282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6486695697246936e-05,
+      "loss": 0.9257,
+      "step": 11283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6486046658345415e-05,
+      "loss": 0.8157,
+      "step": 11284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6485397572276647e-05,
+      "loss": 0.9998,
+      "step": 11285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6484748439045345e-05,
+      "loss": 1.1576,
+      "step": 11286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6484099258656233e-05,
+      "loss": 1.0319,
+      "step": 11287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6483450031114033e-05,
+      "loss": 0.9233,
+      "step": 11288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6482800756423462e-05,
+      "loss": 1.2354,
+      "step": 11289
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6482151434589243e-05,
+      "loss": 1.0295,
+      "step": 11290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6481502065616102e-05,
+      "loss": 1.0725,
+      "step": 11291
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6480852649508756e-05,
+      "loss": 0.6936,
+      "step": 11292
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6480203186271934e-05,
+      "loss": 1.0374,
+      "step": 11293
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647955367591035e-05,
+      "loss": 0.9559,
+      "step": 11294
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478904118428735e-05,
+      "loss": 1.0898,
+      "step": 11295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478254513831807e-05,
+      "loss": 1.1476,
+      "step": 11296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6477604862124293e-05,
+      "loss": 0.8635,
+      "step": 11297
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647695516331092e-05,
+      "loss": 1.1903,
+      "step": 11298
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6476305417396407e-05,
+      "loss": 0.8154,
+      "step": 11299
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647565562438548e-05,
+      "loss": 0.5762,
+      "step": 11300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647500578428287e-05,
+      "loss": 1.0413,
+      "step": 11301
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6474355897093302e-05,
+      "loss": 1.0183,
+      "step": 11302
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6473705962821492e-05,
+      "loss": 1.1837,
+      "step": 11303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647305598147218e-05,
+      "loss": 0.9065,
+      "step": 11304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647240595305008e-05,
+      "loss": 0.9607,
+      "step": 11305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6471755877559927e-05,
+      "loss": 1.1712,
+      "step": 11306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647110575500645e-05,
+      "loss": 0.8799,
+      "step": 11307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6470455585394368e-05,
+      "loss": 1.0291,
+      "step": 11308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6469805368728417e-05,
+      "loss": 1.0549,
+      "step": 11309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6469155105013324e-05,
+      "loss": 0.8042,
+      "step": 11310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6468504794253814e-05,
+      "loss": 0.871,
+      "step": 11311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6467854436454624e-05,
+      "loss": 1.1205,
+      "step": 11312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6467204031620473e-05,
+      "loss": 0.7636,
+      "step": 11313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6466553579756096e-05,
+      "loss": 0.9572,
+      "step": 11314
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6465903080866225e-05,
+      "loss": 1.0103,
+      "step": 11315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646525253495559e-05,
+      "loss": 1.0408,
+      "step": 11316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646460194202892e-05,
+      "loss": 0.8979,
+      "step": 11317
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646395130209095e-05,
+      "loss": 0.7723,
+      "step": 11318
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64633006151464e-05,
+      "loss": 1.0755,
+      "step": 11319
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6462649881200015e-05,
+      "loss": 0.8874,
+      "step": 11320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6461999100256524e-05,
+      "loss": 0.6759,
+      "step": 11321
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6461348272320658e-05,
+      "loss": 1.1278,
+      "step": 11322
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6460697397397147e-05,
+      "loss": 0.9931,
+      "step": 11323
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6460046475490728e-05,
+      "loss": 0.7948,
+      "step": 11324
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6459395506606133e-05,
+      "loss": 1.0639,
+      "step": 11325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6458744490748097e-05,
+      "loss": 1.2102,
+      "step": 11326
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6458093427921355e-05,
+      "loss": 1.0023,
+      "step": 11327
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645744231813064e-05,
+      "loss": 1.0694,
+      "step": 11328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456791161380685e-05,
+      "loss": 0.8882,
+      "step": 11329
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456139957676233e-05,
+      "loss": 0.8409,
+      "step": 11330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645548870702201e-05,
+      "loss": 0.8787,
+      "step": 11331
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645483740942276e-05,
+      "loss": 0.7232,
+      "step": 11332
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645418606488321e-05,
+      "loss": 0.8123,
+      "step": 11333
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6453534673408104e-05,
+      "loss": 1.0551,
+      "step": 11334
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6452883235002176e-05,
+      "loss": 0.5157,
+      "step": 11335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645223174967017e-05,
+      "loss": 0.7155,
+      "step": 11336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6451580217416812e-05,
+      "loss": 1.0894,
+      "step": 11337
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6450928638246853e-05,
+      "loss": 1.187,
+      "step": 11338
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645027701216502e-05,
+      "loss": 1.1545,
+      "step": 11339
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6449625339176056e-05,
+      "loss": 1.1038,
+      "step": 11340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64489736192847e-05,
+      "loss": 1.0148,
+      "step": 11341
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6448321852495694e-05,
+      "loss": 0.948,
+      "step": 11342
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6447670038813773e-05,
+      "loss": 1.1874,
+      "step": 11343
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6447018178243684e-05,
+      "loss": 0.7203,
+      "step": 11344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6446366270790158e-05,
+      "loss": 0.7822,
+      "step": 11345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6445714316457942e-05,
+      "loss": 0.7205,
+      "step": 11346
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6445062315251777e-05,
+      "loss": 0.9554,
+      "step": 11347
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6444410267176403e-05,
+      "loss": 0.881,
+      "step": 11348
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6443758172236562e-05,
+      "loss": 0.9168,
+      "step": 11349
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6443106030436996e-05,
+      "loss": 1.026,
+      "step": 11350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442453841782443e-05,
+      "loss": 0.8756,
+      "step": 11351
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6441801606277657e-05,
+      "loss": 1.0919,
+      "step": 11352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644114932392737e-05,
+      "loss": 1.0779,
+      "step": 11353
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6440496994736332e-05,
+      "loss": 1.1565,
+      "step": 11354
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6439844618709285e-05,
+      "loss": 1.0345,
+      "step": 11355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643919219585097e-05,
+      "loss": 1.2074,
+      "step": 11356
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6438539726166135e-05,
+      "loss": 0.7396,
+      "step": 11357
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437887209659528e-05,
+      "loss": 1.0732,
+      "step": 11358
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6437234646335885e-05,
+      "loss": 0.6696,
+      "step": 11359
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643658203619996e-05,
+      "loss": 0.9299,
+      "step": 11360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6435929379256494e-05,
+      "loss": 1.0526,
+      "step": 11361
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6435276675510235e-05,
+      "loss": 1.188,
+      "step": 11362
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643462392496593e-05,
+      "loss": 1.0166,
+      "step": 11363
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643397112762833e-05,
+      "loss": 1.096,
+      "step": 11364
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6433318283502168e-05,
+      "loss": 1.326,
+      "step": 11365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6432665392592206e-05,
+      "loss": 1.0263,
+      "step": 11366
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643201245490319e-05,
+      "loss": 1.1562,
+      "step": 11367
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643135947043986e-05,
+      "loss": 1.4041,
+      "step": 11368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643070643920697e-05,
+      "loss": 0.6398,
+      "step": 11369
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6430053361209274e-05,
+      "loss": 1.0352,
+      "step": 11370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6429400236451513e-05,
+      "loss": 1.1309,
+      "step": 11371
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6428747064938438e-05,
+      "loss": 0.8527,
+      "step": 11372
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64280938466748e-05,
+      "loss": 1.0668,
+      "step": 11373
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427440581665352e-05,
+      "loss": 0.8212,
+      "step": 11374
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6426787269914842e-05,
+      "loss": 1.129,
+      "step": 11375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6426133911428016e-05,
+      "loss": 1.1418,
+      "step": 11376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6425480506209636e-05,
+      "loss": 1.0392,
+      "step": 11377
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642482705426445e-05,
+      "loss": 0.6914,
+      "step": 11378
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6424173555597205e-05,
+      "loss": 1.0699,
+      "step": 11379
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6423520010212657e-05,
+      "loss": 0.9747,
+      "step": 11380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642286641811556e-05,
+      "loss": 1.3438,
+      "step": 11381
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6422212779310664e-05,
+      "loss": 1.4407,
+      "step": 11382
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6421559093802722e-05,
+      "loss": 0.9501,
+      "step": 11383
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642090536159649e-05,
+      "loss": 1.0614,
+      "step": 11384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642025158269672e-05,
+      "loss": 0.8433,
+      "step": 11385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641959775710817e-05,
+      "loss": 0.7263,
+      "step": 11386
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641894388483559e-05,
+      "loss": 0.9882,
+      "step": 11387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641828996588374e-05,
+      "loss": 1.077,
+      "step": 11388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641763600025737e-05,
+      "loss": 1.0124,
+      "step": 11389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641698198796124e-05,
+      "loss": 0.836,
+      "step": 11390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6416327929000103e-05,
+      "loss": 1.453,
+      "step": 11391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641567382337872e-05,
+      "loss": 0.8546,
+      "step": 11392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641501967110184e-05,
+      "loss": 0.8414,
+      "step": 11393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6414365472174228e-05,
+      "loss": 1.1509,
+      "step": 11394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413711226600634e-05,
+      "loss": 0.9227,
+      "step": 11395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413056934385822e-05,
+      "loss": 1.3299,
+      "step": 11396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641240259553455e-05,
+      "loss": 0.9127,
+      "step": 11397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6411748210051573e-05,
+      "loss": 0.8934,
+      "step": 11398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6411093777941653e-05,
+      "loss": 0.983,
+      "step": 11399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641043929920954e-05,
+      "loss": 1.0446,
+      "step": 11400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6409784773860007e-05,
+      "loss": 1.2882,
+      "step": 11401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6409130201897805e-05,
+      "loss": 1.1351,
+      "step": 11402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6408475583327696e-05,
+      "loss": 0.7045,
+      "step": 11403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6407820918154445e-05,
+      "loss": 1.0977,
+      "step": 11404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6407166206382805e-05,
+      "loss": 0.9801,
+      "step": 11405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6406511448017538e-05,
+      "loss": 0.9374,
+      "step": 11406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6405856643063413e-05,
+      "loss": 0.8967,
+      "step": 11407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6405201791525187e-05,
+      "loss": 0.9287,
+      "step": 11408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640454689340762e-05,
+      "loss": 0.7899,
+      "step": 11409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6403891948715477e-05,
+      "loss": 1.012,
+      "step": 11410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6403236957453522e-05,
+      "loss": 0.9338,
+      "step": 11411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640258191962652e-05,
+      "loss": 0.7372,
+      "step": 11412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6401926835239227e-05,
+      "loss": 0.9848,
+      "step": 11413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640127170429641e-05,
+      "loss": 0.916,
+      "step": 11414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6400616526802835e-05,
+      "loss": 1.2177,
+      "step": 11415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639996130276327e-05,
+      "loss": 1.1539,
+      "step": 11416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6399306032182472e-05,
+      "loss": 0.9093,
+      "step": 11417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639865071506521e-05,
+      "loss": 1.7212,
+      "step": 11418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639799535141625e-05,
+      "loss": 0.7886,
+      "step": 11419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6397339941240355e-05,
+      "loss": 0.8676,
+      "step": 11420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63966844845423e-05,
+      "loss": 0.9671,
+      "step": 11421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639602898132684e-05,
+      "loss": 1.0553,
+      "step": 11422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6395373431598745e-05,
+      "loss": 1.1197,
+      "step": 11423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639471783536279e-05,
+      "loss": 0.7754,
+      "step": 11424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639406219262373e-05,
+      "loss": 1.1905,
+      "step": 11425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6393406503386345e-05,
+      "loss": 1.0247,
+      "step": 11426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63927507676554e-05,
+      "loss": 0.9554,
+      "step": 11427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6392094985435653e-05,
+      "loss": 0.8435,
+      "step": 11428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6391439156731887e-05,
+      "loss": 0.9645,
+      "step": 11429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6390783281548865e-05,
+      "loss": 0.9076,
+      "step": 11430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6390127359891358e-05,
+      "loss": 1.1483,
+      "step": 11431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6389471391764136e-05,
+      "loss": 1.5789,
+      "step": 11432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388815377171966e-05,
+      "loss": 1.0012,
+      "step": 11433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388159316119623e-05,
+      "loss": 0.8945,
+      "step": 11434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6387503208611875e-05,
+      "loss": 1.0303,
+      "step": 11435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6386847054653495e-05,
+      "loss": 0.8552,
+      "step": 11436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6386190854249256e-05,
+      "loss": 1.0221,
+      "step": 11437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6385534607403922e-05,
+      "loss": 1.1309,
+      "step": 11438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6384878314122276e-05,
+      "loss": 1.306,
+      "step": 11439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6384221974409083e-05,
+      "loss": 1.1302,
+      "step": 11440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6383565588269123e-05,
+      "loss": 0.8406,
+      "step": 11441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382909155707163e-05,
+      "loss": 0.7977,
+      "step": 11442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.638225267672798e-05,
+      "loss": 0.9533,
+      "step": 11443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6381596151336342e-05,
+      "loss": 1.1075,
+      "step": 11444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6380939579537033e-05,
+      "loss": 0.9069,
+      "step": 11445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6380282961334823e-05,
+      "loss": 0.859,
+      "step": 11446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6379626296734486e-05,
+      "loss": 1.1751,
+      "step": 11447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6378969585740795e-05,
+      "loss": 0.8042,
+      "step": 11448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6378312828358532e-05,
+      "loss": 1.4655,
+      "step": 11449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637765602459247e-05,
+      "loss": 0.8712,
+      "step": 11450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637699917444738e-05,
+      "loss": 0.8093,
+      "step": 11451
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637634227792805e-05,
+      "loss": 0.873,
+      "step": 11452
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637568533503925e-05,
+      "loss": 1.0092,
+      "step": 11453
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6375028345785756e-05,
+      "loss": 0.9384,
+      "step": 11454
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6374371310172346e-05,
+      "loss": 0.8888,
+      "step": 11455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6373714228203805e-05,
+      "loss": 1.152,
+      "step": 11456
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6373057099884907e-05,
+      "loss": 0.6481,
+      "step": 11457
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372399925220427e-05,
+      "loss": 0.9363,
+      "step": 11458
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6371742704215145e-05,
+      "loss": 1.0166,
+      "step": 11459
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6371085436873847e-05,
+      "loss": 1.1209,
+      "step": 11460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6370428123201306e-05,
+      "loss": 1.0673,
+      "step": 11461
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6369770763202305e-05,
+      "loss": 1.1851,
+      "step": 11462
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6369113356881623e-05,
+      "loss": 1.1507,
+      "step": 11463
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636845590424404e-05,
+      "loss": 0.7908,
+      "step": 11464
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636779840529434e-05,
+      "loss": 0.8071,
+      "step": 11465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6367140860037302e-05,
+      "loss": 1.0874,
+      "step": 11466
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636648326847771e-05,
+      "loss": 1.2085,
+      "step": 11467
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6365825630620346e-05,
+      "loss": 1.2775,
+      "step": 11468
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6365167946469987e-05,
+      "loss": 1.0568,
+      "step": 11469
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364510216031423e-05,
+      "loss": 1.0564,
+      "step": 11470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6363852439309433e-05,
+      "loss": 1.0753,
+      "step": 11471
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6363194616308802e-05,
+      "loss": 1.0358,
+      "step": 11472
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636253674703431e-05,
+      "loss": 1.0287,
+      "step": 11473
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636187883149075e-05,
+      "loss": 0.9201,
+      "step": 11474
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6361220869682896e-05,
+      "loss": 1.0546,
+      "step": 11475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6360562861615538e-05,
+      "loss": 0.9523,
+      "step": 11476
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635990480729346e-05,
+      "loss": 0.9076,
+      "step": 11477
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635924670672145e-05,
+      "loss": 0.8182,
+      "step": 11478
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6358588559904292e-05,
+      "loss": 0.9068,
+      "step": 11479
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635793036684677e-05,
+      "loss": 0.8856,
+      "step": 11480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6357272127553675e-05,
+      "loss": 1.0148,
+      "step": 11481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6356613842029792e-05,
+      "loss": 0.8697,
+      "step": 11482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6355955510279902e-05,
+      "loss": 1.0697,
+      "step": 11483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63552971323088e-05,
+      "loss": 0.9623,
+      "step": 11484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6354638708121273e-05,
+      "loss": 0.7467,
+      "step": 11485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6353980237722106e-05,
+      "loss": 0.8929,
+      "step": 11486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635332172111609e-05,
+      "loss": 1.068,
+      "step": 11487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6352663158308014e-05,
+      "loss": 1.1014,
+      "step": 11488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6352004549302665e-05,
+      "loss": 1.0651,
+      "step": 11489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635134589410483e-05,
+      "loss": 0.8979,
+      "step": 11490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6350687192719308e-05,
+      "loss": 1.0382,
+      "step": 11491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6350028445150882e-05,
+      "loss": 1.2337,
+      "step": 11492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6349369651404344e-05,
+      "loss": 1.0624,
+      "step": 11493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6348710811484484e-05,
+      "loss": 1.1167,
+      "step": 11494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6348051925396095e-05,
+      "loss": 0.8919,
+      "step": 11495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6347392993143962e-05,
+      "loss": 1.1959,
+      "step": 11496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634673401473289e-05,
+      "loss": 0.6108,
+      "step": 11497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634607499016766e-05,
+      "loss": 1.2931,
+      "step": 11498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6345415919453067e-05,
+      "loss": 0.9866,
+      "step": 11499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6344756802593905e-05,
+      "loss": 0.7559,
+      "step": 11500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6344097639594968e-05,
+      "loss": 1.2797,
+      "step": 11501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634343843046105e-05,
+      "loss": 0.8639,
+      "step": 11502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6342779175196942e-05,
+      "loss": 0.9564,
+      "step": 11503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6342119873807438e-05,
+      "loss": 0.8848,
+      "step": 11504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6341460526297335e-05,
+      "loss": 0.9439,
+      "step": 11505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6340801132671425e-05,
+      "loss": 1.1861,
+      "step": 11506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634014169293451e-05,
+      "loss": 0.828,
+      "step": 11507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633948220709138e-05,
+      "loss": 0.9032,
+      "step": 11508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6338822675146825e-05,
+      "loss": 0.9416,
+      "step": 11509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6338163097105654e-05,
+      "loss": 1.0829,
+      "step": 11510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6337503472972653e-05,
+      "loss": 0.9846,
+      "step": 11511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6336843802752626e-05,
+      "loss": 0.8138,
+      "step": 11512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6336184086450365e-05,
+      "loss": 0.93,
+      "step": 11513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633552432407067e-05,
+      "loss": 1.1927,
+      "step": 11514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6334864515618337e-05,
+      "loss": 0.9501,
+      "step": 11515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633420466109817e-05,
+      "loss": 0.9782,
+      "step": 11516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6333544760514965e-05,
+      "loss": 1.2199,
+      "step": 11517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6332884813873512e-05,
+      "loss": 0.9371,
+      "step": 11518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633222482117862e-05,
+      "loss": 1.1445,
+      "step": 11519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6331564782435087e-05,
+      "loss": 0.9509,
+      "step": 11520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633090469764771e-05,
+      "loss": 0.8721,
+      "step": 11521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6330244566821294e-05,
+      "loss": 0.9621,
+      "step": 11522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6329584389960636e-05,
+      "loss": 0.8157,
+      "step": 11523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6328924167070538e-05,
+      "loss": 1.0412,
+      "step": 11524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63282638981558e-05,
+      "loss": 0.827,
+      "step": 11525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632760358322122e-05,
+      "loss": 0.8051,
+      "step": 11526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632694322227161e-05,
+      "loss": 0.9638,
+      "step": 11527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6326282815311764e-05,
+      "loss": 1.0558,
+      "step": 11528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6325622362346484e-05,
+      "loss": 1.1244,
+      "step": 11529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632496186338058e-05,
+      "loss": 1.2446,
+      "step": 11530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632430131841885e-05,
+      "loss": 0.9889,
+      "step": 11531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63236407274661e-05,
+      "loss": 0.762,
+      "step": 11532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6322980090527125e-05,
+      "loss": 0.9781,
+      "step": 11533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6322319407606745e-05,
+      "loss": 0.9515,
+      "step": 11534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6321658678709752e-05,
+      "loss": 0.858,
+      "step": 11535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6320997903840954e-05,
+      "loss": 0.8846,
+      "step": 11536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632033708300516e-05,
+      "loss": 1.0557,
+      "step": 11537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6319676216207176e-05,
+      "loss": 1.2073,
+      "step": 11538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6319015303451798e-05,
+      "loss": 0.8529,
+      "step": 11539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6318354344743843e-05,
+      "loss": 1.0897,
+      "step": 11540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631769334008811e-05,
+      "loss": 0.8421,
+      "step": 11541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6317032289489418e-05,
+      "loss": 0.932,
+      "step": 11542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6316371192952556e-05,
+      "loss": 0.754,
+      "step": 11543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6315710050482345e-05,
+      "loss": 0.836,
+      "step": 11544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6315048862083588e-05,
+      "loss": 1.028,
+      "step": 11545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6314387627761097e-05,
+      "loss": 0.9938,
+      "step": 11546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631372634751968e-05,
+      "loss": 1.0416,
+      "step": 11547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631306502136414e-05,
+      "loss": 0.8485,
+      "step": 11548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6312403649299287e-05,
+      "loss": 1.3045,
+      "step": 11549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6311742231329936e-05,
+      "loss": 1.2327,
+      "step": 11550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63110807674609e-05,
+      "loss": 0.8963,
+      "step": 11551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6310419257696978e-05,
+      "loss": 1.2417,
+      "step": 11552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630975770204299e-05,
+      "loss": 1.2425,
+      "step": 11553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6309096100503737e-05,
+      "loss": 1.0022,
+      "step": 11554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6308434453084043e-05,
+      "loss": 1.1227,
+      "step": 11555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6307772759788713e-05,
+      "loss": 1.1562,
+      "step": 11556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6307111020622554e-05,
+      "loss": 0.9581,
+      "step": 11557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630644923559039e-05,
+      "loss": 1.2281,
+      "step": 11558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630578740469702e-05,
+      "loss": 1.0679,
+      "step": 11559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630512552794727e-05,
+      "loss": 0.8308,
+      "step": 11560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630446360534594e-05,
+      "loss": 1.4635,
+      "step": 11561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6303801636897857e-05,
+      "loss": 1.2033,
+      "step": 11562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6303139622607824e-05,
+      "loss": 0.8682,
+      "step": 11563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6302477562480664e-05,
+      "loss": 0.9344,
+      "step": 11564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6301815456521185e-05,
+      "loss": 0.766,
+      "step": 11565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6301153304734205e-05,
+      "loss": 0.8479,
+      "step": 11566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6300491107124538e-05,
+      "loss": 0.9242,
+      "step": 11567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6299828863697e-05,
+      "loss": 0.691,
+      "step": 11568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6299166574456405e-05,
+      "loss": 0.9696,
+      "step": 11569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629850423940757e-05,
+      "loss": 1.0426,
+      "step": 11570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629784185855532e-05,
+      "loss": 0.8522,
+      "step": 11571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6297179431904458e-05,
+      "loss": 0.7714,
+      "step": 11572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6296516959459813e-05,
+      "loss": 0.6954,
+      "step": 11573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6295854441226193e-05,
+      "loss": 0.8996,
+      "step": 11574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629519187720842e-05,
+      "loss": 0.7968,
+      "step": 11575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6294529267411317e-05,
+      "loss": 1.4163,
+      "step": 11576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6293866611839694e-05,
+      "loss": 0.7612,
+      "step": 11577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6293203910498375e-05,
+      "loss": 1.0832,
+      "step": 11578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629254116339218e-05,
+      "loss": 1.0802,
+      "step": 11579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6291878370525925e-05,
+      "loss": 1.2214,
+      "step": 11580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6291215531904434e-05,
+      "loss": 0.9602,
+      "step": 11581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6290552647532526e-05,
+      "loss": 0.9642,
+      "step": 11582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6289889717415017e-05,
+      "loss": 1.2742,
+      "step": 11583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6289226741556733e-05,
+      "loss": 1.3394,
+      "step": 11584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6288563719962495e-05,
+      "loss": 0.9405,
+      "step": 11585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6287900652637118e-05,
+      "loss": 1.5448,
+      "step": 11586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6287237539585436e-05,
+      "loss": 1.091,
+      "step": 11587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628657438081226e-05,
+      "loss": 0.9483,
+      "step": 11588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628591117632242e-05,
+      "loss": 0.8444,
+      "step": 11589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6285247926120732e-05,
+      "loss": 1.2348,
+      "step": 11590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6284584630212024e-05,
+      "loss": 1.2672,
+      "step": 11591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6283921288601123e-05,
+      "loss": 0.996,
+      "step": 11592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6283257901292847e-05,
+      "loss": 1.1858,
+      "step": 11593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628259446829202e-05,
+      "loss": 1.0751,
+      "step": 11594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6281930989603466e-05,
+      "loss": 1.0333,
+      "step": 11595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6281267465232015e-05,
+      "loss": 1.2567,
+      "step": 11596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6280603895182494e-05,
+      "loss": 0.7872,
+      "step": 11597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627994027945972e-05,
+      "loss": 1.0857,
+      "step": 11598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6279276618068523e-05,
+      "loss": 0.7015,
+      "step": 11599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6278612911013726e-05,
+      "loss": 1.1356,
+      "step": 11600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627794915830016e-05,
+      "loss": 0.9766,
+      "step": 11601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6277285359932654e-05,
+      "loss": 1.0181,
+      "step": 11602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627662151591603e-05,
+      "loss": 1.0484,
+      "step": 11603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6275957626255118e-05,
+      "loss": 0.7739,
+      "step": 11604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6275293690954744e-05,
+      "loss": 1.2215,
+      "step": 11605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6274629710019735e-05,
+      "loss": 1.2258,
+      "step": 11606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6273965683454925e-05,
+      "loss": 1.138,
+      "step": 11607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6273301611265143e-05,
+      "loss": 1.0593,
+      "step": 11608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627263749345521e-05,
+      "loss": 1.1104,
+      "step": 11609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627197333002996e-05,
+      "loss": 0.7343,
+      "step": 11610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6271309120994227e-05,
+      "loss": 0.9048,
+      "step": 11611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6270644866352837e-05,
+      "loss": 1.163,
+      "step": 11612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6269980566110622e-05,
+      "loss": 0.9547,
+      "step": 11613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626931622027241e-05,
+      "loss": 0.8681,
+      "step": 11614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6268651828843035e-05,
+      "loss": 0.9514,
+      "step": 11615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6267987391827326e-05,
+      "loss": 1.0208,
+      "step": 11616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6267322909230122e-05,
+      "loss": 1.0325,
+      "step": 11617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6266658381056243e-05,
+      "loss": 1.0108,
+      "step": 11618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626599380731053e-05,
+      "loss": 1.3343,
+      "step": 11619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626532918799782e-05,
+      "loss": 1.0119,
+      "step": 11620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6264664523122932e-05,
+      "loss": 0.8239,
+      "step": 11621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263999812690717e-05,
+      "loss": 1.0339,
+      "step": 11622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263335056705993e-05,
+      "loss": 0.7736,
+      "step": 11623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6262670255173602e-05,
+      "loss": 0.7215,
+      "step": 11624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6262005408098378e-05,
+      "loss": 1.3315,
+      "step": 11625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261340515485152e-05,
+      "loss": 0.8857,
+      "step": 11626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6260675577338768e-05,
+      "loss": 1.0414,
+      "step": 11627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6260010593664055e-05,
+      "loss": 1.1304,
+      "step": 11628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6259345564465848e-05,
+      "loss": 0.6502,
+      "step": 11629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258680489748984e-05,
+      "loss": 0.7575,
+      "step": 11630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6258015369518303e-05,
+      "loss": 0.9881,
+      "step": 11631
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6257350203778642e-05,
+      "loss": 1.2762,
+      "step": 11632
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625668499253483e-05,
+      "loss": 0.9268,
+      "step": 11633
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6256019735791714e-05,
+      "loss": 1.1998,
+      "step": 11634
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6255354433554123e-05,
+      "loss": 1.1016,
+      "step": 11635
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6254689085826903e-05,
+      "loss": 1.0192,
+      "step": 11636
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6254023692614887e-05,
+      "loss": 0.9025,
+      "step": 11637
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6253358253922918e-05,
+      "loss": 0.6977,
+      "step": 11638
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6252692769755837e-05,
+      "loss": 0.8135,
+      "step": 11639
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6252027240118472e-05,
+      "loss": 1.0271,
+      "step": 11640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6251361665015675e-05,
+      "loss": 1.2014,
+      "step": 11641
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6250696044452285e-05,
+      "loss": 0.7313,
+      "step": 11642
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6250030378433136e-05,
+      "loss": 1.2092,
+      "step": 11643
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624936466696307e-05,
+      "loss": 0.9028,
+      "step": 11644
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6248698910046934e-05,
+      "loss": 1.0025,
+      "step": 11645
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624803310768956e-05,
+      "loss": 0.7584,
+      "step": 11646
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6247367259895803e-05,
+      "loss": 1.071,
+      "step": 11647
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6246701366670494e-05,
+      "loss": 1.1807,
+      "step": 11648
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624603542801848e-05,
+      "loss": 0.8076,
+      "step": 11649
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6245369443944598e-05,
+      "loss": 0.8056,
+      "step": 11650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62447034144537e-05,
+      "loss": 1.1333,
+      "step": 11651
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6244037339550626e-05,
+      "loss": 1.4955,
+      "step": 11652
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624337121924022e-05,
+      "loss": 1.254,
+      "step": 11653
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6242705053527322e-05,
+      "loss": 1.1555,
+      "step": 11654
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624203884241678e-05,
+      "loss": 0.9269,
+      "step": 11655
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624137258591344e-05,
+      "loss": 1.023,
+      "step": 11656
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6240706284022143e-05,
+      "loss": 1.1648,
+      "step": 11657
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6240039936747737e-05,
+      "loss": 1.1069,
+      "step": 11658
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6239373544095068e-05,
+      "loss": 1.0336,
+      "step": 11659
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6238707106068986e-05,
+      "loss": 0.9189,
+      "step": 11660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623804062267433e-05,
+      "loss": 1.3979,
+      "step": 11661
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6237374093915947e-05,
+      "loss": 0.9518,
+      "step": 11662
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623670751979869e-05,
+      "loss": 1.1701,
+      "step": 11663
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6236040900327404e-05,
+      "loss": 1.1239,
+      "step": 11664
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6235374235506936e-05,
+      "loss": 1.0151,
+      "step": 11665
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623470752534213e-05,
+      "loss": 0.826,
+      "step": 11666
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623404076983784e-05,
+      "loss": 1.0159,
+      "step": 11667
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6233373968998918e-05,
+      "loss": 0.8553,
+      "step": 11668
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6232707122830203e-05,
+      "loss": 1.1697,
+      "step": 11669
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6232040231336556e-05,
+      "loss": 0.9453,
+      "step": 11670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6231373294522813e-05,
+      "loss": 1.039,
+      "step": 11671
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623070631239384e-05,
+      "loss": 0.9328,
+      "step": 11672
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623003928495447e-05,
+      "loss": 1.2641,
+      "step": 11673
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6229372212209567e-05,
+      "loss": 1.0271,
+      "step": 11674
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6228705094163975e-05,
+      "loss": 1.1516,
+      "step": 11675
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622803793082255e-05,
+      "loss": 0.673,
+      "step": 11676
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622737072219014e-05,
+      "loss": 1.213,
+      "step": 11677
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62267034682716e-05,
+      "loss": 0.7379,
+      "step": 11678
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6226036169071783e-05,
+      "loss": 1.0835,
+      "step": 11679
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6225368824595534e-05,
+      "loss": 1.12,
+      "step": 11680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6224701434847715e-05,
+      "loss": 0.6022,
+      "step": 11681
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6224033999833174e-05,
+      "loss": 1.1389,
+      "step": 11682
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622336651955677e-05,
+      "loss": 0.887,
+      "step": 11683
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6222698994023353e-05,
+      "loss": 0.8,
+      "step": 11684
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6222031423237776e-05,
+      "loss": 1.0726,
+      "step": 11685
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6221363807204895e-05,
+      "loss": 0.8732,
+      "step": 11686
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6220696145929568e-05,
+      "loss": 1.0651,
+      "step": 11687
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6220028439416648e-05,
+      "loss": 0.9796,
+      "step": 11688
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219360687670986e-05,
+      "loss": 0.8987,
+      "step": 11689
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6218692890697444e-05,
+      "loss": 1.0961,
+      "step": 11690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621802504850088e-05,
+      "loss": 0.8459,
+      "step": 11691
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6217357161086148e-05,
+      "loss": 0.8432,
+      "step": 11692
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62166892284581e-05,
+      "loss": 1.3177,
+      "step": 11693
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6216021250621602e-05,
+      "loss": 1.0619,
+      "step": 11694
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6215353227581502e-05,
+      "loss": 0.8862,
+      "step": 11695
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6214685159342667e-05,
+      "loss": 1.0116,
+      "step": 11696
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6214017045909952e-05,
+      "loss": 1.3388,
+      "step": 11697
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621334888728821e-05,
+      "loss": 1.0395,
+      "step": 11698
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621268068348231e-05,
+      "loss": 1.0778,
+      "step": 11699
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6212012434497103e-05,
+      "loss": 1.0297,
+      "step": 11700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6211344140337454e-05,
+      "loss": 0.7255,
+      "step": 11701
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6210675801008218e-05,
+      "loss": 1.0381,
+      "step": 11702
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621000741651426e-05,
+      "loss": 1.4443,
+      "step": 11703
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620933898686044e-05,
+      "loss": 0.8722,
+      "step": 11704
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620867051205161e-05,
+      "loss": 1.0101,
+      "step": 11705
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620800199209265e-05,
+      "loss": 0.9976,
+      "step": 11706
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62073334269884e-05,
+      "loss": 1.0978,
+      "step": 11707
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206664816743737e-05,
+      "loss": 1.046,
+      "step": 11708
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6205996161363516e-05,
+      "loss": 1.0541,
+      "step": 11709
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6205327460852604e-05,
+      "loss": 0.9501,
+      "step": 11710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620465871521586e-05,
+      "loss": 0.6284,
+      "step": 11711
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620398992445815e-05,
+      "loss": 0.9459,
+      "step": 11712
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6203321088584337e-05,
+      "loss": 0.9771,
+      "step": 11713
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620265220759928e-05,
+      "loss": 1.2886,
+      "step": 11714
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620198328150785e-05,
+      "loss": 1.099,
+      "step": 11715
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6201314310314913e-05,
+      "loss": 1.2262,
+      "step": 11716
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6200645294025324e-05,
+      "loss": 0.9233,
+      "step": 11717
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619997623264396e-05,
+      "loss": 0.9659,
+      "step": 11718
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6199307126175673e-05,
+      "loss": 0.8212,
+      "step": 11719
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619863797462534e-05,
+      "loss": 1.273,
+      "step": 11720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619796877799782e-05,
+      "loss": 1.0484,
+      "step": 11721
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6197299536297988e-05,
+      "loss": 0.9992,
+      "step": 11722
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6196630249530704e-05,
+      "loss": 1.1342,
+      "step": 11723
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6195960917700834e-05,
+      "loss": 0.7833,
+      "step": 11724
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619529154081325e-05,
+      "loss": 0.5085,
+      "step": 11725
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6194622118872817e-05,
+      "loss": 0.9447,
+      "step": 11726
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6193952651884406e-05,
+      "loss": 0.7957,
+      "step": 11727
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619328313985288e-05,
+      "loss": 1.2117,
+      "step": 11728
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6192613582783113e-05,
+      "loss": 1.0124,
+      "step": 11729
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6191943980679975e-05,
+      "loss": 0.9813,
+      "step": 11730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619127433354833e-05,
+      "loss": 0.9721,
+      "step": 11731
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619060464139305e-05,
+      "loss": 0.9375,
+      "step": 11732
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618993490421901e-05,
+      "loss": 0.9811,
+      "step": 11733
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6189265122031073e-05,
+      "loss": 0.9028,
+      "step": 11734
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618859529483411e-05,
+      "loss": 0.8675,
+      "step": 11735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187925422633e-05,
+      "loss": 0.9451,
+      "step": 11736
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187255505432603e-05,
+      "loss": 1.0318,
+      "step": 11737
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6186585543237802e-05,
+      "loss": 0.6229,
+      "step": 11738
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6185915536053462e-05,
+      "loss": 0.9795,
+      "step": 11739
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6185245483884456e-05,
+      "loss": 1.0154,
+      "step": 11740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6184575386735663e-05,
+      "loss": 1.2523,
+      "step": 11741
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6183905244611948e-05,
+      "loss": 0.9698,
+      "step": 11742
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6183235057518186e-05,
+      "loss": 1.1907,
+      "step": 11743
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6182564825459253e-05,
+      "loss": 1.0013,
+      "step": 11744
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6181894548440022e-05,
+      "loss": 1.0416,
+      "step": 11745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6181224226465367e-05,
+      "loss": 1.0801,
+      "step": 11746
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6180553859540165e-05,
+      "loss": 0.8644,
+      "step": 11747
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6179883447669284e-05,
+      "loss": 0.9297,
+      "step": 11748
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617921299085761e-05,
+      "loss": 1.1391,
+      "step": 11749
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617854248911001e-05,
+      "loss": 1.0355,
+      "step": 11750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6177871942431365e-05,
+      "loss": 0.8045,
+      "step": 11751
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6177201350826547e-05,
+      "loss": 1.6281,
+      "step": 11752
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6176530714300437e-05,
+      "loss": 0.9035,
+      "step": 11753
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6175860032857905e-05,
+      "loss": 1.0601,
+      "step": 11754
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6175189306503836e-05,
+      "loss": 0.9417,
+      "step": 11755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6174518535243106e-05,
+      "loss": 0.8036,
+      "step": 11756
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617384771908059e-05,
+      "loss": 0.8781,
+      "step": 11757
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173176858021167e-05,
+      "loss": 0.8491,
+      "step": 11758
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617250595206972e-05,
+      "loss": 1.3937,
+      "step": 11759
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617183500123112e-05,
+      "loss": 0.9844,
+      "step": 11760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6171164005510254e-05,
+      "loss": 0.7076,
+      "step": 11761
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6170492964912e-05,
+      "loss": 0.7989,
+      "step": 11762
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616982187944123e-05,
+      "loss": 1.0649,
+      "step": 11763
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616915074910283e-05,
+      "loss": 0.9557,
+      "step": 11764
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6168479573901687e-05,
+      "loss": 1.191,
+      "step": 11765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616780835384267e-05,
+      "loss": 0.9585,
+      "step": 11766
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6167137088930668e-05,
+      "loss": 0.7807,
+      "step": 11767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616646577917056e-05,
+      "loss": 1.1514,
+      "step": 11768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616579442456723e-05,
+      "loss": 0.87,
+      "step": 11769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6165123025125554e-05,
+      "loss": 1.0298,
+      "step": 11770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616445158085042e-05,
+      "loss": 1.1639,
+      "step": 11771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616378009174671e-05,
+      "loss": 0.9884,
+      "step": 11772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616310855781931e-05,
+      "loss": 0.9457,
+      "step": 11773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6162436979073096e-05,
+      "loss": 0.7195,
+      "step": 11774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6161765355512958e-05,
+      "loss": 1.0813,
+      "step": 11775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6161093687143777e-05,
+      "loss": 1.1366,
+      "step": 11776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616042197397044e-05,
+      "loss": 1.4818,
+      "step": 11777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6159750215997827e-05,
+      "loss": 0.9334,
+      "step": 11778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615907841323083e-05,
+      "loss": 1.232,
+      "step": 11779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615840656567433e-05,
+      "loss": 0.8933,
+      "step": 11780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615773467333321e-05,
+      "loss": 1.2689,
+      "step": 11781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6157062736212365e-05,
+      "loss": 1.3263,
+      "step": 11782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6156390754316673e-05,
+      "loss": 1.0258,
+      "step": 11783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6155718727651024e-05,
+      "loss": 1.1244,
+      "step": 11784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6155046656220307e-05,
+      "loss": 1.0436,
+      "step": 11785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6154374540029404e-05,
+      "loss": 0.8031,
+      "step": 11786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615370237908321e-05,
+      "loss": 1.0179,
+      "step": 11787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6153030173386602e-05,
+      "loss": 1.0974,
+      "step": 11788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615235792294448e-05,
+      "loss": 0.9085,
+      "step": 11789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615168562776173e-05,
+      "loss": 1.7963,
+      "step": 11790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615101328784324e-05,
+      "loss": 0.8415,
+      "step": 11791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615034090319389e-05,
+      "loss": 0.8216,
+      "step": 11792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6149668473818584e-05,
+      "loss": 1.1056,
+      "step": 11793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148995999722207e-05,
+      "loss": 0.6898,
+      "step": 11794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148323480909646e-05,
+      "loss": 0.7776,
+      "step": 11795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6147650917385794e-05,
+      "loss": 1.3062,
+      "step": 11796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146978309155543e-05,
+      "loss": 0.8468,
+      "step": 11797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614630565622378e-05,
+      "loss": 1.1558,
+      "step": 11798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61456329585954e-05,
+      "loss": 0.7491,
+      "step": 11799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6144960216275297e-05,
+      "loss": 0.9961,
+      "step": 11800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6144287429268364e-05,
+      "loss": 0.9819,
+      "step": 11801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6143614597579485e-05,
+      "loss": 0.7249,
+      "step": 11802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614294172121356e-05,
+      "loss": 0.9591,
+      "step": 11803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614226880017548e-05,
+      "loss": 1.1661,
+      "step": 11804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6141595834470142e-05,
+      "loss": 0.7284,
+      "step": 11805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6140922824102437e-05,
+      "loss": 1.1016,
+      "step": 11806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6140249769077255e-05,
+      "loss": 0.6121,
+      "step": 11807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6139576669399498e-05,
+      "loss": 1.6418,
+      "step": 11808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138903525074058e-05,
+      "loss": 0.9524,
+      "step": 11809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138230336105827e-05,
+      "loss": 0.9922,
+      "step": 11810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6137557102499706e-05,
+      "loss": 0.9083,
+      "step": 11811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6136883824260587e-05,
+      "loss": 0.7791,
+      "step": 11812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6136210501393368e-05,
+      "loss": 1.037,
+      "step": 11813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6135537133902945e-05,
+      "loss": 1.1979,
+      "step": 11814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134863721794212e-05,
+      "loss": 0.9505,
+      "step": 11815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134190265072074e-05,
+      "loss": 0.9167,
+      "step": 11816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613351676374142e-05,
+      "loss": 1.0632,
+      "step": 11817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613284321780715e-05,
+      "loss": 1.2417,
+      "step": 11818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6132169627274165e-05,
+      "loss": 1.1616,
+      "step": 11819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6131495992147363e-05,
+      "loss": 0.9632,
+      "step": 11820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6130822312431638e-05,
+      "loss": 1.2403,
+      "step": 11821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6130148588131894e-05,
+      "loss": 1.2053,
+      "step": 11822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6129474819253025e-05,
+      "loss": 0.9131,
+      "step": 11823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612880100579994e-05,
+      "loss": 0.8218,
+      "step": 11824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612812714777753e-05,
+      "loss": 2.0509,
+      "step": 11825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61274532451907e-05,
+      "loss": 1.0921,
+      "step": 11826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612677929804435e-05,
+      "loss": 0.9041,
+      "step": 11827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612610530634338e-05,
+      "loss": 0.8997,
+      "step": 11828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612543127009269e-05,
+      "loss": 1.0625,
+      "step": 11829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6124757189297187e-05,
+      "loss": 0.8912,
+      "step": 11830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612408306396177e-05,
+      "loss": 0.9863,
+      "step": 11831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6123408894091337e-05,
+      "loss": 0.9739,
+      "step": 11832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6122734679690798e-05,
+      "loss": 0.8396,
+      "step": 11833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612206042076505e-05,
+      "loss": 1.0786,
+      "step": 11834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6121386117319e-05,
+      "loss": 0.5631,
+      "step": 11835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6120711769357552e-05,
+      "loss": 1.0426,
+      "step": 11836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6120037376885605e-05,
+      "loss": 1.103,
+      "step": 11837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6119362939908066e-05,
+      "loss": 0.8587,
+      "step": 11838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6118688458429846e-05,
+      "loss": 0.9887,
+      "step": 11839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611801393245584e-05,
+      "loss": 1.0063,
+      "step": 11840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117339361990957e-05,
+      "loss": 0.764,
+      "step": 11841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6116664747040104e-05,
+      "loss": 1.0544,
+      "step": 11842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6115990087608183e-05,
+      "loss": 0.9526,
+      "step": 11843
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6115315383700107e-05,
+      "loss": 1.0111,
+      "step": 11844
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6114640635320775e-05,
+      "loss": 0.7692,
+      "step": 11845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61139658424751e-05,
+      "loss": 1.0582,
+      "step": 11846
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6113291005167984e-05,
+      "loss": 0.9372,
+      "step": 11847
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6112616123404338e-05,
+      "loss": 0.9528,
+      "step": 11848
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6111941197189073e-05,
+      "loss": 1.1306,
+      "step": 11849
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6111266226527086e-05,
+      "loss": 0.7796,
+      "step": 11850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110591211423296e-05,
+      "loss": 1.1059,
+      "step": 11851
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6109916151882606e-05,
+      "loss": 1.147,
+      "step": 11852
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6109241047909932e-05,
+      "loss": 0.7391,
+      "step": 11853
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6108565899510173e-05,
+      "loss": 0.8171,
+      "step": 11854
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6107890706688246e-05,
+      "loss": 0.82,
+      "step": 11855
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6107215469449063e-05,
+      "loss": 0.9943,
+      "step": 11856
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6106540187797526e-05,
+      "loss": 0.6205,
+      "step": 11857
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6105864861738553e-05,
+      "loss": 0.9566,
+      "step": 11858
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6105189491277052e-05,
+      "loss": 0.8291,
+      "step": 11859
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6104514076417937e-05,
+      "loss": 0.8702,
+      "step": 11860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6103838617166116e-05,
+      "loss": 0.9086,
+      "step": 11861
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6103163113526505e-05,
+      "loss": 1.1128,
+      "step": 11862
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6102487565504012e-05,
+      "loss": 1.4186,
+      "step": 11863
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610181197310355e-05,
+      "loss": 1.1542,
+      "step": 11864
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6101136336330037e-05,
+      "loss": 0.9992,
+      "step": 11865
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610046065518838e-05,
+      "loss": 1.8838,
+      "step": 11866
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60997849296835e-05,
+      "loss": 1.0685,
+      "step": 11867
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6099109159820302e-05,
+      "loss": 1.1729,
+      "step": 11868
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609843334560371e-05,
+      "loss": 0.9047,
+      "step": 11869
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609775748703863e-05,
+      "loss": 1.1335,
+      "step": 11870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609708158412998e-05,
+      "loss": 1.3857,
+      "step": 11871
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6096405636882675e-05,
+      "loss": 0.9352,
+      "step": 11872
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6095729645301634e-05,
+      "loss": 1.1652,
+      "step": 11873
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6095053609391768e-05,
+      "loss": 0.8312,
+      "step": 11874
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6094377529158e-05,
+      "loss": 0.9818,
+      "step": 11875
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6093701404605234e-05,
+      "loss": 1.1139,
+      "step": 11876
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60930252357384e-05,
+      "loss": 0.7312,
+      "step": 11877
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609234902256241e-05,
+      "loss": 0.941,
+      "step": 11878
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609167276508218e-05,
+      "loss": 0.86,
+      "step": 11879
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609099646330263e-05,
+      "loss": 1.1165,
+      "step": 11880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6090320117228674e-05,
+      "loss": 0.8371,
+      "step": 11881
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6089643726865237e-05,
+      "loss": 1.0216,
+      "step": 11882
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6088967292217233e-05,
+      "loss": 1.1296,
+      "step": 11883
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6088290813289583e-05,
+      "loss": 1.0592,
+      "step": 11884
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608761429008721e-05,
+      "loss": 1.2233,
+      "step": 11885
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086937722615027e-05,
+      "loss": 1.0375,
+      "step": 11886
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086261110877953e-05,
+      "loss": 1.1701,
+      "step": 11887
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6085584454880918e-05,
+      "loss": 0.9363,
+      "step": 11888
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6084907754628834e-05,
+      "loss": 0.9487,
+      "step": 11889
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6084231010126627e-05,
+      "loss": 1.7719,
+      "step": 11890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6083554221379217e-05,
+      "loss": 1.2364,
+      "step": 11891
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608287738839152e-05,
+      "loss": 0.8912,
+      "step": 11892
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6082200511168466e-05,
+      "loss": 0.923,
+      "step": 11893
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6081523589714975e-05,
+      "loss": 1.3661,
+      "step": 11894
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6080846624035972e-05,
+      "loss": 1.4198,
+      "step": 11895
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6080169614136374e-05,
+      "loss": 1.0431,
+      "step": 11896
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6079492560021107e-05,
+      "loss": 1.0496,
+      "step": 11897
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6078815461695095e-05,
+      "loss": 1.1546,
+      "step": 11898
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6078138319163263e-05,
+      "loss": 0.8934,
+      "step": 11899
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6077461132430535e-05,
+      "loss": 1.1985,
+      "step": 11900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6076783901501833e-05,
+      "loss": 1.0201,
+      "step": 11901
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6076106626382087e-05,
+      "loss": 0.6942,
+      "step": 11902
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6075429307076214e-05,
+      "loss": 0.8405,
+      "step": 11903
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6074751943589147e-05,
+      "loss": 0.6333,
+      "step": 11904
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6074074535925807e-05,
+      "loss": 0.9265,
+      "step": 11905
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6073397084091126e-05,
+      "loss": 1.0312,
+      "step": 11906
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6072719588090023e-05,
+      "loss": 0.7173,
+      "step": 11907
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607204204792743e-05,
+      "loss": 1.1755,
+      "step": 11908
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071364463608276e-05,
+      "loss": 1.1126,
+      "step": 11909
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070686835137484e-05,
+      "loss": 0.7558,
+      "step": 11910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6070009162519983e-05,
+      "loss": 0.9969,
+      "step": 11911
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60693314457607e-05,
+      "loss": 0.9228,
+      "step": 11912
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606865368486457e-05,
+      "loss": 0.9131,
+      "step": 11913
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606797587983651e-05,
+      "loss": 4.605,
+      "step": 11914
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6067298030681457e-05,
+      "loss": 1.6284,
+      "step": 11915
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6066620137404337e-05,
+      "loss": 0.9969,
+      "step": 11916
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6065942200010086e-05,
+      "loss": 0.9602,
+      "step": 11917
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606526421850363e-05,
+      "loss": 1.0677,
+      "step": 11918
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6064586192889894e-05,
+      "loss": 1.0933,
+      "step": 11919
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606390812317382e-05,
+      "loss": 0.782,
+      "step": 11920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606323000936033e-05,
+      "loss": 2.4976,
+      "step": 11921
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6062551851454358e-05,
+      "loss": 1.0035,
+      "step": 11922
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6061873649460834e-05,
+      "loss": 3.4131,
+      "step": 11923
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6061195403384694e-05,
+      "loss": 0.5237,
+      "step": 11924
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6060517113230866e-05,
+      "loss": 1.034,
+      "step": 11925
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6059838779004288e-05,
+      "loss": 1.0883,
+      "step": 11926
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6059160400709885e-05,
+      "loss": 1.1949,
+      "step": 11927
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60584819783526e-05,
+      "loss": 1.0814,
+      "step": 11928
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6057803511937354e-05,
+      "loss": 0.9408,
+      "step": 11929
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6057125001469095e-05,
+      "loss": 1.0588,
+      "step": 11930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6056446446952747e-05,
+      "loss": 1.088,
+      "step": 11931
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055767848393248e-05,
+      "loss": 1.1801,
+      "step": 11932
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055089205795533e-05,
+      "loss": 1.0112,
+      "step": 11933
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605441051916454e-05,
+      "loss": 0.8501,
+      "step": 11934
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053731788505197e-05,
+      "loss": 0.9714,
+      "step": 11935
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6053053013822445e-05,
+      "loss": 0.9381,
+      "step": 11936
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605237419512122e-05,
+      "loss": 0.9101,
+      "step": 11937
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6051695332406456e-05,
+      "loss": 1.0429,
+      "step": 11938
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6051016425683096e-05,
+      "loss": 1.3154,
+      "step": 11939
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605033747495607e-05,
+      "loss": 0.9332,
+      "step": 11940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604965848023032e-05,
+      "loss": 0.9746,
+      "step": 11941
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6048979441510778e-05,
+      "loss": 1.0097,
+      "step": 11942
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6048300358802388e-05,
+      "loss": 1.187,
+      "step": 11943
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6047621232110083e-05,
+      "loss": 0.9709,
+      "step": 11944
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6046942061438805e-05,
+      "loss": 1.0552,
+      "step": 11945
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6046262846793496e-05,
+      "loss": 1.0761,
+      "step": 11946
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6045583588179094e-05,
+      "loss": 0.9367,
+      "step": 11947
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6044904285600533e-05,
+      "loss": 1.0053,
+      "step": 11948
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6044224939062755e-05,
+      "loss": 0.7533,
+      "step": 11949
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6043545548570707e-05,
+      "loss": 1.0523,
+      "step": 11950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6042866114129323e-05,
+      "loss": 1.1706,
+      "step": 11951
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6042186635743545e-05,
+      "loss": 0.9696,
+      "step": 11952
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6041507113418316e-05,
+      "loss": 0.8208,
+      "step": 11953
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6040827547158575e-05,
+      "loss": 1.1337,
+      "step": 11954
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6040147936969263e-05,
+      "loss": 1.1877,
+      "step": 11955
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603946828285533e-05,
+      "loss": 0.8463,
+      "step": 11956
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603878858482171e-05,
+      "loss": 1.0647,
+      "step": 11957
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6038108842873348e-05,
+      "loss": 1.0942,
+      "step": 11958
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603742905701519e-05,
+      "loss": 0.9924,
+      "step": 11959
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6036749227252176e-05,
+      "loss": 1.0715,
+      "step": 11960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603606935358925e-05,
+      "loss": 1.119,
+      "step": 11961
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603538943603136e-05,
+      "loss": 1.0931,
+      "step": 11962
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603470947458345e-05,
+      "loss": 1.1008,
+      "step": 11963
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6034029469250458e-05,
+      "loss": 0.744,
+      "step": 11964
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6033349420037332e-05,
+      "loss": 1.0322,
+      "step": 11965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6032669326949024e-05,
+      "loss": 0.8095,
+      "step": 11966
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603198918999047e-05,
+      "loss": 1.0669,
+      "step": 11967
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6031309009166626e-05,
+      "loss": 1.0699,
+      "step": 11968
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6030628784482432e-05,
+      "loss": 0.8704,
+      "step": 11969
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602994851594283e-05,
+      "loss": 1.0135,
+      "step": 11970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6029268203552783e-05,
+      "loss": 0.9471,
+      "step": 11971
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6028587847317217e-05,
+      "loss": 0.9568,
+      "step": 11972
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60279074472411e-05,
+      "loss": 0.5886,
+      "step": 11973
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6027227003329366e-05,
+      "loss": 0.8846,
+      "step": 11974
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602654651558697e-05,
+      "loss": 0.9345,
+      "step": 11975
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6025865984018855e-05,
+      "loss": 1.1031,
+      "step": 11976
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6025185408629974e-05,
+      "loss": 1.0647,
+      "step": 11977
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6024504789425277e-05,
+      "loss": 0.9648,
+      "step": 11978
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6023824126409713e-05,
+      "loss": 1.1999,
+      "step": 11979
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602314341958823e-05,
+      "loss": 1.1478,
+      "step": 11980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6022462668965782e-05,
+      "loss": 0.9581,
+      "step": 11981
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6021781874547313e-05,
+      "loss": 0.8224,
+      "step": 11982
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602110103633778e-05,
+      "loss": 1.0827,
+      "step": 11983
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602042015434213e-05,
+      "loss": 0.8129,
+      "step": 11984
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6019739228565314e-05,
+      "loss": 1.0025,
+      "step": 11985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601905825901229e-05,
+      "loss": 1.1416,
+      "step": 11986
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6018377245688004e-05,
+      "loss": 0.8275,
+      "step": 11987
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6017696188597408e-05,
+      "loss": 1.2116,
+      "step": 11988
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601701508774546e-05,
+      "loss": 0.8242,
+      "step": 11989
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6016333943137114e-05,
+      "loss": 1.0941,
+      "step": 11990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015652754777312e-05,
+      "loss": 1.0106,
+      "step": 11991
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601497152267102e-05,
+      "loss": 0.9591,
+      "step": 11992
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6014290246823184e-05,
+      "loss": 1.2853,
+      "step": 11993
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6013608927238765e-05,
+      "loss": 1.2527,
+      "step": 11994
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6012927563922713e-05,
+      "loss": 0.701,
+      "step": 11995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6012246156879983e-05,
+      "loss": 1.1206,
+      "step": 11996
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601156470611553e-05,
+      "loss": 0.9755,
+      "step": 11997
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010883211634312e-05,
+      "loss": 0.9571,
+      "step": 11998
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010201673441284e-05,
+      "loss": 1.0434,
+      "step": 11999
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6009520091541403e-05,
+      "loss": 1.1234,
+      "step": 12000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6008838465939623e-05,
+      "loss": 0.8489,
+      "step": 12001
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6008156796640904e-05,
+      "loss": 1.1763,
+      "step": 12002
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60074750836502e-05,
+      "loss": 1.1285,
+      "step": 12003
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600679332697247e-05,
+      "loss": 0.8962,
+      "step": 12004
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600611152661267e-05,
+      "loss": 1.3313,
+      "step": 12005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6005429682575765e-05,
+      "loss": 0.9303,
+      "step": 12006
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6004747794866707e-05,
+      "loss": 0.8975,
+      "step": 12007
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6004065863490454e-05,
+      "loss": 0.7911,
+      "step": 12008
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6003383888451967e-05,
+      "loss": 1.0039,
+      "step": 12009
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600270186975621e-05,
+      "loss": 1.1377,
+      "step": 12010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6002019807408135e-05,
+      "loss": 1.4195,
+      "step": 12011
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6001337701412703e-05,
+      "loss": 0.9639,
+      "step": 12012
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6000655551774885e-05,
+      "loss": 1.0326,
+      "step": 12013
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5999973358499626e-05,
+      "loss": 0.899,
+      "step": 12014
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5999291121591894e-05,
+      "loss": 1.0506,
+      "step": 12015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5998608841056657e-05,
+      "loss": 0.8322,
+      "step": 12016
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5997926516898867e-05,
+      "loss": 1.1634,
+      "step": 12017
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599724414912349e-05,
+      "loss": 0.943,
+      "step": 12018
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5996561737735487e-05,
+      "loss": 0.8326,
+      "step": 12019
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5995879282739818e-05,
+      "loss": 0.9337,
+      "step": 12020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5995196784141457e-05,
+      "loss": 0.9697,
+      "step": 12021
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5994514241945355e-05,
+      "loss": 0.8712,
+      "step": 12022
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5993831656156476e-05,
+      "loss": 1.1323,
+      "step": 12023
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5993149026779792e-05,
+      "loss": 0.9062,
+      "step": 12024
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5992466353820258e-05,
+      "loss": 0.9366,
+      "step": 12025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5991783637282848e-05,
+      "loss": 0.9891,
+      "step": 12026
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5991100877172524e-05,
+      "loss": 0.9686,
+      "step": 12027
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5990418073494245e-05,
+      "loss": 1.339,
+      "step": 12028
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598973522625298e-05,
+      "loss": 0.8278,
+      "step": 12029
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989052335453695e-05,
+      "loss": 1.2676,
+      "step": 12030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598836940110136e-05,
+      "loss": 0.8007,
+      "step": 12031
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5987686423200935e-05,
+      "loss": 1.4235,
+      "step": 12032
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598700340175739e-05,
+      "loss": 1.1764,
+      "step": 12033
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598632033677569e-05,
+      "loss": 1.0158,
+      "step": 12034
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985637228260806e-05,
+      "loss": 0.996,
+      "step": 12035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5984954076217706e-05,
+      "loss": 0.666,
+      "step": 12036
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598427088065135e-05,
+      "loss": 1.9138,
+      "step": 12037
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5983587641566713e-05,
+      "loss": 1.1364,
+      "step": 12038
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598290435896876e-05,
+      "loss": 0.9371,
+      "step": 12039
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982221032862468e-05,
+      "loss": 0.8105,
+      "step": 12040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59815376632528e-05,
+      "loss": 0.9758,
+      "step": 12041
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5980854250144722e-05,
+      "loss": 1.1557,
+      "step": 12042
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5980170793543207e-05,
+      "loss": 1.3247,
+      "step": 12043
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597948729345323e-05,
+      "loss": 0.7993,
+      "step": 12044
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5978803749879754e-05,
+      "loss": 1.0917,
+      "step": 12045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5978120162827756e-05,
+      "loss": 0.9245,
+      "step": 12046
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59774365323022e-05,
+      "loss": 0.9996,
+      "step": 12047
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5976752858308065e-05,
+      "loss": 0.9045,
+      "step": 12048
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597606914085032e-05,
+      "loss": 1.083,
+      "step": 12049
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5975385379933933e-05,
+      "loss": 0.8706,
+      "step": 12050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5974701575563886e-05,
+      "loss": 0.9451,
+      "step": 12051
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597401772774514e-05,
+      "loss": 0.8866,
+      "step": 12052
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597333383648268e-05,
+      "loss": 0.8609,
+      "step": 12053
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5972649901781465e-05,
+      "loss": 1.1944,
+      "step": 12054
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597196592364648e-05,
+      "loss": 0.9094,
+      "step": 12055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5971281902082698e-05,
+      "loss": 0.6823,
+      "step": 12056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5970597837095087e-05,
+      "loss": 1.0744,
+      "step": 12057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596991372868863e-05,
+      "loss": 0.9541,
+      "step": 12058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5969229576868294e-05,
+      "loss": 1.0366,
+      "step": 12059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596854538163906e-05,
+      "loss": 1.0547,
+      "step": 12060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5967861143005897e-05,
+      "loss": 2.0645,
+      "step": 12061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596717686097379e-05,
+      "loss": 0.7818,
+      "step": 12062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596649253554771e-05,
+      "loss": 1.0953,
+      "step": 12063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596580816673263e-05,
+      "loss": 1.2417,
+      "step": 12064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596512375453353e-05,
+      "loss": 0.9712,
+      "step": 12065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964439298955392e-05,
+      "loss": 0.7214,
+      "step": 12066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596375480000319e-05,
+      "loss": 0.8606,
+      "step": 12067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5963070257681898e-05,
+      "loss": 0.9502,
+      "step": 12068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5962385671996493e-05,
+      "loss": 0.8129,
+      "step": 12069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5961701042951963e-05,
+      "loss": 1.2999,
+      "step": 12070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596101637055328e-05,
+      "loss": 1.0114,
+      "step": 12071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5960331654805422e-05,
+      "loss": 0.8509,
+      "step": 12072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5959646895713372e-05,
+      "loss": 1.8229,
+      "step": 12073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5958962093282106e-05,
+      "loss": 0.8606,
+      "step": 12074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595827724751661e-05,
+      "loss": 1.43,
+      "step": 12075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5957592358421855e-05,
+      "loss": 1.184,
+      "step": 12076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595690742600283e-05,
+      "loss": 0.9535,
+      "step": 12077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5956222450264512e-05,
+      "loss": 0.9184,
+      "step": 12078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5955537431211884e-05,
+      "loss": 1.0668,
+      "step": 12079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5954852368849923e-05,
+      "loss": 0.8022,
+      "step": 12080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595416726318362e-05,
+      "loss": 1.0013,
+      "step": 12081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5953482114217947e-05,
+      "loss": 1.183,
+      "step": 12082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5952796921957894e-05,
+      "loss": 0.9828,
+      "step": 12083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5952111686408435e-05,
+      "loss": 1.1184,
+      "step": 12084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5951426407574567e-05,
+      "loss": 0.7753,
+      "step": 12085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595074108546126e-05,
+      "loss": 1.7611,
+      "step": 12086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5950055720073505e-05,
+      "loss": 1.1057,
+      "step": 12087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5949370311416283e-05,
+      "loss": 1.1486,
+      "step": 12088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5948684859494575e-05,
+      "loss": 0.9608,
+      "step": 12089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5947999364313378e-05,
+      "loss": 1.1451,
+      "step": 12090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5947313825877667e-05,
+      "loss": 1.0183,
+      "step": 12091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5946628244192427e-05,
+      "loss": 0.9168,
+      "step": 12092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5945942619262648e-05,
+      "loss": 1.1576,
+      "step": 12093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594525695109331e-05,
+      "loss": 1.0719,
+      "step": 12094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594457123968941e-05,
+      "loss": 1.053,
+      "step": 12095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594388548505592e-05,
+      "loss": 1.2165,
+      "step": 12096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5943199687197844e-05,
+      "loss": 1.0947,
+      "step": 12097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942513846120155e-05,
+      "loss": 1.1266,
+      "step": 12098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5941827961827844e-05,
+      "loss": 1.4821,
+      "step": 12099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59411420343259e-05,
+      "loss": 0.8274,
+      "step": 12100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5940456063619314e-05,
+      "loss": 0.7398,
+      "step": 12101
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5939770049713072e-05,
+      "loss": 0.8692,
+      "step": 12102
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5939083992612163e-05,
+      "loss": 0.9747,
+      "step": 12103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593839789232157e-05,
+      "loss": 1.1361,
+      "step": 12104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5937711748846292e-05,
+      "loss": 0.8251,
+      "step": 12105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5937025562191317e-05,
+      "loss": 0.9242,
+      "step": 12106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5936339332361632e-05,
+      "loss": 0.9588,
+      "step": 12107
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5935653059362226e-05,
+      "loss": 1.0472,
+      "step": 12108
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593496674319809e-05,
+      "loss": 0.9867,
+      "step": 12109
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593428038387422e-05,
+      "loss": 0.8388,
+      "step": 12110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5933593981395605e-05,
+      "loss": 1.0671,
+      "step": 12111
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5932907535767235e-05,
+      "loss": 0.9671,
+      "step": 12112
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59322210469941e-05,
+      "loss": 0.7286,
+      "step": 12113
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59315345150812e-05,
+      "loss": 1.0135,
+      "step": 12114
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930847940033514e-05,
+      "loss": 0.9766,
+      "step": 12115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930161321856052e-05,
+      "loss": 0.7339,
+      "step": 12116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5929474660553795e-05,
+      "loss": 0.9072,
+      "step": 12117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592878795613174e-05,
+      "loss": 1.0538,
+      "step": 12118
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5928101208594877e-05,
+      "loss": 0.8901,
+      "step": 12119
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5927414417948205e-05,
+      "loss": 1.6118,
+      "step": 12120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592672758419672e-05,
+      "loss": 0.8342,
+      "step": 12121
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5926040707345412e-05,
+      "loss": 1.0515,
+      "step": 12122
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592535378739928e-05,
+      "loss": 1.1452,
+      "step": 12123
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5924666824363314e-05,
+      "loss": 1.0356,
+      "step": 12124
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5923979818242515e-05,
+      "loss": 0.9138,
+      "step": 12125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592329276904188e-05,
+      "loss": 1.0382,
+      "step": 12126
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5922605676766395e-05,
+      "loss": 0.7412,
+      "step": 12127
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5921918541421068e-05,
+      "loss": 1.0949,
+      "step": 12128
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592123136301089e-05,
+      "loss": 0.9606,
+      "step": 12129
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5920544141540864e-05,
+      "loss": 0.9198,
+      "step": 12130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919856877015982e-05,
+      "loss": 1.0236,
+      "step": 12131
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5919169569441243e-05,
+      "loss": 1.1597,
+      "step": 12132
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5918482218821642e-05,
+      "loss": 0.8371,
+      "step": 12133
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5917794825162186e-05,
+      "loss": 0.9807,
+      "step": 12134
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5917107388467866e-05,
+      "loss": 0.7422,
+      "step": 12135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5916419908743686e-05,
+      "loss": 1.059,
+      "step": 12136
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5915732385994644e-05,
+      "loss": 1.1543,
+      "step": 12137
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5915044820225736e-05,
+      "loss": 1.4112,
+      "step": 12138
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5914357211441965e-05,
+      "loss": 1.0103,
+      "step": 12139
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5913669559648334e-05,
+      "loss": 0.7812,
+      "step": 12140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591298186484984e-05,
+      "loss": 1.0064,
+      "step": 12141
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5912294127051487e-05,
+      "loss": 1.5578,
+      "step": 12142
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591160634625827e-05,
+      "loss": 1.0344,
+      "step": 12143
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59109185224752e-05,
+      "loss": 0.9935,
+      "step": 12144
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591023065570727e-05,
+      "loss": 0.9841,
+      "step": 12145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5909542745959493e-05,
+      "loss": 0.93,
+      "step": 12146
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5908854793236862e-05,
+      "loss": 0.4115,
+      "step": 12147
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590816679754438e-05,
+      "loss": 0.8827,
+      "step": 12148
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5907478758887053e-05,
+      "loss": 1.2208,
+      "step": 12149
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906790677269887e-05,
+      "loss": 0.9455,
+      "step": 12150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906102552697884e-05,
+      "loss": 1.3209,
+      "step": 12151
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5905414385176046e-05,
+      "loss": 1.2559,
+      "step": 12152
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5904726174709376e-05,
+      "loss": 1.0462,
+      "step": 12153
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5904037921302883e-05,
+      "loss": 0.7892,
+      "step": 12154
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5903349624961574e-05,
+      "loss": 1.0669,
+      "step": 12155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590266128569045e-05,
+      "loss": 0.8463,
+      "step": 12156
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5901972903494515e-05,
+      "loss": 0.8592,
+      "step": 12157
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590128447837878e-05,
+      "loss": 0.7222,
+      "step": 12158
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5900596010348246e-05,
+      "loss": 1.3331,
+      "step": 12159
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5899907499407925e-05,
+      "loss": 1.0195,
+      "step": 12160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5899218945562823e-05,
+      "loss": 1.0552,
+      "step": 12161
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589853034881794e-05,
+      "loss": 1.0921,
+      "step": 12162
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5897841709178295e-05,
+      "loss": 0.8624,
+      "step": 12163
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589715302664889e-05,
+      "loss": 0.8804,
+      "step": 12164
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589646430123473e-05,
+      "loss": 1.3114,
+      "step": 12165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895775532940828e-05,
+      "loss": 1.1681,
+      "step": 12166
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895086721772192e-05,
+      "loss": 1.0197,
+      "step": 12167
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589439786773383e-05,
+      "loss": 1.3015,
+      "step": 12168
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589370897083075e-05,
+      "loss": 0.9305,
+      "step": 12169
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589302003106797e-05,
+      "loss": 1.7317,
+      "step": 12170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589233104845049e-05,
+      "loss": 1.0278,
+      "step": 12171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5891642022983325e-05,
+      "loss": 0.9578,
+      "step": 12172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5890952954671485e-05,
+      "loss": 1.0457,
+      "step": 12173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589026384351998e-05,
+      "loss": 1.5225,
+      "step": 12174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588957468953382e-05,
+      "loss": 0.9398,
+      "step": 12175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5888885492718027e-05,
+      "loss": 1.332,
+      "step": 12176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58881962530776e-05,
+      "loss": 0.9543,
+      "step": 12177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5887506970617554e-05,
+      "loss": 0.5318,
+      "step": 12178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5886817645342905e-05,
+      "loss": 0.7875,
+      "step": 12179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5886128277258665e-05,
+      "loss": 0.8718,
+      "step": 12180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5885438866369845e-05,
+      "loss": 0.9549,
+      "step": 12181
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884749412681463e-05,
+      "loss": 0.9578,
+      "step": 12182
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884059916198524e-05,
+      "loss": 1.4547,
+      "step": 12183
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5883370376926052e-05,
+      "loss": 1.2703,
+      "step": 12184
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588268079486906e-05,
+      "loss": 1.051,
+      "step": 12185
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5881991170032557e-05,
+      "loss": 0.854,
+      "step": 12186
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588130150242156e-05,
+      "loss": 0.817,
+      "step": 12187
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880611792041087e-05,
+      "loss": 1.0069,
+      "step": 12188
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587992203889615e-05,
+      "loss": 0.7709,
+      "step": 12189
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587923224299177e-05,
+      "loss": 0.8307,
+      "step": 12190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5878542404332958e-05,
+      "loss": 0.8272,
+      "step": 12191
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5877852522924733e-05,
+      "loss": 1.4287,
+      "step": 12192
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5877162598772114e-05,
+      "loss": 0.8779,
+      "step": 12193
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587647263188011e-05,
+      "loss": 1.2603,
+      "step": 12194
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587578262225375e-05,
+      "loss": 1.125,
+      "step": 12195
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5875092569898046e-05,
+      "loss": 1.011,
+      "step": 12196
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5874402474818014e-05,
+      "loss": 1.6365,
+      "step": 12197
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5873712337018675e-05,
+      "loss": 0.8526,
+      "step": 12198
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587302215650505e-05,
+      "loss": 0.9469,
+      "step": 12199
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5872331933282154e-05,
+      "loss": 1.2178,
+      "step": 12200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587164166735501e-05,
+      "loss": 0.8763,
+      "step": 12201
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5870951358728635e-05,
+      "loss": 0.8374,
+      "step": 12202
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5870261007408048e-05,
+      "loss": 0.8399,
+      "step": 12203
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586957061339827e-05,
+      "loss": 0.6206,
+      "step": 12204
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5868880176704326e-05,
+      "loss": 0.932,
+      "step": 12205
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5868189697331232e-05,
+      "loss": 0.9079,
+      "step": 12206
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5867499175284013e-05,
+      "loss": 1.0333,
+      "step": 12207
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5866808610567685e-05,
+      "loss": 0.939,
+      "step": 12208
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5866118003187276e-05,
+      "loss": 1.0978,
+      "step": 12209
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5865427353147805e-05,
+      "loss": 0.7981,
+      "step": 12210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5864736660454295e-05,
+      "loss": 0.8084,
+      "step": 12211
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586404592511177e-05,
+      "loss": 1.1032,
+      "step": 12212
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5863355147125247e-05,
+      "loss": 0.8987,
+      "step": 12213
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5862664326499757e-05,
+      "loss": 1.1236,
+      "step": 12214
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586197346324032e-05,
+      "loss": 1.0017,
+      "step": 12215
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5861282557351966e-05,
+      "loss": 0.9907,
+      "step": 12216
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586059160883971e-05,
+      "loss": 1.468,
+      "step": 12217
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5859900617708582e-05,
+      "loss": 1.0178,
+      "step": 12218
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5859209583963603e-05,
+      "loss": 0.9103,
+      "step": 12219
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5858518507609806e-05,
+      "loss": 1.3975,
+      "step": 12220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585782738865221e-05,
+      "loss": 0.9321,
+      "step": 12221
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585713622709584e-05,
+      "loss": 0.9487,
+      "step": 12222
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585644502294573e-05,
+      "loss": 1.0006,
+      "step": 12223
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5855753776206898e-05,
+      "loss": 0.8574,
+      "step": 12224
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5855062486884377e-05,
+      "loss": 0.9786,
+      "step": 12225
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585437115498319e-05,
+      "loss": 1.1381,
+      "step": 12226
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5853679780508365e-05,
+      "loss": 0.5776,
+      "step": 12227
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585298836346493e-05,
+      "loss": 1.073,
+      "step": 12228
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5852296903857914e-05,
+      "loss": 0.9554,
+      "step": 12229
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5851605401692344e-05,
+      "loss": 0.9218,
+      "step": 12230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5850913856973255e-05,
+      "loss": 1.607,
+      "step": 12231
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5850222269705668e-05,
+      "loss": 1.0278,
+      "step": 12232
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584953063989461e-05,
+      "loss": 0.9772,
+      "step": 12233
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5848838967545123e-05,
+      "loss": 0.8378,
+      "step": 12234
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5848147252662224e-05,
+      "loss": 0.8136,
+      "step": 12235
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5847455495250952e-05,
+      "loss": 1.0311,
+      "step": 12236
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5846763695316334e-05,
+      "loss": 1.0726,
+      "step": 12237
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5846071852863398e-05,
+      "loss": 1.2511,
+      "step": 12238
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5845379967897185e-05,
+      "loss": 1.377,
+      "step": 12239
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5844688040422714e-05,
+      "loss": 1.0708,
+      "step": 12240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584399607044502e-05,
+      "loss": 0.8221,
+      "step": 12241
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5843304057969145e-05,
+      "loss": 0.9819,
+      "step": 12242
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5842612003000112e-05,
+      "loss": 1.1324,
+      "step": 12243
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5841919905542954e-05,
+      "loss": 1.0693,
+      "step": 12244
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5841227765602704e-05,
+      "loss": 0.8624,
+      "step": 12245
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58405355831844e-05,
+      "loss": 0.9103,
+      "step": 12246
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583984335829307e-05,
+      "loss": 0.9272,
+      "step": 12247
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5839151090933756e-05,
+      "loss": 0.8351,
+      "step": 12248
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5838458781111483e-05,
+      "loss": 1.1446,
+      "step": 12249
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5837766428831287e-05,
+      "loss": 1.3429,
+      "step": 12250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583707403409821e-05,
+      "loss": 1.0142,
+      "step": 12251
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5836381596917282e-05,
+      "loss": 1.0715,
+      "step": 12252
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5835689117293536e-05,
+      "loss": 1.1135,
+      "step": 12253
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583499659523201e-05,
+      "loss": 0.8787,
+      "step": 12254
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5834304030737744e-05,
+      "loss": 0.8824,
+      "step": 12255
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5833611423815768e-05,
+      "loss": 0.8564,
+      "step": 12256
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5832918774471126e-05,
+      "loss": 1.0268,
+      "step": 12257
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583222608270885e-05,
+      "loss": 0.8001,
+      "step": 12258
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5831533348533972e-05,
+      "loss": 1.241,
+      "step": 12259
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5830840571951544e-05,
+      "loss": 1.1281,
+      "step": 12260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5830147752966592e-05,
+      "loss": 0.5692,
+      "step": 12261
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5829454891584162e-05,
+      "loss": 1.207,
+      "step": 12262
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5828761987809284e-05,
+      "loss": 0.994,
+      "step": 12263
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5828069041647004e-05,
+      "loss": 0.7678,
+      "step": 12264
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5827376053102357e-05,
+      "loss": 1.2341,
+      "step": 12265
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5826683022180387e-05,
+      "loss": 1.098,
+      "step": 12266
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582598994888613e-05,
+      "loss": 1.045,
+      "step": 12267
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5825296833224624e-05,
+      "loss": 0.9012,
+      "step": 12268
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5824603675200915e-05,
+      "loss": 1.0831,
+      "step": 12269
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582391047482004e-05,
+      "loss": 0.8954,
+      "step": 12270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5823217232087046e-05,
+      "loss": 0.5985,
+      "step": 12271
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582252394700696e-05,
+      "loss": 0.756,
+      "step": 12272
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5821830619584845e-05,
+      "loss": 0.8634,
+      "step": 12273
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5821137249825725e-05,
+      "loss": 1.2077,
+      "step": 12274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5820443837734646e-05,
+      "loss": 1.1668,
+      "step": 12275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581975038331666e-05,
+      "loss": 1.2217,
+      "step": 12276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5819056886576798e-05,
+      "loss": 0.6531,
+      "step": 12277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5818363347520108e-05,
+      "loss": 1.0619,
+      "step": 12278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5817669766151636e-05,
+      "loss": 1.062,
+      "step": 12279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5816976142476423e-05,
+      "loss": 1.0005,
+      "step": 12280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5816282476499513e-05,
+      "loss": 1.0189,
+      "step": 12281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581558876822595e-05,
+      "loss": 0.8575,
+      "step": 12282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5814895017660782e-05,
+      "loss": 0.7723,
+      "step": 12283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5814201224809047e-05,
+      "loss": 0.9598,
+      "step": 12284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5813507389675796e-05,
+      "loss": 1.075,
+      "step": 12285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581281351226608e-05,
+      "loss": 1.1015,
+      "step": 12286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5812119592584934e-05,
+      "loss": 1.0233,
+      "step": 12287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581142563063741e-05,
+      "loss": 1.0278,
+      "step": 12288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581073162642855e-05,
+      "loss": 1.0616,
+      "step": 12289
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581003757996341e-05,
+      "loss": 1.4969,
+      "step": 12290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5809343491247026e-05,
+      "loss": 0.9635,
+      "step": 12291
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5808649360284455e-05,
+      "loss": 1.6209,
+      "step": 12292
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580795518708074e-05,
+      "loss": 0.8616,
+      "step": 12293
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5807260971640927e-05,
+      "loss": 1.2513,
+      "step": 12294
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5806566713970074e-05,
+      "loss": 1.0685,
+      "step": 12295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580587241407322e-05,
+      "loss": 1.171,
+      "step": 12296
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580517807195541e-05,
+      "loss": 1.0431,
+      "step": 12297
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5804483687621707e-05,
+      "loss": 0.8096,
+      "step": 12298
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5803789261077156e-05,
+      "loss": 1.3021,
+      "step": 12299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58030947923268e-05,
+      "loss": 0.9834,
+      "step": 12300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58024002813757e-05,
+      "loss": 1.2814,
+      "step": 12301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5801705728228894e-05,
+      "loss": 0.7297,
+      "step": 12302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5801011132891446e-05,
+      "loss": 0.9916,
+      "step": 12303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5800316495368397e-05,
+      "loss": 1.2125,
+      "step": 12304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5799621815664804e-05,
+      "loss": 1.0247,
+      "step": 12305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5798927093785717e-05,
+      "loss": 1.2125,
+      "step": 12306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5798232329736186e-05,
+      "loss": 0.8701,
+      "step": 12307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5797537523521266e-05,
+      "loss": 1.0469,
+      "step": 12308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5796842675146014e-05,
+      "loss": 1.1181,
+      "step": 12309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5796147784615474e-05,
+      "loss": 0.7837,
+      "step": 12310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5795452851934703e-05,
+      "loss": 0.7221,
+      "step": 12311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579475787710876e-05,
+      "loss": 0.8372,
+      "step": 12312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579406286014269e-05,
+      "loss": 0.8253,
+      "step": 12313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5793367801041552e-05,
+      "loss": 0.8946,
+      "step": 12314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.57926726998104e-05,
+      "loss": 1.3761,
+      "step": 12315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579197755645429e-05,
+      "loss": 0.913,
+      "step": 12316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5791282370978276e-05,
+      "loss": 0.8501,
+      "step": 12317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5790587143387413e-05,
+      "loss": 1.2968,
+      "step": 12318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5789891873686758e-05,
+      "loss": 1.0093,
+      "step": 12319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5789196561881362e-05,
+      "loss": 0.9446,
+      "step": 12320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5788501207976295e-05,
+      "loss": 0.8479,
+      "step": 12321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5787805811976597e-05,
+      "loss": 1.7241,
+      "step": 12322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5787110373887333e-05,
+      "loss": 1.2086,
+      "step": 12323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5786414893713563e-05,
+      "loss": 0.6757,
+      "step": 12324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5785719371460338e-05,
+      "loss": 0.8912,
+      "step": 12325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578502380713272e-05,
+      "loss": 0.9549,
+      "step": 12326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5784328200735767e-05,
+      "loss": 1.1068,
+      "step": 12327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5783632552274538e-05,
+      "loss": 1.0376,
+      "step": 12328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578293686175409e-05,
+      "loss": 0.984,
+      "step": 12329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5782241129179482e-05,
+      "loss": 0.8507,
+      "step": 12330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5781545354555775e-05,
+      "loss": 1.0275,
+      "step": 12331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5780849537888026e-05,
+      "loss": 1.1629,
+      "step": 12332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.57801536791813e-05,
+      "loss": 1.0962,
+      "step": 12333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5779457778440656e-05,
+      "loss": 0.9252,
+      "step": 12334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577876183567115e-05,
+      "loss": 0.8749,
+      "step": 12335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5778065850877848e-05,
+      "loss": 0.9145,
+      "step": 12336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577736982406581e-05,
+      "loss": 1.343,
+      "step": 12337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5776673755240096e-05,
+      "loss": 1.3239,
+      "step": 12338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577597764440577e-05,
+      "loss": 1.0178,
+      "step": 12339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577528149156789e-05,
+      "loss": 1.6485,
+      "step": 12340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5774585296731527e-05,
+      "loss": 0.9888,
+      "step": 12341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577388905990173e-05,
+      "loss": 1.2275,
+      "step": 12342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577319278108358e-05,
+      "loss": 1.0538,
+      "step": 12343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5772496460282126e-05,
+      "loss": 0.8583,
+      "step": 12344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5771800097502437e-05,
+      "loss": 0.8414,
+      "step": 12345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.577110369274958e-05,
+      "loss": 1.1997,
+      "step": 12346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.577040724602861e-05,
+      "loss": 0.6165,
+      "step": 12347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5769710757344603e-05,
+      "loss": 0.897,
+      "step": 12348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5769014226702615e-05,
+      "loss": 0.7163,
+      "step": 12349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5768317654107718e-05,
+      "loss": 0.722,
+      "step": 12350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5767621039564972e-05,
+      "loss": 1.2921,
+      "step": 12351
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5766924383079448e-05,
+      "loss": 1.0133,
+      "step": 12352
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576622768465621e-05,
+      "loss": 0.8484,
+      "step": 12353
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576553094430032e-05,
+      "loss": 1.0877,
+      "step": 12354
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5764834162016852e-05,
+      "loss": 0.9383,
+      "step": 12355
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5764137337810866e-05,
+      "loss": 0.7887,
+      "step": 12356
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576344047168744e-05,
+      "loss": 0.8134,
+      "step": 12357
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576274356365163e-05,
+      "loss": 0.9036,
+      "step": 12358
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5762046613708508e-05,
+      "loss": 0.9687,
+      "step": 12359
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5761349621863145e-05,
+      "loss": 1.1319,
+      "step": 12360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5760652588120605e-05,
+      "loss": 0.8402,
+      "step": 12361
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5759955512485963e-05,
+      "loss": 0.8263,
+      "step": 12362
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5759258394964282e-05,
+      "loss": 1.0197,
+      "step": 12363
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575856123556064e-05,
+      "loss": 1.0214,
+      "step": 12364
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5757864034280096e-05,
+      "loss": 0.9665,
+      "step": 12365
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575716679112773e-05,
+      "loss": 1.2717,
+      "step": 12366
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5756469506108602e-05,
+      "loss": 1.3178,
+      "step": 12367
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575577217922779e-05,
+      "loss": 0.9935,
+      "step": 12368
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5755074810490367e-05,
+      "loss": 0.8974,
+      "step": 12369
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5754377399901398e-05,
+      "loss": 0.8241,
+      "step": 12370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575367994746596e-05,
+      "loss": 0.9407,
+      "step": 12371
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575298245318912e-05,
+      "loss": 1.3556,
+      "step": 12372
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5752284917075953e-05,
+      "loss": 1.0675,
+      "step": 12373
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5751587339131532e-05,
+      "loss": 0.9626,
+      "step": 12374
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5750889719360927e-05,
+      "loss": 0.9131,
+      "step": 12375
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5750192057769214e-05,
+      "loss": 0.6584,
+      "step": 12376
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574949435436147e-05,
+      "loss": 0.9464,
+      "step": 12377
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574879660914276e-05,
+      "loss": 1.1501,
+      "step": 12378
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5748098822118165e-05,
+      "loss": 0.9828,
+      "step": 12379
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5747400993292756e-05,
+      "loss": 0.9181,
+      "step": 12380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574670312267161e-05,
+      "loss": 1.0954,
+      "step": 12381
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57460052102598e-05,
+      "loss": 1.0503,
+      "step": 12382
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5745307256062403e-05,
+      "loss": 0.9761,
+      "step": 12383
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5744609260084488e-05,
+      "loss": 1.0696,
+      "step": 12384
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5743911222331143e-05,
+      "loss": 1.081,
+      "step": 12385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5743213142807435e-05,
+      "loss": 0.9396,
+      "step": 12386
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5742515021518447e-05,
+      "loss": 1.15,
+      "step": 12387
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574181685846925e-05,
+      "loss": 1.1053,
+      "step": 12388
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574111865366492e-05,
+      "loss": 0.9837,
+      "step": 12389
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574042040711054e-05,
+      "loss": 1.0219,
+      "step": 12390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5739722118811185e-05,
+      "loss": 1.1703,
+      "step": 12391
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5739023788771933e-05,
+      "loss": 1.1647,
+      "step": 12392
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5738325416997864e-05,
+      "loss": 1.1695,
+      "step": 12393
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5737627003494055e-05,
+      "loss": 1.0514,
+      "step": 12394
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5736928548265584e-05,
+      "loss": 0.8529,
+      "step": 12395
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5736230051317534e-05,
+      "loss": 1.2166,
+      "step": 12396
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573553151265498e-05,
+      "loss": 1.0457,
+      "step": 12397
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5734832932283003e-05,
+      "loss": 0.6823,
+      "step": 12398
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573413431020669e-05,
+      "loss": 1.018,
+      "step": 12399
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5733435646431107e-05,
+      "loss": 0.7752,
+      "step": 12400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5732736940961347e-05,
+      "loss": 1.0203,
+      "step": 12401
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573203819380249e-05,
+      "loss": 1.0509,
+      "step": 12402
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573133940495961e-05,
+      "loss": 0.7339,
+      "step": 12403
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5730640574437796e-05,
+      "loss": 0.8947,
+      "step": 12404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572994170224213e-05,
+      "loss": 0.7767,
+      "step": 12405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572924278837769e-05,
+      "loss": 1.2889,
+      "step": 12406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5728543832849558e-05,
+      "loss": 1.0824,
+      "step": 12407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572784483566282e-05,
+      "loss": 0.9198,
+      "step": 12408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572714579682256e-05,
+      "loss": 0.7878,
+      "step": 12409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5726446716333854e-05,
+      "loss": 1.3213,
+      "step": 12410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5725747594201798e-05,
+      "loss": 0.8816,
+      "step": 12411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.572504843043147e-05,
+      "loss": 1.0228,
+      "step": 12412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5724349225027952e-05,
+      "loss": 0.8042,
+      "step": 12413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5723649977996327e-05,
+      "loss": 0.742,
+      "step": 12414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5722950689341686e-05,
+      "loss": 1.2019,
+      "step": 12415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5722251359069118e-05,
+      "loss": 0.807,
+      "step": 12416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5721551987183694e-05,
+      "loss": 0.8918,
+      "step": 12417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5720852573690515e-05,
+      "loss": 0.989,
+      "step": 12418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5720153118594656e-05,
+      "loss": 1.0154,
+      "step": 12419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571945362190121e-05,
+      "loss": 0.9296,
+      "step": 12420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5718754083615264e-05,
+      "loss": 0.9123,
+      "step": 12421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57180545037419e-05,
+      "loss": 0.7651,
+      "step": 12422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571735488228621e-05,
+      "loss": 1.1999,
+      "step": 12423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5716655219253284e-05,
+      "loss": 0.8189,
+      "step": 12424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5715955514648203e-05,
+      "loss": 0.8888,
+      "step": 12425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571525576847606e-05,
+      "loss": 0.7915,
+      "step": 12426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571455598074194e-05,
+      "loss": 0.8514,
+      "step": 12427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5713856151450936e-05,
+      "loss": 0.6533,
+      "step": 12428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5713156280608134e-05,
+      "loss": 1.1651,
+      "step": 12429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5712456368218628e-05,
+      "loss": 0.7787,
+      "step": 12430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5711756414287505e-05,
+      "loss": 1.1864,
+      "step": 12431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571105641881985e-05,
+      "loss": 0.982,
+      "step": 12432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5710356381820763e-05,
+      "loss": 0.9369,
+      "step": 12433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5709656303295332e-05,
+      "loss": 0.9707,
+      "step": 12434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5708956183248644e-05,
+      "loss": 0.5606,
+      "step": 12435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570825602168579e-05,
+      "loss": 0.8631,
+      "step": 12436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5707555818611867e-05,
+      "loss": 0.8968,
+      "step": 12437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5706855574031966e-05,
+      "loss": 0.8483,
+      "step": 12438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5706155287951174e-05,
+      "loss": 1.2728,
+      "step": 12439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570545496037459e-05,
+      "loss": 0.8143,
+      "step": 12440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5704754591307304e-05,
+      "loss": 0.8989,
+      "step": 12441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5704054180754407e-05,
+      "loss": 0.9528,
+      "step": 12442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5703353728720995e-05,
+      "loss": 1.0759,
+      "step": 12443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5702653235212162e-05,
+      "loss": 1.0682,
+      "step": 12444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5701952700232998e-05,
+      "loss": 0.9561,
+      "step": 12445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5701252123788607e-05,
+      "loss": 1.2455,
+      "step": 12446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570055150588407e-05,
+      "loss": 1.0425,
+      "step": 12447
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5699850846524496e-05,
+      "loss": 1.2742,
+      "step": 12448
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5699150145714973e-05,
+      "loss": 0.7713,
+      "step": 12449
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5698449403460593e-05,
+      "loss": 0.9017,
+      "step": 12450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569774861976646e-05,
+      "loss": 0.7742,
+      "step": 12451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5697047794637665e-05,
+      "loss": 1.0572,
+      "step": 12452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5696346928079306e-05,
+      "loss": 0.845,
+      "step": 12453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569564602009648e-05,
+      "loss": 0.9096,
+      "step": 12454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5694945070694282e-05,
+      "loss": 1.2638,
+      "step": 12455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5694244079877812e-05,
+      "loss": 0.8888,
+      "step": 12456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569354304765217e-05,
+      "loss": 0.9588,
+      "step": 12457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5692841974022446e-05,
+      "loss": 1.2693,
+      "step": 12458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569214085899374e-05,
+      "loss": 0.9341,
+      "step": 12459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569143970257116e-05,
+      "loss": 0.4618,
+      "step": 12460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56907385047598e-05,
+      "loss": 0.9075,
+      "step": 12461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5690037265564753e-05,
+      "loss": 1.1993,
+      "step": 12462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5689335984991123e-05,
+      "loss": 0.6557,
+      "step": 12463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5688634663044012e-05,
+      "loss": 1.0142,
+      "step": 12464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5687933299728517e-05,
+      "loss": 0.8299,
+      "step": 12465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568723189504974e-05,
+      "loss": 0.6132,
+      "step": 12466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5686530449012784e-05,
+      "loss": 0.9942,
+      "step": 12467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568582896162274e-05,
+      "loss": 1.1367,
+      "step": 12468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5685127432884723e-05,
+      "loss": 1.1616,
+      "step": 12469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5684425862803823e-05,
+      "loss": 0.8485,
+      "step": 12470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568372425138515e-05,
+      "loss": 1.5377,
+      "step": 12471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5683022598633802e-05,
+      "loss": 0.8552,
+      "step": 12472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5682320904554882e-05,
+      "loss": 0.8867,
+      "step": 12473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5681619169153494e-05,
+      "loss": 1.2237,
+      "step": 12474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568091739243474e-05,
+      "loss": 0.8996,
+      "step": 12475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5680215574403728e-05,
+      "loss": 1.0044,
+      "step": 12476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5679513715065553e-05,
+      "loss": 1.0939,
+      "step": 12477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5678811814425326e-05,
+      "loss": 1.0648,
+      "step": 12478
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5678109872488147e-05,
+      "loss": 0.8482,
+      "step": 12479
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567740788925912e-05,
+      "loss": 1.2615,
+      "step": 12480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5676705864743358e-05,
+      "loss": 0.9324,
+      "step": 12481
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5676003798945957e-05,
+      "loss": 0.7927,
+      "step": 12482
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5675301691872027e-05,
+      "loss": 0.9924,
+      "step": 12483
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5674599543526674e-05,
+      "loss": 1.1468,
+      "step": 12484
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5673897353915e-05,
+      "loss": 1.1703,
+      "step": 12485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567319512304212e-05,
+      "loss": 0.8937,
+      "step": 12486
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567249285091313e-05,
+      "loss": 0.8627,
+      "step": 12487
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5671790537533143e-05,
+      "loss": 0.8791,
+      "step": 12488
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5671088182907265e-05,
+      "loss": 1.0,
+      "step": 12489
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5670385787040606e-05,
+      "loss": 0.9321,
+      "step": 12490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5669683349938273e-05,
+      "loss": 0.7859,
+      "step": 12491
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5668980871605367e-05,
+      "loss": 0.9999,
+      "step": 12492
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5668278352047005e-05,
+      "loss": 1.3145,
+      "step": 12493
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5667575791268297e-05,
+      "loss": 0.7783,
+      "step": 12494
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5666873189274344e-05,
+      "loss": 0.4676,
+      "step": 12495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5666170546070266e-05,
+      "loss": 1.0823,
+      "step": 12496
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566546786166116e-05,
+      "loss": 0.9414,
+      "step": 12497
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5664765136052147e-05,
+      "loss": 1.0754,
+      "step": 12498
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566406236924833e-05,
+      "loss": 0.9428,
+      "step": 12499
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5663359561254823e-05,
+      "loss": 0.8549,
+      "step": 12500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5662656712076737e-05,
+      "loss": 0.9582,
+      "step": 12501
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5661953821719183e-05,
+      "loss": 1.2777,
+      "step": 12502
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5661250890187272e-05,
+      "loss": 0.9541,
+      "step": 12503
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5660547917486114e-05,
+      "loss": 1.0531,
+      "step": 12504
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659844903620826e-05,
+      "loss": 1.029,
+      "step": 12505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659141848596515e-05,
+      "loss": 1.084,
+      "step": 12506
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5658438752418296e-05,
+      "loss": 1.215,
+      "step": 12507
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5657735615091284e-05,
+      "loss": 0.6026,
+      "step": 12508
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565703243662059e-05,
+      "loss": 0.93,
+      "step": 12509
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5656329217011322e-05,
+      "loss": 0.7406,
+      "step": 12510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5655625956268608e-05,
+      "loss": 0.8486,
+      "step": 12511
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565492265439755e-05,
+      "loss": 0.811,
+      "step": 12512
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565421931140327e-05,
+      "loss": 0.9562,
+      "step": 12513
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5653515927290874e-05,
+      "loss": 0.7109,
+      "step": 12514
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652812502065488e-05,
+      "loss": 0.9734,
+      "step": 12515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652109035732217e-05,
+      "loss": 1.1981,
+      "step": 12516
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565140552829618e-05,
+      "loss": 0.8828,
+      "step": 12517
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56507019797625e-05,
+      "loss": 1.0093,
+      "step": 12518
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5649998390136286e-05,
+      "loss": 0.9339,
+      "step": 12519
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5649294759422655e-05,
+      "loss": 1.1356,
+      "step": 12520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5648591087626724e-05,
+      "loss": 1.1426,
+      "step": 12521
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5647887374753618e-05,
+      "loss": 1.0424,
+      "step": 12522
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5647183620808443e-05,
+      "loss": 0.8257,
+      "step": 12523
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564647982579632e-05,
+      "loss": 1.1003,
+      "step": 12524
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5645775989722366e-05,
+      "loss": 0.8454,
+      "step": 12525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564507211259171e-05,
+      "loss": 1.1954,
+      "step": 12526
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564436819440946e-05,
+      "loss": 0.9533,
+      "step": 12527
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5643664235180736e-05,
+      "loss": 0.649,
+      "step": 12528
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564296023491066e-05,
+      "loss": 1.1595,
+      "step": 12529
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564225619360435e-05,
+      "loss": 1.0035,
+      "step": 12530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5641552111266927e-05,
+      "loss": 0.9274,
+      "step": 12531
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5640847987903508e-05,
+      "loss": 1.204,
+      "step": 12532
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564014382351922e-05,
+      "loss": 0.7403,
+      "step": 12533
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5639439618119178e-05,
+      "loss": 0.9966,
+      "step": 12534
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5638735371708502e-05,
+      "loss": 0.7971,
+      "step": 12535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563803108429232e-05,
+      "loss": 0.7442,
+      "step": 12536
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563732675587575e-05,
+      "loss": 0.7416,
+      "step": 12537
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563662238646391e-05,
+      "loss": 1.0841,
+      "step": 12538
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563591797606193e-05,
+      "loss": 1.4253,
+      "step": 12539
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563521352467493e-05,
+      "loss": 0.9592,
+      "step": 12540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5634509032308027e-05,
+      "loss": 0.8914,
+      "step": 12541
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563380449896635e-05,
+      "loss": 0.9896,
+      "step": 12542
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563309992465502e-05,
+      "loss": 0.9501,
+      "step": 12543
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5632395309379166e-05,
+      "loss": 1.1338,
+      "step": 12544
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5631690653143905e-05,
+      "loss": 1.2918,
+      "step": 12545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5630985955954366e-05,
+      "loss": 1.0417,
+      "step": 12546
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5630281217815667e-05,
+      "loss": 1.1666,
+      "step": 12547
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5629576438732944e-05,
+      "loss": 0.9894,
+      "step": 12548
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5628871618711313e-05,
+      "loss": 0.8334,
+      "step": 12549
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56281667577559e-05,
+      "loss": 1.095,
+      "step": 12550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562746185587184e-05,
+      "loss": 1.1681,
+      "step": 12551
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5626756913064248e-05,
+      "loss": 1.4005,
+      "step": 12552
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5626051929338253e-05,
+      "loss": 0.9723,
+      "step": 12553
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5625346904698984e-05,
+      "loss": 1.0617,
+      "step": 12554
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562464183915157e-05,
+      "loss": 1.0373,
+      "step": 12555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5623936732701138e-05,
+      "loss": 0.7265,
+      "step": 12556
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5623231585352812e-05,
+      "loss": 1.0155,
+      "step": 12557
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562252639711172e-05,
+      "loss": 1.0457,
+      "step": 12558
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5621821167982992e-05,
+      "loss": 1.0905,
+      "step": 12559
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5621115897971753e-05,
+      "loss": 1.089,
+      "step": 12560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562041058708314e-05,
+      "loss": 1.2536,
+      "step": 12561
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5619705235322276e-05,
+      "loss": 0.9386,
+      "step": 12562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561899984269429e-05,
+      "loss": 1.1354,
+      "step": 12563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5618294409204313e-05,
+      "loss": 0.8174,
+      "step": 12564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561758893485748e-05,
+      "loss": 1.3571,
+      "step": 12565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561688341965891e-05,
+      "loss": 1.1566,
+      "step": 12566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5616177863613744e-05,
+      "loss": 1.0169,
+      "step": 12567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5615472266727107e-05,
+      "loss": 1.1818,
+      "step": 12568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561476662900413e-05,
+      "loss": 0.778,
+      "step": 12569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5614060950449948e-05,
+      "loss": 0.8354,
+      "step": 12570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5613355231069696e-05,
+      "loss": 0.9762,
+      "step": 12571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56126494708685e-05,
+      "loss": 0.9568,
+      "step": 12572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561194366985149e-05,
+      "loss": 0.8956,
+      "step": 12573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5611237828023805e-05,
+      "loss": 1.0682,
+      "step": 12574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5610531945390573e-05,
+      "loss": 1.047,
+      "step": 12575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5609826021956932e-05,
+      "loss": 0.8346,
+      "step": 12576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5609120057728013e-05,
+      "loss": 0.8958,
+      "step": 12577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560841405270895e-05,
+      "loss": 1.2027,
+      "step": 12578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5607708006904875e-05,
+      "loss": 1.0325,
+      "step": 12579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560700192032093e-05,
+      "loss": 0.6814,
+      "step": 12580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560629579296224e-05,
+      "loss": 0.8937,
+      "step": 12581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5605589624833944e-05,
+      "loss": 0.8811,
+      "step": 12582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5604883415941177e-05,
+      "loss": 0.8459,
+      "step": 12583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5604177166289077e-05,
+      "loss": 0.9342,
+      "step": 12584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560347087588278e-05,
+      "loss": 0.9986,
+      "step": 12585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560276454472742e-05,
+      "loss": 0.9339,
+      "step": 12586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5602058172828134e-05,
+      "loss": 0.9935,
+      "step": 12587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5601351760190056e-05,
+      "loss": 0.6423,
+      "step": 12588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560064530681833e-05,
+      "loss": 1.1001,
+      "step": 12589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5599938812718084e-05,
+      "loss": 1.0479,
+      "step": 12590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5599232277894467e-05,
+      "loss": 0.8285,
+      "step": 12591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559852570235261e-05,
+      "loss": 1.1024,
+      "step": 12592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559781908609765e-05,
+      "loss": 0.7172,
+      "step": 12593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5597112429134728e-05,
+      "loss": 1.0487,
+      "step": 12594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5596405731468985e-05,
+      "loss": 0.9142,
+      "step": 12595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5595698993105556e-05,
+      "loss": 1.0159,
+      "step": 12596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5594992214049583e-05,
+      "loss": 0.8911,
+      "step": 12597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5594285394306207e-05,
+      "loss": 0.903,
+      "step": 12598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5593578533880566e-05,
+      "loss": 1.0633,
+      "step": 12599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5592871632777798e-05,
+      "loss": 0.7453,
+      "step": 12600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5592164691003052e-05,
+      "loss": 0.9929,
+      "step": 12601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559145770856146e-05,
+      "loss": 0.9575,
+      "step": 12602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5590750685458163e-05,
+      "loss": 1.0779,
+      "step": 12603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559004362169831e-05,
+      "loss": 1.0003,
+      "step": 12604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5589336517287038e-05,
+      "loss": 1.1745,
+      "step": 12605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558862937222949e-05,
+      "loss": 0.7576,
+      "step": 12606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558792218653081e-05,
+      "loss": 1.1405,
+      "step": 12607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558721496019614e-05,
+      "loss": 0.9548,
+      "step": 12608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586507693230622e-05,
+      "loss": 0.8762,
+      "step": 12609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55858003856394e-05,
+      "loss": 0.8656,
+      "step": 12610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5585093037427613e-05,
+      "loss": 1.2733,
+      "step": 12611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5584385648600414e-05,
+      "loss": 1.089,
+      "step": 12612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558367821916294e-05,
+      "loss": 1.2152,
+      "step": 12613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5582970749120337e-05,
+      "loss": 1.1365,
+      "step": 12614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5582263238477753e-05,
+      "loss": 0.9023,
+      "step": 12615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558155568724033e-05,
+      "loss": 1.4109,
+      "step": 12616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5580848095413216e-05,
+      "loss": 1.165,
+      "step": 12617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5580140463001554e-05,
+      "loss": 0.9481,
+      "step": 12618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557943279001049e-05,
+      "loss": 0.7323,
+      "step": 12619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557872507644517e-05,
+      "loss": 1.0885,
+      "step": 12620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5578017322310746e-05,
+      "loss": 1.0059,
+      "step": 12621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5577309527612356e-05,
+      "loss": 1.0811,
+      "step": 12622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5576601692355155e-05,
+      "loss": 0.9802,
+      "step": 12623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5575893816544283e-05,
+      "loss": 1.116,
+      "step": 12624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5575185900184893e-05,
+      "loss": 1.064,
+      "step": 12625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5574477943282136e-05,
+      "loss": 0.9063,
+      "step": 12626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5573769945841155e-05,
+      "loss": 1.0819,
+      "step": 12627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55730619078671e-05,
+      "loss": 0.8259,
+      "step": 12628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557235382936512e-05,
+      "loss": 0.9435,
+      "step": 12629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557164571034036e-05,
+      "loss": 0.9938,
+      "step": 12630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5570937550797978e-05,
+      "loss": 0.7427,
+      "step": 12631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5570229350743118e-05,
+      "loss": 1.1063,
+      "step": 12632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5569521110180932e-05,
+      "loss": 0.9764,
+      "step": 12633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556881282911657e-05,
+      "loss": 1.1163,
+      "step": 12634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5568104507555185e-05,
+      "loss": 0.9995,
+      "step": 12635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5567396145501922e-05,
+      "loss": 0.9363,
+      "step": 12636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5566687742961938e-05,
+      "loss": 1.3687,
+      "step": 12637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5565979299940383e-05,
+      "loss": 0.9498,
+      "step": 12638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5565270816442403e-05,
+      "loss": 0.9533,
+      "step": 12639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556456229247316e-05,
+      "loss": 0.9641,
+      "step": 12640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5563853728037805e-05,
+      "loss": 0.7202,
+      "step": 12641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5563145123141484e-05,
+      "loss": 0.8478,
+      "step": 12642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5562436477789355e-05,
+      "loss": 1.0668,
+      "step": 12643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556172779198657e-05,
+      "loss": 0.8597,
+      "step": 12644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5561019065738282e-05,
+      "loss": 0.9072,
+      "step": 12645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556031029904965e-05,
+      "loss": 1.153,
+      "step": 12646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5559601491925818e-05,
+      "loss": 0.8716,
+      "step": 12647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555889264437195e-05,
+      "loss": 0.9207,
+      "step": 12648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5558183756393198e-05,
+      "loss": 0.8554,
+      "step": 12649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5557474827994713e-05,
+      "loss": 1.2871,
+      "step": 12650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5556765859181657e-05,
+      "loss": 0.8562,
+      "step": 12651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5556056849959183e-05,
+      "loss": 1.2953,
+      "step": 12652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5555347800332445e-05,
+      "loss": 0.8154,
+      "step": 12653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55546387103066e-05,
+      "loss": 0.9463,
+      "step": 12654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555392957988681e-05,
+      "loss": 0.9866,
+      "step": 12655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555322040907822e-05,
+      "loss": 0.9403,
+      "step": 12656
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5552511197886e-05,
+      "loss": 0.9375,
+      "step": 12657
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55518019463153e-05,
+      "loss": 0.9226,
+      "step": 12658
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555109265437128e-05,
+      "loss": 0.4662,
+      "step": 12659
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55503833220591e-05,
+      "loss": 0.6708,
+      "step": 12660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5549673949383913e-05,
+      "loss": 0.8115,
+      "step": 12661
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5548964536350882e-05,
+      "loss": 0.6342,
+      "step": 12662
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5548255082965165e-05,
+      "loss": 0.9523,
+      "step": 12663
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554754558923192e-05,
+      "loss": 0.92,
+      "step": 12664
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554683605515631e-05,
+      "loss": 0.7385,
+      "step": 12665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554612648074349e-05,
+      "loss": 0.9748,
+      "step": 12666
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554541686599862e-05,
+      "loss": 0.9324,
+      "step": 12667
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5544707210926867e-05,
+      "loss": 0.9314,
+      "step": 12668
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5543997515533384e-05,
+      "loss": 0.9703,
+      "step": 12669
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554328777982334e-05,
+      "loss": 1.0092,
+      "step": 12670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554257800380189e-05,
+      "loss": 1.2331,
+      "step": 12671
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5541868187474195e-05,
+      "loss": 0.9029,
+      "step": 12672
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554115833084542e-05,
+      "loss": 1.1424,
+      "step": 12673
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5540448433920724e-05,
+      "loss": 1.1023,
+      "step": 12674
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539738496705277e-05,
+      "loss": 0.801,
+      "step": 12675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539028519204232e-05,
+      "loss": 1.2245,
+      "step": 12676
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553831850142276e-05,
+      "loss": 0.6804,
+      "step": 12677
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553760844336602e-05,
+      "loss": 0.7683,
+      "step": 12678
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5536898345039174e-05,
+      "loss": 0.9024,
+      "step": 12679
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553618820644739e-05,
+      "loss": 0.6991,
+      "step": 12680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5535478027595832e-05,
+      "loss": 0.8341,
+      "step": 12681
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553476780848966e-05,
+      "loss": 1.3424,
+      "step": 12682
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5534057549134045e-05,
+      "loss": 0.777,
+      "step": 12683
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553334724953415e-05,
+      "loss": 1.1869,
+      "step": 12684
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5532636909695136e-05,
+      "loss": 0.9582,
+      "step": 12685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553192652962217e-05,
+      "loss": 0.8375,
+      "step": 12686
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5531216109320425e-05,
+      "loss": 0.8412,
+      "step": 12687
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5530505648795057e-05,
+      "loss": 1.0908,
+      "step": 12688
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552979514805124e-05,
+      "loss": 1.1082,
+      "step": 12689
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5529084607094144e-05,
+      "loss": 0.7447,
+      "step": 12690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5528374025928926e-05,
+      "loss": 1.1719,
+      "step": 12691
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5527663404560756e-05,
+      "loss": 0.8365,
+      "step": 12692
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5526952742994805e-05,
+      "loss": 1.123,
+      "step": 12693
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5526242041236243e-05,
+      "loss": 1.0508,
+      "step": 12694
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5525531299290232e-05,
+      "loss": 0.6329,
+      "step": 12695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5524820517161948e-05,
+      "loss": 0.8859,
+      "step": 12696
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552410969485655e-05,
+      "loss": 0.7924,
+      "step": 12697
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552339883237922e-05,
+      "loss": 0.9853,
+      "step": 12698
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552268792973511e-05,
+      "loss": 0.935,
+      "step": 12699
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552197698692941e-05,
+      "loss": 1.1537,
+      "step": 12700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5521266003967276e-05,
+      "loss": 0.9248,
+      "step": 12701
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5520554980853883e-05,
+      "loss": 0.9741,
+      "step": 12702
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55198439175944e-05,
+      "loss": 1.0667,
+      "step": 12703
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5519132814194e-05,
+      "loss": 0.774,
+      "step": 12704
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5518421670657856e-05,
+      "loss": 1.158,
+      "step": 12705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5517710486991133e-05,
+      "loss": 0.9658,
+      "step": 12706
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5516999263199006e-05,
+      "loss": 1.1566,
+      "step": 12707
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551628799928665e-05,
+      "loss": 0.7545,
+      "step": 12708
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5515576695259234e-05,
+      "loss": 0.979,
+      "step": 12709
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5514865351121933e-05,
+      "loss": 1.0298,
+      "step": 12710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551415396687992e-05,
+      "loss": 1.1191,
+      "step": 12711
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5513442542538363e-05,
+      "loss": 0.7843,
+      "step": 12712
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551273107810244e-05,
+      "loss": 0.8553,
+      "step": 12713
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5512019573577327e-05,
+      "loss": 0.773,
+      "step": 12714
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5511308028968195e-05,
+      "loss": 0.9244,
+      "step": 12715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.551059644428022e-05,
+      "loss": 0.9716,
+      "step": 12716
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5509884819518574e-05,
+      "loss": 0.9033,
+      "step": 12717
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5509173154688433e-05,
+      "loss": 0.9008,
+      "step": 12718
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5508461449794975e-05,
+      "loss": 0.774,
+      "step": 12719
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550774970484337e-05,
+      "loss": 1.2045,
+      "step": 12720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5507037919838804e-05,
+      "loss": 0.7131,
+      "step": 12721
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550632609478644e-05,
+      "loss": 1.3284,
+      "step": 12722
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5505614229691467e-05,
+      "loss": 0.8721,
+      "step": 12723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550490232455905e-05,
+      "loss": 1.1859,
+      "step": 12724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5504190379394377e-05,
+      "loss": 0.8233,
+      "step": 12725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5503478394202616e-05,
+      "loss": 0.8472,
+      "step": 12726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550276636898895e-05,
+      "loss": 0.9134,
+      "step": 12727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550205430375856e-05,
+      "loss": 1.0869,
+      "step": 12728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5501342198516617e-05,
+      "loss": 1.1252,
+      "step": 12729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5500630053268304e-05,
+      "loss": 1.0043,
+      "step": 12730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5499917868018795e-05,
+      "loss": 0.9203,
+      "step": 12731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549920564277328e-05,
+      "loss": 0.8604,
+      "step": 12732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549849337753692e-05,
+      "loss": 1.0108,
+      "step": 12733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5497781072314913e-05,
+      "loss": 1.2338,
+      "step": 12734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5497068727112435e-05,
+      "loss": 0.9614,
+      "step": 12735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5496356341934654e-05,
+      "loss": 1.0224,
+      "step": 12736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5495643916786763e-05,
+      "loss": 1.1661,
+      "step": 12737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549493145167394e-05,
+      "loss": 0.7686,
+      "step": 12738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5494218946601366e-05,
+      "loss": 1.0681,
+      "step": 12739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549350640157422e-05,
+      "loss": 1.003,
+      "step": 12740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5492793816597687e-05,
+      "loss": 0.8748,
+      "step": 12741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549208119167694e-05,
+      "loss": 1.1221,
+      "step": 12742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5491368526817178e-05,
+      "loss": 1.6627,
+      "step": 12743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5490655822023572e-05,
+      "loss": 1.2826,
+      "step": 12744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5489943077301307e-05,
+      "loss": 1.2181,
+      "step": 12745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5489230292655565e-05,
+      "loss": 0.9936,
+      "step": 12746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548851746809153e-05,
+      "loss": 1.0982,
+      "step": 12747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5487804603614386e-05,
+      "loss": 1.1028,
+      "step": 12748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5487091699229318e-05,
+      "loss": 0.9022,
+      "step": 12749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5486378754941514e-05,
+      "loss": 1.1481,
+      "step": 12750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548566577075615e-05,
+      "loss": 0.9481,
+      "step": 12751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548495274667842e-05,
+      "loss": 1.0495,
+      "step": 12752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54842396827135e-05,
+      "loss": 1.0479,
+      "step": 12753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548352657886658e-05,
+      "loss": 1.3536,
+      "step": 12754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548281343514285e-05,
+      "loss": 0.7956,
+      "step": 12755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5482100251547486e-05,
+      "loss": 1.0213,
+      "step": 12756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5481387028085683e-05,
+      "loss": 1.1479,
+      "step": 12757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5480673764762624e-05,
+      "loss": 0.7693,
+      "step": 12758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54799604615835e-05,
+      "loss": 0.8986,
+      "step": 12759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479247118553494e-05,
+      "loss": 0.9464,
+      "step": 12760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5478533735677793e-05,
+      "loss": 0.7512,
+      "step": 12761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5477820312961585e-05,
+      "loss": 1.3202,
+      "step": 12762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5477106850410062e-05,
+      "loss": 1.5375,
+      "step": 12763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547639334802841e-05,
+      "loss": 1.0303,
+      "step": 12764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5475679805821814e-05,
+      "loss": 0.853,
+      "step": 12765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547496622379547e-05,
+      "loss": 0.7184,
+      "step": 12766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547425260195456e-05,
+      "loss": 1.1196,
+      "step": 12767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547353894030428e-05,
+      "loss": 0.9156,
+      "step": 12768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5472825238849817e-05,
+      "loss": 0.7754,
+      "step": 12769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5472111497596363e-05,
+      "loss": 0.9906,
+      "step": 12770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54713977165491e-05,
+      "loss": 0.9253,
+      "step": 12771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547068389571323e-05,
+      "loss": 0.9299,
+      "step": 12772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546997003509394e-05,
+      "loss": 1.2401,
+      "step": 12773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5469256134696417e-05,
+      "loss": 0.9062,
+      "step": 12774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546854219452586e-05,
+      "loss": 1.1633,
+      "step": 12775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5467828214587453e-05,
+      "loss": 1.1862,
+      "step": 12776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5467114194886396e-05,
+      "loss": 0.9113,
+      "step": 12777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5466400135427874e-05,
+      "loss": 0.7716,
+      "step": 12778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5465686036217084e-05,
+      "loss": 0.9868,
+      "step": 12779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546497189725922e-05,
+      "loss": 1.6766,
+      "step": 12780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546425771855947e-05,
+      "loss": 1.0437,
+      "step": 12781
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5463543500123034e-05,
+      "loss": 1.188,
+      "step": 12782
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5462829241955103e-05,
+      "loss": 0.6991,
+      "step": 12783
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546211494406087e-05,
+      "loss": 1.2333,
+      "step": 12784
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5461400606445535e-05,
+      "loss": 0.9919,
+      "step": 12785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546068622911428e-05,
+      "loss": 0.7263,
+      "step": 12786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5459971812072315e-05,
+      "loss": 1.2327,
+      "step": 12787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5459257355324824e-05,
+      "loss": 0.9306,
+      "step": 12788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5458542858877012e-05,
+      "loss": 0.996,
+      "step": 12789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5457828322734067e-05,
+      "loss": 1.0904,
+      "step": 12790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5457113746901188e-05,
+      "loss": 0.7669,
+      "step": 12791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5456399131383572e-05,
+      "loss": 0.902,
+      "step": 12792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5455684476186417e-05,
+      "loss": 1.1339,
+      "step": 12793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545496978131492e-05,
+      "loss": 1.1379,
+      "step": 12794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5454255046774273e-05,
+      "loss": 1.0374,
+      "step": 12795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545354027256968e-05,
+      "loss": 0.7029,
+      "step": 12796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5452825458706336e-05,
+      "loss": 0.6925,
+      "step": 12797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5452110605189443e-05,
+      "loss": 0.7824,
+      "step": 12798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5451395712024192e-05,
+      "loss": 0.8564,
+      "step": 12799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545068077921579e-05,
+      "loss": 0.9796,
+      "step": 12800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5449965806769426e-05,
+      "loss": 1.0342,
+      "step": 12801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544925079469031e-05,
+      "loss": 0.793,
+      "step": 12802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5448535742983635e-05,
+      "loss": 0.9608,
+      "step": 12803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5447820651654603e-05,
+      "loss": 1.5464,
+      "step": 12804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544710552070842e-05,
+      "loss": 0.9008,
+      "step": 12805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5446390350150272e-05,
+      "loss": 1.0013,
+      "step": 12806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5445675139985373e-05,
+      "loss": 1.0386,
+      "step": 12807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544495989021892e-05,
+      "loss": 1.0337,
+      "step": 12808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5444244600856113e-05,
+      "loss": 0.9476,
+      "step": 12809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5443529271902155e-05,
+      "loss": 0.9403,
+      "step": 12810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5442813903362247e-05,
+      "loss": 0.8982,
+      "step": 12811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544209849524159e-05,
+      "loss": 1.0176,
+      "step": 12812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5441383047545394e-05,
+      "loss": 0.9646,
+      "step": 12813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5440667560278853e-05,
+      "loss": 0.9737,
+      "step": 12814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439952033447173e-05,
+      "loss": 1.096,
+      "step": 12815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439236467055557e-05,
+      "loss": 0.7382,
+      "step": 12816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543852086110921e-05,
+      "loss": 0.9863,
+      "step": 12817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5437805215613338e-05,
+      "loss": 1.0217,
+      "step": 12818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5437089530573142e-05,
+      "loss": 0.8198,
+      "step": 12819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5436373805993824e-05,
+      "loss": 1.2445,
+      "step": 12820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5435658041880596e-05,
+      "loss": 1.0405,
+      "step": 12821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434942238238654e-05,
+      "loss": 1.1014,
+      "step": 12822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434226395073212e-05,
+      "loss": 1.022,
+      "step": 12823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5433510512389472e-05,
+      "loss": 1.657,
+      "step": 12824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543279459019264e-05,
+      "loss": 1.1393,
+      "step": 12825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5432078628487924e-05,
+      "loss": 1.0245,
+      "step": 12826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5431362627280524e-05,
+      "loss": 1.0431,
+      "step": 12827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5430646586575656e-05,
+      "loss": 0.9959,
+      "step": 12828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542993050637852e-05,
+      "loss": 0.912,
+      "step": 12829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542921438669433e-05,
+      "loss": 0.7789,
+      "step": 12830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542849822752829e-05,
+      "loss": 1.018,
+      "step": 12831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5427782028885605e-05,
+      "loss": 0.8711,
+      "step": 12832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542706579077149e-05,
+      "loss": 0.8983,
+      "step": 12833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5426349513191143e-05,
+      "loss": 0.7392,
+      "step": 12834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542563319614979e-05,
+      "loss": 0.9788,
+      "step": 12835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542491683965262e-05,
+      "loss": 0.833,
+      "step": 12836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5424200443704857e-05,
+      "loss": 0.7208,
+      "step": 12837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5423484008311706e-05,
+      "loss": 1.026,
+      "step": 12838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5422767533478376e-05,
+      "loss": 0.7351,
+      "step": 12839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5422051019210082e-05,
+      "loss": 0.8572,
+      "step": 12840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5421334465512026e-05,
+      "loss": 0.8105,
+      "step": 12841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5420617872389427e-05,
+      "loss": 1.0447,
+      "step": 12842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541990123984749e-05,
+      "loss": 0.9585,
+      "step": 12843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541918456789143e-05,
+      "loss": 0.8052,
+      "step": 12844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541846785652646e-05,
+      "loss": 1.0351,
+      "step": 12845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541775110575779e-05,
+      "loss": 0.9509,
+      "step": 12846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541703431559063e-05,
+      "loss": 0.9902,
+      "step": 12847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5416317486030195e-05,
+      "loss": 0.8681,
+      "step": 12848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5415600617081698e-05,
+      "loss": 0.7578,
+      "step": 12849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5414883708750354e-05,
+      "loss": 0.8325,
+      "step": 12850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5414166761041376e-05,
+      "loss": 1.0622,
+      "step": 12851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5413449773959972e-05,
+      "loss": 0.8695,
+      "step": 12852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541273274751136e-05,
+      "loss": 0.8762,
+      "step": 12853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5412015681700758e-05,
+      "loss": 0.9415,
+      "step": 12854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5411298576533376e-05,
+      "loss": 1.0991,
+      "step": 12855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5410581432014426e-05,
+      "loss": 0.6461,
+      "step": 12856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5409864248149128e-05,
+      "loss": 1.1396,
+      "step": 12857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5409147024942698e-05,
+      "loss": 1.0079,
+      "step": 12858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408429762400352e-05,
+      "loss": 0.7137,
+      "step": 12859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5407712460527305e-05,
+      "loss": 1.175,
+      "step": 12860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5406995119328768e-05,
+      "loss": 0.816,
+      "step": 12861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5406277738809966e-05,
+      "loss": 0.7326,
+      "step": 12862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540556031897611e-05,
+      "loss": 0.9064,
+      "step": 12863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5404842859832415e-05,
+      "loss": 0.8948,
+      "step": 12864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540412536138411e-05,
+      "loss": 0.816,
+      "step": 12865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54034078236364e-05,
+      "loss": 1.1229,
+      "step": 12866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540269024659451e-05,
+      "loss": 1.0705,
+      "step": 12867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5401972630263656e-05,
+      "loss": 1.0156,
+      "step": 12868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5401254974649057e-05,
+      "loss": 0.9473,
+      "step": 12869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5400537279755935e-05,
+      "loss": 0.9428,
+      "step": 12870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5399819545589503e-05,
+      "loss": 1.1175,
+      "step": 12871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5399101772154986e-05,
+      "loss": 0.7867,
+      "step": 12872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53983839594576e-05,
+      "loss": 1.1446,
+      "step": 12873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5397666107502566e-05,
+      "loss": 0.9138,
+      "step": 12874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5396948216295104e-05,
+      "loss": 1.2251,
+      "step": 12875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539623028584044e-05,
+      "loss": 0.6603,
+      "step": 12876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5395512316143783e-05,
+      "loss": 1.0679,
+      "step": 12877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5394794307210363e-05,
+      "loss": 0.8567,
+      "step": 12878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5394076259045403e-05,
+      "loss": 1.2116,
+      "step": 12879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539335817165412e-05,
+      "loss": 0.9886,
+      "step": 12880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5392640045041738e-05,
+      "loss": 0.8932,
+      "step": 12881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539192187921348e-05,
+      "loss": 1.0617,
+      "step": 12882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539120367417456e-05,
+      "loss": 1.1887,
+      "step": 12883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5390485429930214e-05,
+      "loss": 0.9165,
+      "step": 12884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538976714648566e-05,
+      "loss": 1.1124,
+      "step": 12885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538904882384612e-05,
+      "loss": 0.6905,
+      "step": 12886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538833046201682e-05,
+      "loss": 0.8471,
+      "step": 12887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538761206100298e-05,
+      "loss": 0.8518,
+      "step": 12888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5386893620809825e-05,
+      "loss": 0.817,
+      "step": 12889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5386175141442586e-05,
+      "loss": 0.9349,
+      "step": 12890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5385456622906484e-05,
+      "loss": 0.8137,
+      "step": 12891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5384738065206737e-05,
+      "loss": 1.2237,
+      "step": 12892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538401946834858e-05,
+      "loss": 1.3644,
+      "step": 12893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538330083233724e-05,
+      "loss": 0.7606,
+      "step": 12894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538258215717793e-05,
+      "loss": 1.0866,
+      "step": 12895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538186344287589e-05,
+      "loss": 1.0093,
+      "step": 12896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5381144689436343e-05,
+      "loss": 1.1799,
+      "step": 12897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5380425896864506e-05,
+      "loss": 1.3429,
+      "step": 12898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5379707065165623e-05,
+      "loss": 0.6966,
+      "step": 12899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5378988194344913e-05,
+      "loss": 0.9847,
+      "step": 12900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53782692844076e-05,
+      "loss": 0.8977,
+      "step": 12901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5377550335358917e-05,
+      "loss": 1.1068,
+      "step": 12902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537683134720409e-05,
+      "loss": 0.88,
+      "step": 12903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5376112319948345e-05,
+      "loss": 1.088,
+      "step": 12904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537539325359692e-05,
+      "loss": 1.1413,
+      "step": 12905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5374674148155037e-05,
+      "loss": 0.7767,
+      "step": 12906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373955003627923e-05,
+      "loss": 1.0683,
+      "step": 12907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373235820020815e-05,
+      "loss": 0.8973,
+      "step": 12908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5372516597338937e-05,
+      "loss": 0.8898,
+      "step": 12909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5371797335587528e-05,
+      "loss": 1.1043,
+      "step": 12910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5371078034771806e-05,
+      "loss": 0.8201,
+      "step": 12911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537035869489701e-05,
+      "loss": 0.849,
+      "step": 12912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5369639315968368e-05,
+      "loss": 0.8352,
+      "step": 12913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5368919897991113e-05,
+      "loss": 1.0173,
+      "step": 12914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5368200440970478e-05,
+      "loss": 1.5365,
+      "step": 12915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536748094491169e-05,
+      "loss": 1.1188,
+      "step": 12916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5366761409819988e-05,
+      "loss": 1.1055,
+      "step": 12917
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53660418357006e-05,
+      "loss": 1.3283,
+      "step": 12918
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5365322222558758e-05,
+      "loss": 1.0525,
+      "step": 12919
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53646025703997e-05,
+      "loss": 0.7208,
+      "step": 12920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5363882879228654e-05,
+      "loss": 0.8942,
+      "step": 12921
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536316314905086e-05,
+      "loss": 0.7515,
+      "step": 12922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536244337987154e-05,
+      "loss": 1.0057,
+      "step": 12923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361723571695945e-05,
+      "loss": 1.064,
+      "step": 12924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5361003724529295e-05,
+      "loss": 1.078,
+      "step": 12925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5360283838376832e-05,
+      "loss": 0.9515,
+      "step": 12926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535956391324379e-05,
+      "loss": 1.1712,
+      "step": 12927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5358843949135404e-05,
+      "loss": 0.9195,
+      "step": 12928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535812394605691e-05,
+      "loss": 1.0682,
+      "step": 12929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5357403904013546e-05,
+      "loss": 0.9303,
+      "step": 12930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5356683823010542e-05,
+      "loss": 1.0941,
+      "step": 12931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535596370305314e-05,
+      "loss": 0.7747,
+      "step": 12932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5355243544146575e-05,
+      "loss": 0.9787,
+      "step": 12933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5354523346296084e-05,
+      "loss": 1.035,
+      "step": 12934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353803109506906e-05,
+      "loss": 1.0701,
+      "step": 12935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353082833784276e-05,
+      "loss": 0.9613,
+      "step": 12936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5352362519133434e-05,
+      "loss": 0.9886,
+      "step": 12937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535164216555962e-05,
+      "loss": 0.9282,
+      "step": 12938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350921773068065e-05,
+      "loss": 0.8367,
+      "step": 12939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350201341664014e-05,
+      "loss": 0.8188,
+      "step": 12940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5349480871352704e-05,
+      "loss": 0.8302,
+      "step": 12941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5348760362139377e-05,
+      "loss": 0.8191,
+      "step": 12942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534803981402927e-05,
+      "loss": 0.9665,
+      "step": 12943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5347319227027622e-05,
+      "loss": 0.7658,
+      "step": 12944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5346598601139677e-05,
+      "loss": 0.6853,
+      "step": 12945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5345877936370673e-05,
+      "loss": 0.824,
+      "step": 12946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534515723272585e-05,
+      "loss": 1.0884,
+      "step": 12947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534443649021045e-05,
+      "loss": 0.6636,
+      "step": 12948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5343715708829714e-05,
+      "loss": 1.0355,
+      "step": 12949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5342994888588887e-05,
+      "loss": 1.1251,
+      "step": 12950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5342274029493205e-05,
+      "loss": 0.9357,
+      "step": 12951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5341553131547914e-05,
+      "loss": 1.2019,
+      "step": 12952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5340832194758257e-05,
+      "loss": 1.036,
+      "step": 12953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534011121912947e-05,
+      "loss": 0.7856,
+      "step": 12954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5339390204666806e-05,
+      "loss": 0.8858,
+      "step": 12955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53386691513755e-05,
+      "loss": 1.0038,
+      "step": 12956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5337948059260802e-05,
+      "loss": 1.0313,
+      "step": 12957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533722692832795e-05,
+      "loss": 1.2045,
+      "step": 12958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5336505758582197e-05,
+      "loss": 0.7039,
+      "step": 12959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533578455002878e-05,
+      "loss": 1.1378,
+      "step": 12960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533506330267294e-05,
+      "loss": 0.9687,
+      "step": 12961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533434201651993e-05,
+      "loss": 1.4597,
+      "step": 12962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5333620691574995e-05,
+      "loss": 1.3869,
+      "step": 12963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5332899327843375e-05,
+      "loss": 1.0421,
+      "step": 12964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5332177925330317e-05,
+      "loss": 0.86,
+      "step": 12965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533145648404107e-05,
+      "loss": 0.9518,
+      "step": 12966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5330735003980883e-05,
+      "loss": 0.8841,
+      "step": 12967
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5330013485154996e-05,
+      "loss": 1.0866,
+      "step": 12968
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532929192756866e-05,
+      "loss": 0.8772,
+      "step": 12969
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5328570331227127e-05,
+      "loss": 0.9148,
+      "step": 12970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532784869613563e-05,
+      "loss": 0.9768,
+      "step": 12971
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5327127022299428e-05,
+      "loss": 0.9571,
+      "step": 12972
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532640530972377e-05,
+      "loss": 0.8617,
+      "step": 12973
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.53256835584139e-05,
+      "loss": 1.1544,
+      "step": 12974
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5324961768375065e-05,
+      "loss": 0.9389,
+      "step": 12975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5324239939612518e-05,
+      "loss": 0.8525,
+      "step": 12976
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5323518072131505e-05,
+      "loss": 0.9109,
+      "step": 12977
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532279616593728e-05,
+      "loss": 1.0519,
+      "step": 12978
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532207422103509e-05,
+      "loss": 0.8637,
+      "step": 12979
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5321352237430183e-05,
+      "loss": 0.856,
+      "step": 12980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5320630215127814e-05,
+      "loss": 1.0439,
+      "step": 12981
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531990815413323e-05,
+      "loss": 1.0743,
+      "step": 12982
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5319186054451688e-05,
+      "loss": 0.9369,
+      "step": 12983
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5318463916088428e-05,
+      "loss": 0.9106,
+      "step": 12984
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531774173904871e-05,
+      "loss": 1.0994,
+      "step": 12985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5317019523337783e-05,
+      "loss": 0.8145,
+      "step": 12986
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5316297268960902e-05,
+      "loss": 0.8975,
+      "step": 12987
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5315574975923313e-05,
+      "loss": 0.9601,
+      "step": 12988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5314852644230275e-05,
+      "loss": 0.8374,
+      "step": 12989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531413027388704e-05,
+      "loss": 0.831,
+      "step": 12990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5313407864898858e-05,
+      "loss": 0.9855,
+      "step": 12991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5312685417270984e-05,
+      "loss": 0.9939,
+      "step": 12992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531196293100867e-05,
+      "loss": 0.9468,
+      "step": 12993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5311240406117173e-05,
+      "loss": 1.1713,
+      "step": 12994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5310517842601744e-05,
+      "loss": 0.9302,
+      "step": 12995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309795240467643e-05,
+      "loss": 0.9791,
+      "step": 12996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309072599720122e-05,
+      "loss": 0.9749,
+      "step": 12997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5308349920364433e-05,
+      "loss": 1.0051,
+      "step": 12998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530762720240583e-05,
+      "loss": 1.2657,
+      "step": 12999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530690444584958e-05,
+      "loss": 1.119,
+      "step": 13000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530618165070093e-05,
+      "loss": 1.1323,
+      "step": 13001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5305458816965134e-05,
+      "loss": 1.0742,
+      "step": 13002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304735944647452e-05,
+      "loss": 1.2244,
+      "step": 13003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304013033753145e-05,
+      "loss": 0.8766,
+      "step": 13004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5303290084287465e-05,
+      "loss": 1.446,
+      "step": 13005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530256709625567e-05,
+      "loss": 0.9492,
+      "step": 13006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530184406966302e-05,
+      "loss": 1.0159,
+      "step": 13007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5301121004514766e-05,
+      "loss": 0.7068,
+      "step": 13008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5300397900816174e-05,
+      "loss": 0.8645,
+      "step": 13009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5299674758572507e-05,
+      "loss": 0.8651,
+      "step": 13010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298951577789007e-05,
+      "loss": 1.1175,
+      "step": 13011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298228358470945e-05,
+      "loss": 0.8441,
+      "step": 13012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529750510062358e-05,
+      "loss": 0.8855,
+      "step": 13013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5296781804252166e-05,
+      "loss": 0.8815,
+      "step": 13014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5296058469361975e-05,
+      "loss": 1.1442,
+      "step": 13015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5295335095958253e-05,
+      "loss": 1.1909,
+      "step": 13016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5294611684046262e-05,
+      "loss": 0.6719,
+      "step": 13017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5293888233631277e-05,
+      "loss": 1.0761,
+      "step": 13018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529316474471854e-05,
+      "loss": 0.9202,
+      "step": 13019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5292441217313324e-05,
+      "loss": 1.065,
+      "step": 13020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529171765142089e-05,
+      "loss": 0.8635,
+      "step": 13021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5290994047046494e-05,
+      "loss": 0.9855,
+      "step": 13022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5290270404195404e-05,
+      "loss": 0.7933,
+      "step": 13023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5289546722872878e-05,
+      "loss": 0.9153,
+      "step": 13024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5288823003084187e-05,
+      "loss": 1.1324,
+      "step": 13025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528809924483458e-05,
+      "loss": 0.8496,
+      "step": 13026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528737544812933e-05,
+      "loss": 0.8931,
+      "step": 13027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52866516129737e-05,
+      "loss": 1.0536,
+      "step": 13028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528592773937295e-05,
+      "loss": 0.7822,
+      "step": 13029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5285203827332352e-05,
+      "loss": 1.1262,
+      "step": 13030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5284479876857162e-05,
+      "loss": 0.9232,
+      "step": 13031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5283755887952646e-05,
+      "loss": 0.8459,
+      "step": 13032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5283031860624073e-05,
+      "loss": 0.7519,
+      "step": 13033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5282307794876704e-05,
+      "loss": 0.5562,
+      "step": 13034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5281583690715805e-05,
+      "loss": 1.4065,
+      "step": 13035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5280859548146648e-05,
+      "loss": 1.0662,
+      "step": 13036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528013536717449e-05,
+      "loss": 1.0355,
+      "step": 13037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5279411147804598e-05,
+      "loss": 0.8812,
+      "step": 13038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5278686890042246e-05,
+      "loss": 0.8886,
+      "step": 13039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5277962593892697e-05,
+      "loss": 0.8102,
+      "step": 13040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5277238259361218e-05,
+      "loss": 0.8627,
+      "step": 13041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5276513886453072e-05,
+      "loss": 0.7272,
+      "step": 13042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275789475173537e-05,
+      "loss": 0.8263,
+      "step": 13043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275065025527872e-05,
+      "loss": 1.1631,
+      "step": 13044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527434053752135e-05,
+      "loss": 0.876,
+      "step": 13045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5273616011159235e-05,
+      "loss": 0.9717,
+      "step": 13046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52728914464468e-05,
+      "loss": 0.8776,
+      "step": 13047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5272166843389313e-05,
+      "loss": 1.1887,
+      "step": 13048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5271442201992048e-05,
+      "loss": 1.1366,
+      "step": 13049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5270717522260264e-05,
+      "loss": 0.993,
+      "step": 13050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526999280419924e-05,
+      "loss": 0.9831,
+      "step": 13051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526926804781424e-05,
+      "loss": 1.0841,
+      "step": 13052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526854325311054e-05,
+      "loss": 0.8651,
+      "step": 13053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5267818420093402e-05,
+      "loss": 0.636,
+      "step": 13054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5267093548768112e-05,
+      "loss": 0.7907,
+      "step": 13055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5266368639139928e-05,
+      "loss": 1.1353,
+      "step": 13056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5265643691214128e-05,
+      "loss": 1.0226,
+      "step": 13057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5264918704995984e-05,
+      "loss": 0.6296,
+      "step": 13058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5264193680490765e-05,
+      "loss": 0.7714,
+      "step": 13059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5263468617703743e-05,
+      "loss": 1.0767,
+      "step": 13060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5262743516640197e-05,
+      "loss": 1.1433,
+      "step": 13061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5262018377305395e-05,
+      "loss": 1.0826,
+      "step": 13062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526129319970461e-05,
+      "loss": 1.0308,
+      "step": 13063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5260567983843114e-05,
+      "loss": 0.9617,
+      "step": 13064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5259842729726186e-05,
+      "loss": 1.0532,
+      "step": 13065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5259117437359097e-05,
+      "loss": 1.3113,
+      "step": 13066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5258392106747125e-05,
+      "loss": 0.8243,
+      "step": 13067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5257666737895539e-05,
+      "loss": 1.0481,
+      "step": 13068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5256941330809614e-05,
+      "loss": 0.836,
+      "step": 13069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5256215885494632e-05,
+      "loss": 1.0888,
+      "step": 13070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5255490401955864e-05,
+      "loss": 0.8552,
+      "step": 13071
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5254764880198583e-05,
+      "loss": 1.2936,
+      "step": 13072
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5254039320228068e-05,
+      "loss": 0.9247,
+      "step": 13073
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52533137220496e-05,
+      "loss": 1.005,
+      "step": 13074
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5252588085668449e-05,
+      "loss": 1.1389,
+      "step": 13075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5251862411089895e-05,
+      "loss": 0.8627,
+      "step": 13076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5251136698319214e-05,
+      "loss": 1.0209,
+      "step": 13077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5250410947361682e-05,
+      "loss": 0.6321,
+      "step": 13078
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5249685158222578e-05,
+      "loss": 0.9658,
+      "step": 13079
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5248959330907186e-05,
+      "loss": 0.8109,
+      "step": 13080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5248233465420773e-05,
+      "loss": 1.0276,
+      "step": 13081
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5247507561768626e-05,
+      "loss": 1.4398,
+      "step": 13082
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5246781619956022e-05,
+      "loss": 1.0538,
+      "step": 13083
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5246055639988235e-05,
+      "loss": 1.0926,
+      "step": 13084
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5245329621870557e-05,
+      "loss": 1.1023,
+      "step": 13085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5244603565608254e-05,
+      "loss": 1.0762,
+      "step": 13086
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243877471206609e-05,
+      "loss": 1.02,
+      "step": 13087
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243151338670909e-05,
+      "loss": 0.7022,
+      "step": 13088
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5242425168006429e-05,
+      "loss": 0.9856,
+      "step": 13089
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5241698959218454e-05,
+      "loss": 0.9102,
+      "step": 13090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5240972712312259e-05,
+      "loss": 1.1281,
+      "step": 13091
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5240246427293129e-05,
+      "loss": 1.006,
+      "step": 13092
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5239520104166343e-05,
+      "loss": 1.0129,
+      "step": 13093
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5238793742937189e-05,
+      "loss": 1.0642,
+      "step": 13094
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5238067343610943e-05,
+      "loss": 0.93,
+      "step": 13095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5237340906192888e-05,
+      "loss": 0.951,
+      "step": 13096
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523661443068831e-05,
+      "loss": 0.9229,
+      "step": 13097
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5235887917102491e-05,
+      "loss": 0.8481,
+      "step": 13098
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5235161365440711e-05,
+      "loss": 0.9308,
+      "step": 13099
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.523443477570826e-05,
+      "loss": 1.0243,
+      "step": 13100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5233708147910418e-05,
+      "loss": 1.254,
+      "step": 13101
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5232981482052465e-05,
+      "loss": 0.8271,
+      "step": 13102
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5232254778139691e-05,
+      "loss": 0.9083,
+      "step": 13103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5231528036177382e-05,
+      "loss": 0.8306,
+      "step": 13104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230801256170818e-05,
+      "loss": 0.8701,
+      "step": 13105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230074438125287e-05,
+      "loss": 1.1536,
+      "step": 13106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5229347582046073e-05,
+      "loss": 1.5247,
+      "step": 13107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5228620687938462e-05,
+      "loss": 1.106,
+      "step": 13108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522789375580774e-05,
+      "loss": 0.864,
+      "step": 13109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.52271667856592e-05,
+      "loss": 0.7356,
+      "step": 13110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5226439777498116e-05,
+      "loss": 1.1391,
+      "step": 13111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5225712731329785e-05,
+      "loss": 0.8545,
+      "step": 13112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224985647159489e-05,
+      "loss": 0.9422,
+      "step": 13113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522425852499252e-05,
+      "loss": 0.8504,
+      "step": 13114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522353136483416e-05,
+      "loss": 0.9367,
+      "step": 13115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5222804166689703e-05,
+      "loss": 1.1308,
+      "step": 13116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5222076930564428e-05,
+      "loss": 1.0385,
+      "step": 13117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5221349656463636e-05,
+      "loss": 0.9708,
+      "step": 13118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5220622344392607e-05,
+      "loss": 1.0421,
+      "step": 13119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5219894994356632e-05,
+      "loss": 1.1463,
+      "step": 13120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5219167606361e-05,
+      "loss": 1.1801,
+      "step": 13121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5218440180411005e-05,
+      "loss": 0.8544,
+      "step": 13122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5217712716511932e-05,
+      "loss": 1.1491,
+      "step": 13123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216985214669071e-05,
+      "loss": 0.9762,
+      "step": 13124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216257674887718e-05,
+      "loss": 1.1654,
+      "step": 13125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5215530097173155e-05,
+      "loss": 1.2461,
+      "step": 13126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5214802481530681e-05,
+      "loss": 0.9968,
+      "step": 13127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5214074827965583e-05,
+      "loss": 0.8929,
+      "step": 13128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5213347136483156e-05,
+      "loss": 0.9067,
+      "step": 13129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212619407088689e-05,
+      "loss": 1.1215,
+      "step": 13130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5211891639787475e-05,
+      "loss": 0.964,
+      "step": 13131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5211163834584805e-05,
+      "loss": 1.1188,
+      "step": 13132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5210435991485971e-05,
+      "loss": 0.8458,
+      "step": 13133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5209708110496273e-05,
+      "loss": 1.4199,
+      "step": 13134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5208980191620994e-05,
+      "loss": 1.1463,
+      "step": 13135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5208252234865435e-05,
+      "loss": 0.9211,
+      "step": 13136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5207524240234889e-05,
+      "loss": 0.8074,
+      "step": 13137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206796207734646e-05,
+      "loss": 1.1091,
+      "step": 13138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206068137370002e-05,
+      "loss": 1.1161,
+      "step": 13139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5205340029146256e-05,
+      "loss": 0.8533,
+      "step": 13140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5204611883068699e-05,
+      "loss": 1.0503,
+      "step": 13141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5203883699142621e-05,
+      "loss": 1.0452,
+      "step": 13142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5203155477373326e-05,
+      "loss": 1.3658,
+      "step": 13143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5202427217766107e-05,
+      "loss": 1.3259,
+      "step": 13144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5201698920326263e-05,
+      "loss": 1.0623,
+      "step": 13145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520097058505908e-05,
+      "loss": 1.2988,
+      "step": 13146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5200242211969867e-05,
+      "loss": 0.7574,
+      "step": 13147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5199513801063911e-05,
+      "loss": 0.8457,
+      "step": 13148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198785352346514e-05,
+      "loss": 1.0078,
+      "step": 13149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198056865822976e-05,
+      "loss": 1.0017,
+      "step": 13150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519732834149859e-05,
+      "loss": 0.7996,
+      "step": 13151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5196599779378654e-05,
+      "loss": 1.2033,
+      "step": 13152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5195871179468467e-05,
+      "loss": 0.8525,
+      "step": 13153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5195142541773326e-05,
+      "loss": 1.0789,
+      "step": 13154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5194413866298536e-05,
+      "loss": 0.8959,
+      "step": 13155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519368515304939e-05,
+      "loss": 1.2031,
+      "step": 13156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5192956402031188e-05,
+      "loss": 0.9556,
+      "step": 13157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5192227613249229e-05,
+      "loss": 0.9866,
+      "step": 13158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5191498786708816e-05,
+      "loss": 0.778,
+      "step": 13159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5190769922415248e-05,
+      "loss": 0.7803,
+      "step": 13160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5190041020373823e-05,
+      "loss": 1.0005,
+      "step": 13161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5189312080589843e-05,
+      "loss": 1.1254,
+      "step": 13162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518858310306861e-05,
+      "loss": 0.6551,
+      "step": 13163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187854087815424e-05,
+      "loss": 0.8399,
+      "step": 13164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187125034835591e-05,
+      "loss": 1.2739,
+      "step": 13165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5186395944134403e-05,
+      "loss": 1.078,
+      "step": 13166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5185666815717173e-05,
+      "loss": 0.5982,
+      "step": 13167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5184937649589193e-05,
+      "loss": 0.8295,
+      "step": 13168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5184208445755776e-05,
+      "loss": 0.9885,
+      "step": 13169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5183479204222216e-05,
+      "loss": 1.0295,
+      "step": 13170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182749924993818e-05,
+      "loss": 1.0232,
+      "step": 13171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182020608075889e-05,
+      "loss": 1.0072,
+      "step": 13172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.518129125347373e-05,
+      "loss": 0.6927,
+      "step": 13173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5180561861192646e-05,
+      "loss": 0.8528,
+      "step": 13174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5179832431237942e-05,
+      "loss": 0.7112,
+      "step": 13175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5179102963614922e-05,
+      "loss": 1.2142,
+      "step": 13176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5178373458328887e-05,
+      "loss": 1.002,
+      "step": 13177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.517764391538515e-05,
+      "loss": 0.8088,
+      "step": 13178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5176914334789005e-05,
+      "loss": 0.9147,
+      "step": 13179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.517618471654577e-05,
+      "loss": 0.8289,
+      "step": 13180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5175455060660741e-05,
+      "loss": 0.9728,
+      "step": 13181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5174725367139228e-05,
+      "loss": 0.8542,
+      "step": 13182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173995635986542e-05,
+      "loss": 0.6797,
+      "step": 13183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173265867207983e-05,
+      "loss": 1.0362,
+      "step": 13184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5172536060808857e-05,
+      "loss": 0.8691,
+      "step": 13185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5171806216794479e-05,
+      "loss": 0.8674,
+      "step": 13186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.517107633517015e-05,
+      "loss": 0.5949,
+      "step": 13187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5170346415941177e-05,
+      "loss": 0.6494,
+      "step": 13188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5169616459112874e-05,
+      "loss": 0.7374,
+      "step": 13189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5168886464690547e-05,
+      "loss": 1.1431,
+      "step": 13190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51681564326795e-05,
+      "loss": 1.132,
+      "step": 13191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5167426363085048e-05,
+      "loss": 0.7897,
+      "step": 13192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166696255912496e-05,
+      "loss": 1.139,
+      "step": 13193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5165966111167154e-05,
+      "loss": 0.796,
+      "step": 13194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5165235928854337e-05,
+      "loss": 0.9335,
+      "step": 13195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5164505708979349e-05,
+      "loss": 0.84,
+      "step": 13196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51637754515475e-05,
+      "loss": 1.2238,
+      "step": 13197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5163045156564105e-05,
+      "loss": 1.5205,
+      "step": 13198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516231482403447e-05,
+      "loss": 0.8924,
+      "step": 13199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5161584453963908e-05,
+      "loss": 1.0372,
+      "step": 13200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5160854046357734e-05,
+      "loss": 0.949,
+      "step": 13201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5160123601221255e-05,
+      "loss": 1.0455,
+      "step": 13202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5159393118559783e-05,
+      "loss": 0.7946,
+      "step": 13203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5158662598378633e-05,
+      "loss": 0.8847,
+      "step": 13204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5157932040683116e-05,
+      "loss": 1.0938,
+      "step": 13205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5157201445478543e-05,
+      "loss": 0.8313,
+      "step": 13206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5156470812770225e-05,
+      "loss": 0.841,
+      "step": 13207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5155740142563482e-05,
+      "loss": 0.6792,
+      "step": 13208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5155009434863623e-05,
+      "loss": 1.4606,
+      "step": 13209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5154278689675967e-05,
+      "loss": 0.8389,
+      "step": 13210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5153547907005823e-05,
+      "loss": 0.6213,
+      "step": 13211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51528170868585e-05,
+      "loss": 0.9986,
+      "step": 13212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5152086229239325e-05,
+      "loss": 0.8848,
+      "step": 13213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5151355334153605e-05,
+      "loss": 0.8786,
+      "step": 13214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150624401606658e-05,
+      "loss": 0.5656,
+      "step": 13215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5149893431603796e-05,
+      "loss": 1.3797,
+      "step": 13216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514916242415034e-05,
+      "loss": 0.9308,
+      "step": 13217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5148431379251603e-05,
+      "loss": 1.2191,
+      "step": 13218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5147700296912899e-05,
+      "loss": 1.2395,
+      "step": 13219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514696917713955e-05,
+      "loss": 1.0282,
+      "step": 13220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5146238019936865e-05,
+      "loss": 1.3369,
+      "step": 13221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514550682531017e-05,
+      "loss": 0.7841,
+      "step": 13222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5144775593264776e-05,
+      "loss": 0.784,
+      "step": 13223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5144044323806e-05,
+      "loss": 1.1181,
+      "step": 13224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5143313016939167e-05,
+      "loss": 1.2265,
+      "step": 13225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514258167266959e-05,
+      "loss": 0.8709,
+      "step": 13226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5141850291002585e-05,
+      "loss": 1.0385,
+      "step": 13227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5141118871943474e-05,
+      "loss": 0.9184,
+      "step": 13228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514038741549758e-05,
+      "loss": 1.0152,
+      "step": 13229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5139655921670213e-05,
+      "loss": 0.9336,
+      "step": 13230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51389243904667e-05,
+      "loss": 0.835,
+      "step": 13231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5138192821892357e-05,
+      "loss": 1.4188,
+      "step": 13232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5137461215952505e-05,
+      "loss": 0.8669,
+      "step": 13233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5136729572652465e-05,
+      "loss": 0.7596,
+      "step": 13234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513599789199756e-05,
+      "loss": 1.0219,
+      "step": 13235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5135266173993101e-05,
+      "loss": 1.2191,
+      "step": 13236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5134534418644422e-05,
+      "loss": 0.8553,
+      "step": 13237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5133802625956838e-05,
+      "loss": 0.9278,
+      "step": 13238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5133070795935672e-05,
+      "loss": 0.5943,
+      "step": 13239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5132338928586241e-05,
+      "loss": 0.8163,
+      "step": 13240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5131607023913875e-05,
+      "loss": 0.9706,
+      "step": 13241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130875081923888e-05,
+      "loss": 0.8948,
+      "step": 13242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130143102621613e-05,
+      "loss": 0.7697,
+      "step": 13243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5129411086012367e-05,
+      "loss": 0.8876,
+      "step": 13244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5128679032101472e-05,
+      "loss": 0.9626,
+      "step": 13245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5127946940894256e-05,
+      "loss": 0.814,
+      "step": 13246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5127214812396038e-05,
+      "loss": 1.0023,
+      "step": 13247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126482646612146e-05,
+      "loss": 0.8858,
+      "step": 13248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5125750443547902e-05,
+      "loss": 0.781,
+      "step": 13249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5125018203208633e-05,
+      "loss": 1.0621,
+      "step": 13250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512428592559966e-05,
+      "loss": 1.1919,
+      "step": 13251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5123553610726312e-05,
+      "loss": 1.4943,
+      "step": 13252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5122821258593913e-05,
+      "loss": 1.0673,
+      "step": 13253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512208886920779e-05,
+      "loss": 0.9857,
+      "step": 13254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5121356442573265e-05,
+      "loss": 1.0028,
+      "step": 13255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5120623978695669e-05,
+      "loss": 0.8913,
+      "step": 13256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5119891477580328e-05,
+      "loss": 1.1195,
+      "step": 13257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5119158939232566e-05,
+      "loss": 1.0948,
+      "step": 13258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5118426363657714e-05,
+      "loss": 1.2238,
+      "step": 13259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5117693750861096e-05,
+      "loss": 0.9159,
+      "step": 13260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511696110084804e-05,
+      "loss": 0.8616,
+      "step": 13261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5116228413623875e-05,
+      "loss": 1.0115,
+      "step": 13262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511549568919393e-05,
+      "loss": 0.9111,
+      "step": 13263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114762927563528e-05,
+      "loss": 1.3472,
+      "step": 13264
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114030128738008e-05,
+      "loss": 0.9974,
+      "step": 13265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511329729272269e-05,
+      "loss": 0.8274,
+      "step": 13266
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5112564419522905e-05,
+      "loss": 1.0858,
+      "step": 13267
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111831509143984e-05,
+      "loss": 0.7184,
+      "step": 13268
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511109856159126e-05,
+      "loss": 1.176,
+      "step": 13269
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5110365576870059e-05,
+      "loss": 0.7947,
+      "step": 13270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109632554985709e-05,
+      "loss": 0.553,
+      "step": 13271
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5108899495943545e-05,
+      "loss": 1.0334,
+      "step": 13272
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5108166399748896e-05,
+      "loss": 0.9102,
+      "step": 13273
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5107433266407094e-05,
+      "loss": 1.0087,
+      "step": 13274
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5106700095923471e-05,
+      "loss": 1.0929,
+      "step": 13275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5105966888303356e-05,
+      "loss": 1.0174,
+      "step": 13276
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5105233643552083e-05,
+      "loss": 0.932,
+      "step": 13277
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.510450036167499e-05,
+      "loss": 0.8836,
+      "step": 13278
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5103767042677395e-05,
+      "loss": 1.2619,
+      "step": 13279
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5103033686564645e-05,
+      "loss": 0.5615,
+      "step": 13280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5102300293342068e-05,
+      "loss": 1.1851,
+      "step": 13281
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5101566863014992e-05,
+      "loss": 0.9187,
+      "step": 13282
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5100833395588757e-05,
+      "loss": 1.0906,
+      "step": 13283
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5100099891068696e-05,
+      "loss": 0.7782,
+      "step": 13284
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5099366349460141e-05,
+      "loss": 1.3413,
+      "step": 13285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098632770768427e-05,
+      "loss": 0.9925,
+      "step": 13286
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5097899154998892e-05,
+      "loss": 0.99,
+      "step": 13287
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5097165502156865e-05,
+      "loss": 0.9333,
+      "step": 13288
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5096431812247687e-05,
+      "loss": 0.6283,
+      "step": 13289
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5095698085276692e-05,
+      "loss": 1.1545,
+      "step": 13290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509496432124921e-05,
+      "loss": 0.9926,
+      "step": 13291
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5094230520170586e-05,
+      "loss": 0.992,
+      "step": 13292
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509349668204615e-05,
+      "loss": 1.0385,
+      "step": 13293
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5092762806881238e-05,
+      "loss": 0.7356,
+      "step": 13294
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5092028894681194e-05,
+      "loss": 0.806,
+      "step": 13295
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509129494545135e-05,
+      "loss": 0.9221,
+      "step": 13296
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509056095919704e-05,
+      "loss": 1.2239,
+      "step": 13297
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508982693592361e-05,
+      "loss": 0.9495,
+      "step": 13298
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508909287563639e-05,
+      "loss": 1.0981,
+      "step": 13299
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5088358778340725e-05,
+      "loss": 0.9188,
+      "step": 13300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5087624644041947e-05,
+      "loss": 1.0497,
+      "step": 13301
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5086890472745399e-05,
+      "loss": 0.7008,
+      "step": 13302
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5086156264456415e-05,
+      "loss": 0.9022,
+      "step": 13303
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5085422019180341e-05,
+      "loss": 0.6485,
+      "step": 13304
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5084687736922514e-05,
+      "loss": 0.9191,
+      "step": 13305
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5083953417688273e-05,
+      "loss": 1.0014,
+      "step": 13306
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5083219061482957e-05,
+      "loss": 1.0803,
+      "step": 13307
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5082484668311907e-05,
+      "loss": 1.1558,
+      "step": 13308
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5081750238180464e-05,
+      "loss": 0.7799,
+      "step": 13309
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5081015771093971e-05,
+      "loss": 0.9197,
+      "step": 13310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5080281267057765e-05,
+      "loss": 0.9316,
+      "step": 13311
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507954672607719e-05,
+      "loss": 0.8826,
+      "step": 13312
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5078812148157589e-05,
+      "loss": 1.1736,
+      "step": 13313
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5078077533304298e-05,
+      "loss": 1.052,
+      "step": 13314
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5077342881522667e-05,
+      "loss": 1.2896,
+      "step": 13315
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5076608192818034e-05,
+      "loss": 0.7002,
+      "step": 13316
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5075873467195744e-05,
+      "loss": 1.1872,
+      "step": 13317
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5075138704661137e-05,
+      "loss": 0.8379,
+      "step": 13318
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5074403905219554e-05,
+      "loss": 0.9202,
+      "step": 13319
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5073669068876348e-05,
+      "loss": 1.0445,
+      "step": 13320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5072934195636854e-05,
+      "loss": 1.0226,
+      "step": 13321
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5072199285506419e-05,
+      "loss": 0.9291,
+      "step": 13322
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507146433849039e-05,
+      "loss": 0.9004,
+      "step": 13323
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507072935459411e-05,
+      "loss": 1.4585,
+      "step": 13324
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506999433382292e-05,
+      "loss": 0.7389,
+      "step": 13325
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5069259276182168e-05,
+      "loss": 1.5583,
+      "step": 13326
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.50685241816772e-05,
+      "loss": 0.8688,
+      "step": 13327
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5067789050313359e-05,
+      "loss": 1.0082,
+      "step": 13328
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5067053882096e-05,
+      "loss": 0.506,
+      "step": 13329
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5066318677030456e-05,
+      "loss": 1.0955,
+      "step": 13330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5065583435122082e-05,
+      "loss": 0.8164,
+      "step": 13331
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5064848156376224e-05,
+      "loss": 1.0668,
+      "step": 13332
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5064112840798229e-05,
+      "loss": 0.8936,
+      "step": 13333
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5063377488393438e-05,
+      "loss": 0.9735,
+      "step": 13334
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5062642099167208e-05,
+      "loss": 1.1072,
+      "step": 13335
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5061906673124881e-05,
+      "loss": 0.9567,
+      "step": 13336
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5061171210271807e-05,
+      "loss": 1.0615,
+      "step": 13337
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5060435710613334e-05,
+      "loss": 0.7089,
+      "step": 13338
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5059700174154811e-05,
+      "loss": 1.1165,
+      "step": 13339
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5058964600901584e-05,
+      "loss": 0.9998,
+      "step": 13340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5058228990859007e-05,
+      "loss": 1.2524,
+      "step": 13341
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5057493344032428e-05,
+      "loss": 1.3027,
+      "step": 13342
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5056757660427193e-05,
+      "loss": 0.9204,
+      "step": 13343
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5056021940048658e-05,
+      "loss": 1.259,
+      "step": 13344
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5055286182902169e-05,
+      "loss": 1.232,
+      "step": 13345
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5054550388993074e-05,
+      "loss": 1.1333,
+      "step": 13346
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5053814558326729e-05,
+      "loss": 0.9076,
+      "step": 13347
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5053078690908485e-05,
+      "loss": 0.9132,
+      "step": 13348
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5052342786743688e-05,
+      "loss": 1.37,
+      "step": 13349
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5051606845837699e-05,
+      "loss": 0.8834,
+      "step": 13350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5050870868195862e-05,
+      "loss": 0.862,
+      "step": 13351
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5050134853823529e-05,
+      "loss": 0.7141,
+      "step": 13352
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5049398802726055e-05,
+      "loss": 0.8312,
+      "step": 13353
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5048662714908792e-05,
+      "loss": 1.062,
+      "step": 13354
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5047926590377095e-05,
+      "loss": 0.5433,
+      "step": 13355
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5047190429136315e-05,
+      "loss": 0.9283,
+      "step": 13356
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5046454231191805e-05,
+      "loss": 1.3069,
+      "step": 13357
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5045717996548917e-05,
+      "loss": 1.0877,
+      "step": 13358
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504498172521301e-05,
+      "loss": 1.1338,
+      "step": 13359
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5044245417189438e-05,
+      "loss": 0.9486,
+      "step": 13360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5043509072483549e-05,
+      "loss": 1.2218,
+      "step": 13361
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042772691100705e-05,
+      "loss": 0.8554,
+      "step": 13362
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5042036273046257e-05,
+      "loss": 1.0955,
+      "step": 13363
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5041299818325556e-05,
+      "loss": 0.9917,
+      "step": 13364
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5040563326943974e-05,
+      "loss": 1.075,
+      "step": 13365
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.503982679890685e-05,
+      "loss": 0.9377,
+      "step": 13366
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.503909023421954e-05,
+      "loss": 1.0087,
+      "step": 13367
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038353632887415e-05,
+      "loss": 0.7457,
+      "step": 13368
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5037616994915818e-05,
+      "loss": 0.6415,
+      "step": 13369
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036880320310112e-05,
+      "loss": 0.8963,
+      "step": 13370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036143609075655e-05,
+      "loss": 0.75,
+      "step": 13371
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5035406861217801e-05,
+      "loss": 1.0271,
+      "step": 13372
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5034670076741907e-05,
+      "loss": 1.1407,
+      "step": 13373
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5033933255653335e-05,
+      "loss": 0.8102,
+      "step": 13374
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5033196397957444e-05,
+      "loss": 1.4695,
+      "step": 13375
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5032459503659585e-05,
+      "loss": 1.1309,
+      "step": 13376
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5031722572765125e-05,
+      "loss": 0.9608,
+      "step": 13377
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5030985605279418e-05,
+      "loss": 0.9772,
+      "step": 13378
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5030248601207826e-05,
+      "loss": 0.8982,
+      "step": 13379
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5029511560555707e-05,
+      "loss": 1.3485,
+      "step": 13380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502877448332842e-05,
+      "loss": 1.1145,
+      "step": 13381
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5028037369531327e-05,
+      "loss": 0.9191,
+      "step": 13382
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5027300219169791e-05,
+      "loss": 0.8893,
+      "step": 13383
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5026563032249167e-05,
+      "loss": 0.8458,
+      "step": 13384
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5025825808774818e-05,
+      "loss": 1.0818,
+      "step": 13385
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502508854875211e-05,
+      "loss": 0.8917,
+      "step": 13386
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5024351252186393e-05,
+      "loss": 1.3008,
+      "step": 13387
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5023613919083042e-05,
+      "loss": 1.0375,
+      "step": 13388
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5022876549447407e-05,
+      "loss": 0.9947,
+      "step": 13389
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5022139143284862e-05,
+      "loss": 0.8388,
+      "step": 13390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5021401700600757e-05,
+      "loss": 1.1678,
+      "step": 13391
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5020664221400467e-05,
+      "loss": 1.0664,
+      "step": 13392
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5019926705689347e-05,
+      "loss": 1.2652,
+      "step": 13393
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5019189153472761e-05,
+      "loss": 0.8149,
+      "step": 13394
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5018451564756078e-05,
+      "loss": 1.1392,
+      "step": 13395
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5017713939544657e-05,
+      "loss": 1.0037,
+      "step": 13396
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501697627784386e-05,
+      "loss": 1.2703,
+      "step": 13397
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5016238579659055e-05,
+      "loss": 0.6502,
+      "step": 13398
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5015500844995608e-05,
+      "loss": 0.8647,
+      "step": 13399
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501476307385888e-05,
+      "loss": 1.117,
+      "step": 13400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501402526625424e-05,
+      "loss": 1.2117,
+      "step": 13401
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013287422187049e-05,
+      "loss": 1.1161,
+      "step": 13402
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5012549541662677e-05,
+      "loss": 1.1002,
+      "step": 13403
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5011811624686488e-05,
+      "loss": 1.1905,
+      "step": 13404
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5011073671263847e-05,
+      "loss": 1.1485,
+      "step": 13405
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5010335681400125e-05,
+      "loss": 0.8642,
+      "step": 13406
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5009597655100683e-05,
+      "loss": 1.0356,
+      "step": 13407
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008859592370888e-05,
+      "loss": 1.0366,
+      "step": 13408
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008121493216115e-05,
+      "loss": 0.6287,
+      "step": 13409
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5007383357641723e-05,
+      "loss": 1.202,
+      "step": 13410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5006645185653086e-05,
+      "loss": 0.894,
+      "step": 13411
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5005906977255567e-05,
+      "loss": 1.2195,
+      "step": 13412
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5005168732454535e-05,
+      "loss": 0.9865,
+      "step": 13413
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5004430451255361e-05,
+      "loss": 0.7703,
+      "step": 13414
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003692133663417e-05,
+      "loss": 0.9455,
+      "step": 13415
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5002953779684063e-05,
+      "loss": 0.6693,
+      "step": 13416
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5002215389322676e-05,
+      "loss": 0.9998,
+      "step": 13417
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5001476962584624e-05,
+      "loss": 0.9133,
+      "step": 13418
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5000738499475275e-05,
+      "loss": 0.8585,
+      "step": 13419
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.9684,
+      "step": 13420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4999261464164172e-05,
+      "loss": 1.3399,
+      "step": 13421
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4998522891973157e-05,
+      "loss": 0.9728,
+      "step": 13422
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499778428343233e-05,
+      "loss": 0.7427,
+      "step": 13423
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4997045638547062e-05,
+      "loss": 0.9856,
+      "step": 13424
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499630695732272e-05,
+      "loss": 1.2445,
+      "step": 13425
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499556823976468e-05,
+      "loss": 0.8295,
+      "step": 13426
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4994829485878312e-05,
+      "loss": 1.0469,
+      "step": 13427
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499409069566899e-05,
+      "loss": 0.895,
+      "step": 13428
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4993351869142085e-05,
+      "loss": 1.1332,
+      "step": 13429
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4992613006302975e-05,
+      "loss": 0.6977,
+      "step": 13430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991874107157025e-05,
+      "loss": 1.0903,
+      "step": 13431
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4991135171709611e-05,
+      "loss": 0.6965,
+      "step": 13432
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4990396199966112e-05,
+      "loss": 0.6831,
+      "step": 13433
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4989657191931896e-05,
+      "loss": 1.1877,
+      "step": 13434
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4988918147612338e-05,
+      "loss": 0.7762,
+      "step": 13435
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4988179067012814e-05,
+      "loss": 1.0303,
+      "step": 13436
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4987439950138695e-05,
+      "loss": 0.9766,
+      "step": 13437
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4986700796995362e-05,
+      "loss": 0.7985,
+      "step": 13438
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4985961607588187e-05,
+      "loss": 0.9392,
+      "step": 13439
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4985222381922543e-05,
+      "loss": 0.9526,
+      "step": 13440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4984483120003812e-05,
+      "loss": 0.9042,
+      "step": 13441
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4983743821837364e-05,
+      "loss": 1.1603,
+      "step": 13442
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4983004487428579e-05,
+      "loss": 0.9311,
+      "step": 13443
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4982265116782828e-05,
+      "loss": 0.8383,
+      "step": 13444
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4981525709905498e-05,
+      "loss": 0.6913,
+      "step": 13445
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980786266801954e-05,
+      "loss": 1.1621,
+      "step": 13446
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980046787477583e-05,
+      "loss": 0.9592,
+      "step": 13447
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4979307271937757e-05,
+      "loss": 1.0344,
+      "step": 13448
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4978567720187855e-05,
+      "loss": 0.9179,
+      "step": 13449
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4977828132233257e-05,
+      "loss": 0.8499,
+      "step": 13450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497708850807934e-05,
+      "loss": 1.0248,
+      "step": 13451
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4976348847731484e-05,
+      "loss": 1.1696,
+      "step": 13452
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4975609151195063e-05,
+      "loss": 0.9755,
+      "step": 13453
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4974869418475462e-05,
+      "loss": 1.2066,
+      "step": 13454
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4974129649578058e-05,
+      "loss": 1.2862,
+      "step": 13455
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497338984450823e-05,
+      "loss": 1.1266,
+      "step": 13456
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4972650003271358e-05,
+      "loss": 1.0046,
+      "step": 13457
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4971910125872827e-05,
+      "loss": 1.5777,
+      "step": 13458
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497117021231801e-05,
+      "loss": 1.1682,
+      "step": 13459
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4970430262612293e-05,
+      "loss": 1.1891,
+      "step": 13460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4969690276761056e-05,
+      "loss": 0.9706,
+      "step": 13461
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4968950254769674e-05,
+      "loss": 0.7527,
+      "step": 13462
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4968210196643538e-05,
+      "loss": 0.8915,
+      "step": 13463
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4967470102388025e-05,
+      "loss": 1.4268,
+      "step": 13464
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4966729972008518e-05,
+      "loss": 1.2406,
+      "step": 13465
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4965989805510401e-05,
+      "loss": 1.2345,
+      "step": 13466
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4965249602899052e-05,
+      "loss": 1.0832,
+      "step": 13467
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4964509364179856e-05,
+      "loss": 1.1113,
+      "step": 13468
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4963769089358197e-05,
+      "loss": 0.8695,
+      "step": 13469
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496302877843946e-05,
+      "loss": 0.9543,
+      "step": 13470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4962288431429026e-05,
+      "loss": 1.0683,
+      "step": 13471
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496154804833228e-05,
+      "loss": 0.784,
+      "step": 13472
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4960807629154603e-05,
+      "loss": 1.1368,
+      "step": 13473
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4960067173901382e-05,
+      "loss": 1.0493,
+      "step": 13474
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4959326682578006e-05,
+      "loss": 0.8214,
+      "step": 13475
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4958586155189854e-05,
+      "loss": 1.2726,
+      "step": 13476
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495784559174231e-05,
+      "loss": 0.7375,
+      "step": 13477
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4957104992240765e-05,
+      "loss": 0.7669,
+      "step": 13478
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.49563643566906e-05,
+      "loss": 0.9057,
+      "step": 13479
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4955623685097205e-05,
+      "loss": 0.6601,
+      "step": 13480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4954882977465964e-05,
+      "loss": 0.6916,
+      "step": 13481
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4954142233802262e-05,
+      "loss": 0.6837,
+      "step": 13482
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.495340145411149e-05,
+      "loss": 0.9932,
+      "step": 13483
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952660638399031e-05,
+      "loss": 1.257,
+      "step": 13484
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4951919786670274e-05,
+      "loss": 0.7948,
+      "step": 13485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4951178898930607e-05,
+      "loss": 0.8874,
+      "step": 13486
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4950437975185417e-05,
+      "loss": 0.5045,
+      "step": 13487
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4949697015440093e-05,
+      "loss": 0.8696,
+      "step": 13488
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4948956019700018e-05,
+      "loss": 0.8206,
+      "step": 13489
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4948214987970594e-05,
+      "loss": 0.842,
+      "step": 13490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4947473920257194e-05,
+      "loss": 0.7138,
+      "step": 13491
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946732816565214e-05,
+      "loss": 0.899,
+      "step": 13492
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4945991676900047e-05,
+      "loss": 1.0108,
+      "step": 13493
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4945250501267079e-05,
+      "loss": 0.9015,
+      "step": 13494
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.49445092896717e-05,
+      "loss": 1.1371,
+      "step": 13495
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.49437680421193e-05,
+      "loss": 1.0327,
+      "step": 13496
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4943026758615271e-05,
+      "loss": 0.9664,
+      "step": 13497
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4942285439164999e-05,
+      "loss": 1.3683,
+      "step": 13498
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494154408377388e-05,
+      "loss": 0.7988,
+      "step": 13499
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4940802692447306e-05,
+      "loss": 1.0718,
+      "step": 13500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4940061265190664e-05,
+      "loss": 0.7915,
+      "step": 13501
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493931980200935e-05,
+      "loss": 1.1364,
+      "step": 13502
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4938578302908751e-05,
+      "loss": 1.0437,
+      "step": 13503
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4937836767894266e-05,
+      "loss": 0.9857,
+      "step": 13504
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4937095196971279e-05,
+      "loss": 1.035,
+      "step": 13505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493635359014519e-05,
+      "loss": 0.8032,
+      "step": 13506
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4935611947421386e-05,
+      "loss": 0.9814,
+      "step": 13507
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4934870268805268e-05,
+      "loss": 1.2549,
+      "step": 13508
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4934128554302225e-05,
+      "loss": 0.9689,
+      "step": 13509
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493338680391765e-05,
+      "loss": 1.1886,
+      "step": 13510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4932645017656939e-05,
+      "loss": 0.9789,
+      "step": 13511
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4931903195525487e-05,
+      "loss": 0.652,
+      "step": 13512
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4931161337528686e-05,
+      "loss": 1.0984,
+      "step": 13513
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4930419443671931e-05,
+      "loss": 0.8818,
+      "step": 13514
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4929677513960621e-05,
+      "loss": 1.1756,
+      "step": 13515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928935548400144e-05,
+      "loss": 0.9699,
+      "step": 13516
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928193546995906e-05,
+      "loss": 0.971,
+      "step": 13517
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4927451509753294e-05,
+      "loss": 0.8645,
+      "step": 13518
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4926709436677709e-05,
+      "loss": 1.0328,
+      "step": 13519
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925967327774544e-05,
+      "loss": 0.9454,
+      "step": 13520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.49252251830492e-05,
+      "loss": 1.2526,
+      "step": 13521
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4924483002507071e-05,
+      "loss": 0.9056,
+      "step": 13522
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.492374078615355e-05,
+      "loss": 0.6859,
+      "step": 13523
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4922998533994045e-05,
+      "loss": 0.8285,
+      "step": 13524
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4922256246033946e-05,
+      "loss": 1.0216,
+      "step": 13525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4921513922278651e-05,
+      "loss": 0.7813,
+      "step": 13526
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920771562733566e-05,
+      "loss": 1.0095,
+      "step": 13527
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920029167404077e-05,
+      "loss": 0.9666,
+      "step": 13528
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4919286736295593e-05,
+      "loss": 0.9634,
+      "step": 13529
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4918544269413511e-05,
+      "loss": 1.1583,
+      "step": 13530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4917801766763227e-05,
+      "loss": 1.0276,
+      "step": 13531
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491705922835014e-05,
+      "loss": 0.7301,
+      "step": 13532
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4916316654179656e-05,
+      "loss": 0.7291,
+      "step": 13533
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4915574044257169e-05,
+      "loss": 1.3976,
+      "step": 13534
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4914831398588084e-05,
+      "loss": 1.0403,
+      "step": 13535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4914088717177798e-05,
+      "loss": 1.25,
+      "step": 13536
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4913346000031714e-05,
+      "loss": 0.9058,
+      "step": 13537
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4912603247155229e-05,
+      "loss": 1.0218,
+      "step": 13538
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491186045855375e-05,
+      "loss": 0.9809,
+      "step": 13539
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4911117634232679e-05,
+      "loss": 0.9983,
+      "step": 13540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4910374774197412e-05,
+      "loss": 0.9573,
+      "step": 13541
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909631878453355e-05,
+      "loss": 1.1815,
+      "step": 13542
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908888947005911e-05,
+      "loss": 1.0287,
+      "step": 13543
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908145979860478e-05,
+      "loss": 0.9274,
+      "step": 13544
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4907402977022465e-05,
+      "loss": 0.9084,
+      "step": 13545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4906659938497274e-05,
+      "loss": 0.9407,
+      "step": 13546
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4905916864290301e-05,
+      "loss": 0.818,
+      "step": 13547
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490517375440696e-05,
+      "loss": 0.7895,
+      "step": 13548
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4904430608852654e-05,
+      "loss": 0.7408,
+      "step": 13549
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490368742763278e-05,
+      "loss": 0.993,
+      "step": 13550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4902944210752747e-05,
+      "loss": 0.9564,
+      "step": 13551
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490220095821796e-05,
+      "loss": 0.8153,
+      "step": 13552
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490145767003382e-05,
+      "loss": 0.8378,
+      "step": 13553
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490071434620574e-05,
+      "loss": 0.8306,
+      "step": 13554
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489997098673912e-05,
+      "loss": 0.7615,
+      "step": 13555
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4899227591639366e-05,
+      "loss": 1.0891,
+      "step": 13556
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4898484160911885e-05,
+      "loss": 0.809,
+      "step": 13557
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897740694562084e-05,
+      "loss": 0.9928,
+      "step": 13558
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4896997192595367e-05,
+      "loss": 0.7981,
+      "step": 13559
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4896253655017146e-05,
+      "loss": 0.8351,
+      "step": 13560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4895510081832823e-05,
+      "loss": 1.1681,
+      "step": 13561
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4894766473047803e-05,
+      "loss": 0.8621,
+      "step": 13562
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4894022828667503e-05,
+      "loss": 0.7435,
+      "step": 13563
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4893279148697324e-05,
+      "loss": 1.2308,
+      "step": 13564
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4892535433142674e-05,
+      "loss": 0.5288,
+      "step": 13565
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4891791682008965e-05,
+      "loss": 1.1596,
+      "step": 13566
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4891047895301603e-05,
+      "loss": 0.9489,
+      "step": 13567
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4890304073025997e-05,
+      "loss": 0.9572,
+      "step": 13568
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4889560215187558e-05,
+      "loss": 0.9609,
+      "step": 13569
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888816321791695e-05,
+      "loss": 0.6884,
+      "step": 13570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888072392843814e-05,
+      "loss": 1.2198,
+      "step": 13571
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4887328428349327e-05,
+      "loss": 1.0205,
+      "step": 13572
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4886584428313648e-05,
+      "loss": 0.9181,
+      "step": 13573
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885840392742183e-05,
+      "loss": 0.9712,
+      "step": 13574
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885096321640346e-05,
+      "loss": 0.7919,
+      "step": 13575
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4884352215013545e-05,
+      "loss": 1.2618,
+      "step": 13576
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4883608072867191e-05,
+      "loss": 1.1734,
+      "step": 13577
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4882863895206696e-05,
+      "loss": 1.256,
+      "step": 13578
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4882119682037476e-05,
+      "loss": 1.0541,
+      "step": 13579
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4881375433364937e-05,
+      "loss": 1.1823,
+      "step": 13580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4880631149194495e-05,
+      "loss": 0.8052,
+      "step": 13581
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487988682953156e-05,
+      "loss": 1.2143,
+      "step": 13582
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4879142474381547e-05,
+      "loss": 1.4827,
+      "step": 13583
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4878398083749867e-05,
+      "loss": 1.2941,
+      "step": 13584
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4877653657641936e-05,
+      "loss": 0.8162,
+      "step": 13585
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4876909196063165e-05,
+      "loss": 0.6998,
+      "step": 13586
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487616469901897e-05,
+      "loss": 1.1704,
+      "step": 13587
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4875420166514763e-05,
+      "loss": 0.782,
+      "step": 13588
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487467559855596e-05,
+      "loss": 0.8048,
+      "step": 13589
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4873930995147971e-05,
+      "loss": 0.7243,
+      "step": 13590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4873186356296219e-05,
+      "loss": 1.0956,
+      "step": 13591
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4872441682006113e-05,
+      "loss": 0.9724,
+      "step": 13592
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487169697228307e-05,
+      "loss": 1.1942,
+      "step": 13593
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4870952227132507e-05,
+      "loss": 0.9444,
+      "step": 13594
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4870207446559835e-05,
+      "loss": 0.7273,
+      "step": 13595
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4869462630570474e-05,
+      "loss": 0.8366,
+      "step": 13596
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4868717779169844e-05,
+      "loss": 1.256,
+      "step": 13597
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867972892363356e-05,
+      "loss": 0.8401,
+      "step": 13598
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4867227970156428e-05,
+      "loss": 0.976,
+      "step": 13599
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4866483012554478e-05,
+      "loss": 0.925,
+      "step": 13600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4865738019562923e-05,
+      "loss": 0.7997,
+      "step": 13601
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4864992991187178e-05,
+      "loss": 0.732,
+      "step": 13602
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4864247927432667e-05,
+      "loss": 0.7571,
+      "step": 13603
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4863502828304803e-05,
+      "loss": 0.7121,
+      "step": 13604
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4862757693809009e-05,
+      "loss": 1.2019,
+      "step": 13605
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48620125239507e-05,
+      "loss": 0.8663,
+      "step": 13606
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4861267318735295e-05,
+      "loss": 1.193,
+      "step": 13607
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4860522078168213e-05,
+      "loss": 1.1269,
+      "step": 13608
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485977680225488e-05,
+      "loss": 1.076,
+      "step": 13609
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4859031491000706e-05,
+      "loss": 1.0554,
+      "step": 13610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4858286144411116e-05,
+      "loss": 0.796,
+      "step": 13611
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485754076249153e-05,
+      "loss": 1.0823,
+      "step": 13612
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4856795345247368e-05,
+      "loss": 1.1221,
+      "step": 13613
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4856049892684053e-05,
+      "loss": 0.864,
+      "step": 13614
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4855304404807e-05,
+      "loss": 1.0356,
+      "step": 13615
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4854558881621636e-05,
+      "loss": 0.9426,
+      "step": 13616
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4853813323133382e-05,
+      "loss": 1.0596,
+      "step": 13617
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4853067729347656e-05,
+      "loss": 0.7529,
+      "step": 13618
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4852322100269887e-05,
+      "loss": 1.0228,
+      "step": 13619
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4851576435905489e-05,
+      "loss": 1.0847,
+      "step": 13620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485083073625989e-05,
+      "loss": 0.9035,
+      "step": 13621
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485008500133851e-05,
+      "loss": 1.1214,
+      "step": 13622
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484933923114677e-05,
+      "loss": 1.1256,
+      "step": 13623
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4848593425690104e-05,
+      "loss": 1.0316,
+      "step": 13624
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4847847584973924e-05,
+      "loss": 0.8479,
+      "step": 13625
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4847101709003656e-05,
+      "loss": 0.8448,
+      "step": 13626
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484635579778473e-05,
+      "loss": 1.0723,
+      "step": 13627
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4845609851322566e-05,
+      "loss": 0.9718,
+      "step": 13628
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844863869622587e-05,
+      "loss": 0.7846,
+      "step": 13629
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844117852690223e-05,
+      "loss": 0.8462,
+      "step": 13630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4843371800530894e-05,
+      "loss": 0.777,
+      "step": 13631
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4842625713150028e-05,
+      "loss": 0.9685,
+      "step": 13632
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841879590553048e-05,
+      "loss": 1.2023,
+      "step": 13633
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841133432745385e-05,
+      "loss": 0.5959,
+      "step": 13634
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484038723973246e-05,
+      "loss": 1.1665,
+      "step": 13635
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4839641011519704e-05,
+      "loss": 0.8886,
+      "step": 13636
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4838894748112538e-05,
+      "loss": 1.025,
+      "step": 13637
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4838148449516391e-05,
+      "loss": 1.4395,
+      "step": 13638
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4837402115736696e-05,
+      "loss": 0.8221,
+      "step": 13639
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4836655746778875e-05,
+      "loss": 1.005,
+      "step": 13640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4835909342648352e-05,
+      "loss": 1.1849,
+      "step": 13641
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4835162903350561e-05,
+      "loss": 1.1159,
+      "step": 13642
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483441642889093e-05,
+      "loss": 1.0766,
+      "step": 13643
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4833669919274885e-05,
+      "loss": 0.8609,
+      "step": 13644
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4832923374507857e-05,
+      "loss": 1.0331,
+      "step": 13645
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4832176794595271e-05,
+      "loss": 0.5757,
+      "step": 13646
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483143017954256e-05,
+      "loss": 0.7927,
+      "step": 13647
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4830683529355152e-05,
+      "loss": 1.1778,
+      "step": 13648
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482993684403848e-05,
+      "loss": 0.6933,
+      "step": 13649
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4829190123597965e-05,
+      "loss": 0.9612,
+      "step": 13650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4828443368039047e-05,
+      "loss": 0.6264,
+      "step": 13651
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4827696577367152e-05,
+      "loss": 1.0776,
+      "step": 13652
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4826949751587712e-05,
+      "loss": 0.9994,
+      "step": 13653
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4826202890706157e-05,
+      "loss": 0.848,
+      "step": 13654
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4825455994727918e-05,
+      "loss": 0.9336,
+      "step": 13655
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4824709063658428e-05,
+      "loss": 0.9682,
+      "step": 13656
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4823962097503115e-05,
+      "loss": 0.9594,
+      "step": 13657
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4823215096267415e-05,
+      "loss": 1.0179,
+      "step": 13658
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4822468059956763e-05,
+      "loss": 1.1769,
+      "step": 13659
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4821720988576585e-05,
+      "loss": 1.0526,
+      "step": 13660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4820973882132318e-05,
+      "loss": 1.0175,
+      "step": 13661
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482022674062939e-05,
+      "loss": 0.9918,
+      "step": 13662
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481947956407324e-05,
+      "loss": 0.9787,
+      "step": 13663
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4818732352469303e-05,
+      "loss": 0.7952,
+      "step": 13664
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4817985105823003e-05,
+      "loss": 1.1803,
+      "step": 13665
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4817237824139783e-05,
+      "loss": 0.9153,
+      "step": 13666
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4816490507425076e-05,
+      "loss": 1.0053,
+      "step": 13667
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4815743155684312e-05,
+      "loss": 0.8467,
+      "step": 13668
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4814995768922933e-05,
+      "loss": 0.9679,
+      "step": 13669
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4814248347146368e-05,
+      "loss": 1.1356,
+      "step": 13670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4813500890360053e-05,
+      "loss": 0.7563,
+      "step": 13671
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4812753398569427e-05,
+      "loss": 0.9412,
+      "step": 13672
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4812005871779921e-05,
+      "loss": 1.1132,
+      "step": 13673
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4811258309996976e-05,
+      "loss": 1.0009,
+      "step": 13674
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481051071322603e-05,
+      "loss": 0.8502,
+      "step": 13675
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.480976308147251e-05,
+      "loss": 0.8742,
+      "step": 13676
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4809015414741862e-05,
+      "loss": 1.0133,
+      "step": 13677
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4808267713039518e-05,
+      "loss": 0.8265,
+      "step": 13678
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4807519976370918e-05,
+      "loss": 1.1277,
+      "step": 13679
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4806772204741503e-05,
+      "loss": 1.0859,
+      "step": 13680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48060243981567e-05,
+      "loss": 0.839,
+      "step": 13681
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4805276556621956e-05,
+      "loss": 1.2755,
+      "step": 13682
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4804528680142708e-05,
+      "loss": 1.1986,
+      "step": 13683
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4803780768724396e-05,
+      "loss": 0.9708,
+      "step": 13684
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4803032822372453e-05,
+      "loss": 0.8416,
+      "step": 13685
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4802284841092324e-05,
+      "loss": 1.1297,
+      "step": 13686
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4801536824889447e-05,
+      "loss": 0.6545,
+      "step": 13687
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4800788773769259e-05,
+      "loss": 1.0475,
+      "step": 13688
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4800040687737205e-05,
+      "loss": 0.8173,
+      "step": 13689
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4799292566798718e-05,
+      "loss": 0.5425,
+      "step": 13690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4798544410959243e-05,
+      "loss": 1.1035,
+      "step": 13691
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797796220224223e-05,
+      "loss": 1.1487,
+      "step": 13692
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797047994599095e-05,
+      "loss": 0.9695,
+      "step": 13693
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4796299734089301e-05,
+      "loss": 0.7666,
+      "step": 13694
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4795551438700283e-05,
+      "loss": 1.0904,
+      "step": 13695
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.479480310843748e-05,
+      "loss": 1.0831,
+      "step": 13696
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4794054743306342e-05,
+      "loss": 1.2149,
+      "step": 13697
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4793306343312302e-05,
+      "loss": 0.9239,
+      "step": 13698
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4792557908460805e-05,
+      "loss": 0.9741,
+      "step": 13699
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4791809438757296e-05,
+      "loss": 1.1775,
+      "step": 13700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4791060934207217e-05,
+      "loss": 0.9896,
+      "step": 13701
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4790312394816008e-05,
+      "loss": 1.3477,
+      "step": 13702
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478956382058912e-05,
+      "loss": 0.8934,
+      "step": 13703
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478881521153199e-05,
+      "loss": 1.3264,
+      "step": 13704
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4788066567650062e-05,
+      "loss": 0.8405,
+      "step": 13705
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4787317888948784e-05,
+      "loss": 0.8395,
+      "step": 13706
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4786569175433598e-05,
+      "loss": 0.9055,
+      "step": 13707
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785820427109951e-05,
+      "loss": 1.2498,
+      "step": 13708
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785071643983284e-05,
+      "loss": 0.7611,
+      "step": 13709
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4784322826059048e-05,
+      "loss": 0.994,
+      "step": 13710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783573973342681e-05,
+      "loss": 0.9338,
+      "step": 13711
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4782825085839635e-05,
+      "loss": 0.9618,
+      "step": 13712
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4782076163555355e-05,
+      "loss": 0.9604,
+      "step": 13713
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4781327206495285e-05,
+      "loss": 0.8695,
+      "step": 13714
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4780578214664871e-05,
+      "loss": 1.0631,
+      "step": 13715
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4779829188069563e-05,
+      "loss": 0.7515,
+      "step": 13716
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4779080126714807e-05,
+      "loss": 1.0356,
+      "step": 13717
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4778331030606048e-05,
+      "loss": 1.0637,
+      "step": 13718
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477758189974874e-05,
+      "loss": 0.9838,
+      "step": 13719
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4776832734148319e-05,
+      "loss": 0.9053,
+      "step": 13720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4776083533810242e-05,
+      "loss": 0.7967,
+      "step": 13721
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4775334298739956e-05,
+      "loss": 0.9964,
+      "step": 13722
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4774585028942909e-05,
+      "loss": 0.9778,
+      "step": 13723
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477383572442455e-05,
+      "loss": 1.0036,
+      "step": 13724
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4773086385190328e-05,
+      "loss": 1.1684,
+      "step": 13725
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477233701124569e-05,
+      "loss": 1.2655,
+      "step": 13726
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4771587602596085e-05,
+      "loss": 1.2409,
+      "step": 13727
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770838159246968e-05,
+      "loss": 0.7864,
+      "step": 13728
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770088681203786e-05,
+      "loss": 1.2625,
+      "step": 13729
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.476933916847199e-05,
+      "loss": 0.9695,
+      "step": 13730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4768589621057028e-05,
+      "loss": 1.0112,
+      "step": 13731
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4767840038964356e-05,
+      "loss": 1.1477,
+      "step": 13732
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4767090422199418e-05,
+      "loss": 1.2774,
+      "step": 13733
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4766340770767672e-05,
+      "loss": 0.9925,
+      "step": 13734
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4765591084674567e-05,
+      "loss": 1.1342,
+      "step": 13735
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764841363925551e-05,
+      "loss": 0.884,
+      "step": 13736
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764091608526081e-05,
+      "loss": 1.1141,
+      "step": 13737
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4763341818481606e-05,
+      "loss": 1.0532,
+      "step": 13738
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4762591993797584e-05,
+      "loss": 1.1958,
+      "step": 13739
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4761842134479463e-05,
+      "loss": 1.1461,
+      "step": 13740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4761092240532698e-05,
+      "loss": 0.9554,
+      "step": 13741
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4760342311962739e-05,
+      "loss": 1.3325,
+      "step": 13742
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4759592348775042e-05,
+      "loss": 1.2579,
+      "step": 13743
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4758842350975064e-05,
+      "loss": 0.7617,
+      "step": 13744
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4758092318568253e-05,
+      "loss": 0.8887,
+      "step": 13745
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475734225156007e-05,
+      "loss": 0.9207,
+      "step": 13746
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4756592149955962e-05,
+      "loss": 0.839,
+      "step": 13747
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755842013761387e-05,
+      "loss": 1.0665,
+      "step": 13748
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755091842981806e-05,
+      "loss": 0.861,
+      "step": 13749
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4754341637622665e-05,
+      "loss": 0.7893,
+      "step": 13750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4753591397689424e-05,
+      "loss": 1.1704,
+      "step": 13751
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752841123187538e-05,
+      "loss": 1.2077,
+      "step": 13752
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752090814122466e-05,
+      "loss": 0.9941,
+      "step": 13753
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475134047049966e-05,
+      "loss": 1.0112,
+      "step": 13754
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4750590092324579e-05,
+      "loss": 0.9836,
+      "step": 13755
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474983967960268e-05,
+      "loss": 0.9602,
+      "step": 13756
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4749089232339417e-05,
+      "loss": 1.4163,
+      "step": 13757
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4748338750540251e-05,
+      "loss": 0.9986,
+      "step": 13758
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474758823421064e-05,
+      "loss": 0.8673,
+      "step": 13759
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474683768335604e-05,
+      "loss": 0.9324,
+      "step": 13760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4746087097981907e-05,
+      "loss": 1.1801,
+      "step": 13761
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4745336478093704e-05,
+      "loss": 0.6769,
+      "step": 13762
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4744585823696885e-05,
+      "loss": 0.9336,
+      "step": 13763
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4743835134796913e-05,
+      "loss": 0.9814,
+      "step": 13764
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4743084411399245e-05,
+      "loss": 0.8318,
+      "step": 13765
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4742333653509336e-05,
+      "loss": 0.7821,
+      "step": 13766
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4741582861132655e-05,
+      "loss": 0.9283,
+      "step": 13767
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4740832034274655e-05,
+      "loss": 1.4873,
+      "step": 13768
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4740081172940798e-05,
+      "loss": 1.3927,
+      "step": 13769
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4739330277136546e-05,
+      "loss": 1.0144,
+      "step": 13770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4738579346867357e-05,
+      "loss": 1.1302,
+      "step": 13771
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473782838213869e-05,
+      "loss": 0.9452,
+      "step": 13772
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4737077382956012e-05,
+      "loss": 1.0616,
+      "step": 13773
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4736326349324783e-05,
+      "loss": 1.0412,
+      "step": 13774
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4735575281250458e-05,
+      "loss": 0.7414,
+      "step": 13775
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4734824178738508e-05,
+      "loss": 0.8399,
+      "step": 13776
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4734073041794387e-05,
+      "loss": 1.3133,
+      "step": 13777
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4733321870423565e-05,
+      "loss": 1.0965,
+      "step": 13778
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4732570664631498e-05,
+      "loss": 0.9318,
+      "step": 13779
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4731819424423653e-05,
+      "loss": 1.0314,
+      "step": 13780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473106814980549e-05,
+      "loss": 0.9826,
+      "step": 13781
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4730316840782473e-05,
+      "loss": 1.1474,
+      "step": 13782
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472956549736007e-05,
+      "loss": 1.2319,
+      "step": 13783
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728814119543739e-05,
+      "loss": 0.8201,
+      "step": 13784
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728062707338949e-05,
+      "loss": 1.0012,
+      "step": 13785
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472731126075116e-05,
+      "loss": 1.1738,
+      "step": 13786
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4726559779785841e-05,
+      "loss": 1.2589,
+      "step": 13787
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472580826444845e-05,
+      "loss": 0.8978,
+      "step": 13788
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4725056714744463e-05,
+      "loss": 0.9371,
+      "step": 13789
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4724305130679333e-05,
+      "loss": 1.1331,
+      "step": 13790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4723553512258532e-05,
+      "loss": 1.038,
+      "step": 13791
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722801859487528e-05,
+      "loss": 0.8405,
+      "step": 13792
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722050172371785e-05,
+      "loss": 0.8969,
+      "step": 13793
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4721298450916765e-05,
+      "loss": 0.8798,
+      "step": 13794
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4720546695127942e-05,
+      "loss": 0.9626,
+      "step": 13795
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4719794905010778e-05,
+      "loss": 1.1748,
+      "step": 13796
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4719043080570738e-05,
+      "loss": 1.0555,
+      "step": 13797
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4718291221813297e-05,
+      "loss": 0.7475,
+      "step": 13798
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4717539328743916e-05,
+      "loss": 0.9039,
+      "step": 13799
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4716787401368067e-05,
+      "loss": 1.2272,
+      "step": 13800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4716035439691217e-05,
+      "loss": 1.0282,
+      "step": 13801
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471528344371883e-05,
+      "loss": 0.9612,
+      "step": 13802
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471453141345638e-05,
+      "loss": 1.0225,
+      "step": 13803
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713779348909336e-05,
+      "loss": 1.2312,
+      "step": 13804
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713027250083162e-05,
+      "loss": 1.3792,
+      "step": 13805
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.471227511698333e-05,
+      "loss": 0.8344,
+      "step": 13806
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4711522949615313e-05,
+      "loss": 1.0977,
+      "step": 13807
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4710770747984578e-05,
+      "loss": 0.6517,
+      "step": 13808
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4710018512096593e-05,
+      "loss": 0.8871,
+      "step": 13809
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4709266241956832e-05,
+      "loss": 1.0198,
+      "step": 13810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4708513937570763e-05,
+      "loss": 0.9591,
+      "step": 13811
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4707761598943855e-05,
+      "loss": 1.0991,
+      "step": 13812
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4707009226081586e-05,
+      "loss": 0.9266,
+      "step": 13813
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4706256818989424e-05,
+      "loss": 1.0201,
+      "step": 13814
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4705504377672834e-05,
+      "loss": 0.7198,
+      "step": 13815
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4704751902137297e-05,
+      "loss": 0.8745,
+      "step": 13816
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4703999392388285e-05,
+      "loss": 1.2156,
+      "step": 13817
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470324684843126e-05,
+      "loss": 1.0343,
+      "step": 13818
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.470249427027171e-05,
+      "loss": 1.081,
+      "step": 13819
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4701741657915094e-05,
+      "loss": 0.9196,
+      "step": 13820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4700989011366891e-05,
+      "loss": 1.0599,
+      "step": 13821
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4700236330632574e-05,
+      "loss": 0.8525,
+      "step": 13822
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4699483615717618e-05,
+      "loss": 1.0695,
+      "step": 13823
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4698730866627493e-05,
+      "loss": 1.0514,
+      "step": 13824
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4697978083367677e-05,
+      "loss": 1.0102,
+      "step": 13825
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4697225265943641e-05,
+      "loss": 0.8825,
+      "step": 13826
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4696472414360863e-05,
+      "loss": 0.6848,
+      "step": 13827
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4695719528624817e-05,
+      "loss": 0.8409,
+      "step": 13828
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4694966608740978e-05,
+      "loss": 0.9125,
+      "step": 13829
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4694213654714816e-05,
+      "loss": 1.027,
+      "step": 13830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4693460666551813e-05,
+      "loss": 0.6814,
+      "step": 13831
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4692707644257442e-05,
+      "loss": 0.8389,
+      "step": 13832
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4691954587837181e-05,
+      "loss": 1.006,
+      "step": 13833
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4691201497296508e-05,
+      "loss": 0.9266,
+      "step": 13834
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4690448372640894e-05,
+      "loss": 1.0197,
+      "step": 13835
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4689695213875816e-05,
+      "loss": 1.0942,
+      "step": 13836
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4688942021006755e-05,
+      "loss": 0.9128,
+      "step": 13837
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4688188794039188e-05,
+      "loss": 0.742,
+      "step": 13838
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4687435532978594e-05,
+      "loss": 0.9214,
+      "step": 13839
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4686682237830445e-05,
+      "loss": 1.1994,
+      "step": 13840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4685928908600224e-05,
+      "loss": 0.7398,
+      "step": 13841
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4685175545293404e-05,
+      "loss": 0.9068,
+      "step": 13842
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4684422147915471e-05,
+      "loss": 1.1251,
+      "step": 13843
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.46836687164719e-05,
+      "loss": 0.8795,
+      "step": 13844
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4682915250968169e-05,
+      "loss": 1.1429,
+      "step": 13845
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4682161751409758e-05,
+      "loss": 0.8664,
+      "step": 13846
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4681408217802147e-05,
+      "loss": 0.7534,
+      "step": 13847
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680654650150814e-05,
+      "loss": 1.0192,
+      "step": 13848
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4679901048461243e-05,
+      "loss": 0.9509,
+      "step": 13849
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4679147412738913e-05,
+      "loss": 0.7275,
+      "step": 13850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.46783937429893e-05,
+      "loss": 1.1387,
+      "step": 13851
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4677640039217889e-05,
+      "loss": 0.81,
+      "step": 13852
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4676886301430159e-05,
+      "loss": 0.9329,
+      "step": 13853
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4676132529631595e-05,
+      "loss": 0.9864,
+      "step": 13854
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4675378723827675e-05,
+      "loss": 0.7538,
+      "step": 13855
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4674624884023883e-05,
+      "loss": 0.6724,
+      "step": 13856
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.46738710102257e-05,
+      "loss": 0.7375,
+      "step": 13857
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4673117102438606e-05,
+      "loss": 1.0238,
+      "step": 13858
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4672363160668087e-05,
+      "loss": 1.1628,
+      "step": 13859
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4671609184919622e-05,
+      "loss": 0.841,
+      "step": 13860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4670855175198698e-05,
+      "loss": 1.087,
+      "step": 13861
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4670101131510796e-05,
+      "loss": 0.9643,
+      "step": 13862
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4669347053861401e-05,
+      "loss": 1.1693,
+      "step": 13863
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4668592942255995e-05,
+      "loss": 0.9067,
+      "step": 13864
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667838796700062e-05,
+      "loss": 1.14,
+      "step": 13865
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4667084617199088e-05,
+      "loss": 0.7237,
+      "step": 13866
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4666330403758555e-05,
+      "loss": 0.8952,
+      "step": 13867
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466557615638395e-05,
+      "loss": 0.8328,
+      "step": 13868
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4664821875080757e-05,
+      "loss": 1.4228,
+      "step": 13869
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466406755985446e-05,
+      "loss": 1.2202,
+      "step": 13870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4663313210710548e-05,
+      "loss": 1.1334,
+      "step": 13871
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4662558827654505e-05,
+      "loss": 1.048,
+      "step": 13872
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661804410691814e-05,
+      "loss": 0.862,
+      "step": 13873
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661049959827964e-05,
+      "loss": 1.0588,
+      "step": 13874
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4660295475068443e-05,
+      "loss": 0.9903,
+      "step": 13875
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4659540956418731e-05,
+      "loss": 0.6786,
+      "step": 13876
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4658786403884326e-05,
+      "loss": 0.8602,
+      "step": 13877
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4658031817470705e-05,
+      "loss": 0.7633,
+      "step": 13878
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4657277197183361e-05,
+      "loss": 1.4441,
+      "step": 13879
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465652254302778e-05,
+      "loss": 0.6615,
+      "step": 13880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465576785500945e-05,
+      "loss": 0.9889,
+      "step": 13881
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4655013133133857e-05,
+      "loss": 0.7017,
+      "step": 13882
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4654258377406493e-05,
+      "loss": 1.1902,
+      "step": 13883
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4653503587832848e-05,
+      "loss": 1.0365,
+      "step": 13884
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4652748764418404e-05,
+      "loss": 0.9333,
+      "step": 13885
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4651993907168658e-05,
+      "loss": 0.6041,
+      "step": 13886
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4651239016089094e-05,
+      "loss": 0.83,
+      "step": 13887
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46504840911852e-05,
+      "loss": 1.0464,
+      "step": 13888
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649729132462476e-05,
+      "loss": 0.7502,
+      "step": 13889
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4648974139926403e-05,
+      "loss": 1.1924,
+      "step": 13890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464821911358247e-05,
+      "loss": 0.8377,
+      "step": 13891
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647464053436173e-05,
+      "loss": 1.3175,
+      "step": 13892
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4646708959493004e-05,
+      "loss": 0.7581,
+      "step": 13893
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4645953831758448e-05,
+      "loss": 1.0964,
+      "step": 13894
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4645198670238e-05,
+      "loss": 1.0155,
+      "step": 13895
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4644443474937156e-05,
+      "loss": 0.7972,
+      "step": 13896
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4643688245861398e-05,
+      "loss": 0.8916,
+      "step": 13897
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4642932983016227e-05,
+      "loss": 1.0135,
+      "step": 13898
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464217768640713e-05,
+      "loss": 1.2296,
+      "step": 13899
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4641422356039603e-05,
+      "loss": 0.7564,
+      "step": 13900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4640666991919135e-05,
+      "loss": 1.0217,
+      "step": 13901
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4639911594051221e-05,
+      "loss": 0.8826,
+      "step": 13902
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4639156162441356e-05,
+      "loss": 0.89,
+      "step": 13903
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4638400697095034e-05,
+      "loss": 0.8781,
+      "step": 13904
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4637645198017745e-05,
+      "loss": 0.8485,
+      "step": 13905
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4636889665214984e-05,
+      "loss": 1.1042,
+      "step": 13906
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4636134098692249e-05,
+      "loss": 1.2209,
+      "step": 13907
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463537849845503e-05,
+      "loss": 0.9632,
+      "step": 13908
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4634622864508825e-05,
+      "loss": 0.9392,
+      "step": 13909
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463386719685913e-05,
+      "loss": 0.8899,
+      "step": 13910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4633111495511434e-05,
+      "loss": 1.143,
+      "step": 13911
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4632355760471238e-05,
+      "loss": 1.3624,
+      "step": 13912
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4631599991744037e-05,
+      "loss": 0.8201,
+      "step": 13913
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630844189335327e-05,
+      "loss": 1.2254,
+      "step": 13914
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630088353250603e-05,
+      "loss": 0.8138,
+      "step": 13915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4629332483495363e-05,
+      "loss": 1.2239,
+      "step": 13916
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4628576580075102e-05,
+      "loss": 1.0717,
+      "step": 13917
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4627820642995318e-05,
+      "loss": 0.894,
+      "step": 13918
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4627064672261511e-05,
+      "loss": 1.0459,
+      "step": 13919
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4626308667879175e-05,
+      "loss": 0.8302,
+      "step": 13920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4625552629853803e-05,
+      "loss": 1.0339,
+      "step": 13921
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4624796558190904e-05,
+      "loss": 1.016,
+      "step": 13922
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4624040452895971e-05,
+      "loss": 0.9732,
+      "step": 13923
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4623284313974502e-05,
+      "loss": 0.9266,
+      "step": 13924
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4622528141431996e-05,
+      "loss": 1.251,
+      "step": 13925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4621771935273951e-05,
+      "loss": 0.8931,
+      "step": 13926
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4621015695505869e-05,
+      "loss": 0.8759,
+      "step": 13927
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4620259422133244e-05,
+      "loss": 0.915,
+      "step": 13928
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4619503115161584e-05,
+      "loss": 0.901,
+      "step": 13929
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4618746774596382e-05,
+      "loss": 0.897,
+      "step": 13930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4617990400443141e-05,
+      "loss": 0.877,
+      "step": 13931
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461723399270736e-05,
+      "loss": 0.8147,
+      "step": 13932
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4616477551394544e-05,
+      "loss": 0.8708,
+      "step": 13933
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4615721076510188e-05,
+      "loss": 1.3497,
+      "step": 13934
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4614964568059795e-05,
+      "loss": 0.8514,
+      "step": 13935
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4614208026048869e-05,
+      "loss": 0.9487,
+      "step": 13936
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4613451450482905e-05,
+      "loss": 1.4771,
+      "step": 13937
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4612694841367417e-05,
+      "loss": 1.1879,
+      "step": 13938
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611938198707897e-05,
+      "loss": 0.8024,
+      "step": 13939
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611181522509847e-05,
+      "loss": 0.8304,
+      "step": 13940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4610424812778776e-05,
+      "loss": 0.7513,
+      "step": 13941
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4609668069520184e-05,
+      "loss": 0.9874,
+      "step": 13942
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4608911292739572e-05,
+      "loss": 0.7021,
+      "step": 13943
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.460815448244245e-05,
+      "loss": 0.6613,
+      "step": 13944
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4607397638634313e-05,
+      "loss": 0.7139,
+      "step": 13945
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4606640761320668e-05,
+      "loss": 1.1696,
+      "step": 13946
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4605883850507021e-05,
+      "loss": 0.9828,
+      "step": 13947
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4605126906198876e-05,
+      "loss": 0.6581,
+      "step": 13948
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4604369928401736e-05,
+      "loss": 0.855,
+      "step": 13949
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4603612917121107e-05,
+      "loss": 1.2294,
+      "step": 13950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4602855872362494e-05,
+      "loss": 0.9735,
+      "step": 13951
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.46020987941314e-05,
+      "loss": 0.8353,
+      "step": 13952
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4601341682433332e-05,
+      "loss": 1.0651,
+      "step": 13953
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4600584537273801e-05,
+      "loss": 1.2922,
+      "step": 13954
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4599827358658303e-05,
+      "loss": 0.5432,
+      "step": 13955
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4599070146592353e-05,
+      "loss": 1.1949,
+      "step": 13956
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4598312901081452e-05,
+      "loss": 0.9387,
+      "step": 13957
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459755562213111e-05,
+      "loss": 0.8464,
+      "step": 13958
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4596798309746835e-05,
+      "loss": 0.9351,
+      "step": 13959
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4596040963934131e-05,
+      "loss": 1.0015,
+      "step": 13960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595283584698504e-05,
+      "loss": 1.3769,
+      "step": 13961
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4594526172045466e-05,
+      "loss": 0.7703,
+      "step": 13962
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4593768725980526e-05,
+      "loss": 1.2138,
+      "step": 13963
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4593011246509187e-05,
+      "loss": 0.7982,
+      "step": 13964
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4592253733636961e-05,
+      "loss": 1.0232,
+      "step": 13965
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4591496187369356e-05,
+      "loss": 0.8972,
+      "step": 13966
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459073860771188e-05,
+      "loss": 1.025,
+      "step": 13967
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4589980994670044e-05,
+      "loss": 1.2679,
+      "step": 13968
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4589223348249357e-05,
+      "loss": 1.04,
+      "step": 13969
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4588465668455327e-05,
+      "loss": 0.7381,
+      "step": 13970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4587707955293467e-05,
+      "loss": 1.0111,
+      "step": 13971
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4586950208769284e-05,
+      "loss": 0.7664,
+      "step": 13972
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4586192428888288e-05,
+      "loss": 0.9523,
+      "step": 13973
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4585434615655993e-05,
+      "loss": 1.047,
+      "step": 13974
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.458467676907791e-05,
+      "loss": 0.9682,
+      "step": 13975
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4583918889159545e-05,
+      "loss": 0.9406,
+      "step": 13976
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4583160975906413e-05,
+      "loss": 0.5998,
+      "step": 13977
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4582403029324028e-05,
+      "loss": 1.0684,
+      "step": 13978
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4581645049417899e-05,
+      "loss": 0.7466,
+      "step": 13979
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4580887036193539e-05,
+      "loss": 1.0372,
+      "step": 13980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4580128989656456e-05,
+      "loss": 0.8785,
+      "step": 13981
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4579370909812172e-05,
+      "loss": 1.1581,
+      "step": 13982
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4578612796666187e-05,
+      "loss": 1.1563,
+      "step": 13983
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4577854650224028e-05,
+      "loss": 0.8779,
+      "step": 13984
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4577096470491197e-05,
+      "loss": 1.1007,
+      "step": 13985
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4576338257473215e-05,
+      "loss": 0.9459,
+      "step": 13986
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4575580011175592e-05,
+      "loss": 1.0084,
+      "step": 13987
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4574821731603843e-05,
+      "loss": 1.0792,
+      "step": 13988
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4574063418763484e-05,
+      "loss": 1.0732,
+      "step": 13989
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4573305072660025e-05,
+      "loss": 0.6917,
+      "step": 13990
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4572546693298985e-05,
+      "loss": 0.7422,
+      "step": 13991
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4571788280685877e-05,
+      "loss": 1.4833,
+      "step": 13992
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4571029834826217e-05,
+      "loss": 1.083,
+      "step": 13993
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4570271355725522e-05,
+      "loss": 0.929,
+      "step": 13994
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4569512843389306e-05,
+      "loss": 1.3772,
+      "step": 13995
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4568754297823084e-05,
+      "loss": 1.0593,
+      "step": 13996
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4567995719032376e-05,
+      "loss": 1.1062,
+      "step": 13997
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4567237107022689e-05,
+      "loss": 0.7411,
+      "step": 13998
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4566478461799556e-05,
+      "loss": 0.8961,
+      "step": 13999
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456571978336848e-05,
+      "loss": 1.0364,
+      "step": 14000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456496107173498e-05,
+      "loss": 1.1233,
+      "step": 14001
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4564202326904579e-05,
+      "loss": 1.055,
+      "step": 14002
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4563443548882792e-05,
+      "loss": 0.7811,
+      "step": 14003
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4562684737675137e-05,
+      "loss": 0.933,
+      "step": 14004
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456192589328713e-05,
+      "loss": 0.8679,
+      "step": 14005
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4561167015724293e-05,
+      "loss": 0.8785,
+      "step": 14006
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4560408104992142e-05,
+      "loss": 1.195,
+      "step": 14007
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4559649161096196e-05,
+      "loss": 0.7886,
+      "step": 14008
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4558890184041979e-05,
+      "loss": 0.9724,
+      "step": 14009
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4558131173835002e-05,
+      "loss": 0.778,
+      "step": 14010
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4557372130480791e-05,
+      "loss": 0.7521,
+      "step": 14011
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4556613053984863e-05,
+      "loss": 1.0658,
+      "step": 14012
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4555853944352737e-05,
+      "loss": 0.5968,
+      "step": 14013
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4555094801589942e-05,
+      "loss": 1.0139,
+      "step": 14014
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4554335625701988e-05,
+      "loss": 1.0021,
+      "step": 14015
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4553576416694394e-05,
+      "loss": 1.0915,
+      "step": 14016
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4552817174572693e-05,
+      "loss": 1.0327,
+      "step": 14017
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4552057899342398e-05,
+      "loss": 1.0031,
+      "step": 14018
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4551298591009033e-05,
+      "loss": 1.4705,
+      "step": 14019
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455053924957812e-05,
+      "loss": 1.1974,
+      "step": 14020
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4549779875055179e-05,
+      "loss": 0.7892,
+      "step": 14021
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4549020467445732e-05,
+      "loss": 1.384,
+      "step": 14022
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4548261026755305e-05,
+      "loss": 1.0956,
+      "step": 14023
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454750155298942e-05,
+      "loss": 0.8371,
+      "step": 14024
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4546742046153596e-05,
+      "loss": 0.8427,
+      "step": 14025
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454598250625336e-05,
+      "loss": 0.9911,
+      "step": 14026
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4545222933294237e-05,
+      "loss": 0.9433,
+      "step": 14027
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4544463327281743e-05,
+      "loss": 0.9567,
+      "step": 14028
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543703688221411e-05,
+      "loss": 1.013,
+      "step": 14029
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454294401611876e-05,
+      "loss": 1.0656,
+      "step": 14030
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4542184310979314e-05,
+      "loss": 0.9782,
+      "step": 14031
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4541424572808602e-05,
+      "loss": 0.6495,
+      "step": 14032
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4540664801612145e-05,
+      "loss": 1.0005,
+      "step": 14033
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4539904997395468e-05,
+      "loss": 1.1038,
+      "step": 14034
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45391451601641e-05,
+      "loss": 0.8445,
+      "step": 14035
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4538385289923564e-05,
+      "loss": 1.0601,
+      "step": 14036
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4537625386679383e-05,
+      "loss": 0.9472,
+      "step": 14037
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453686545043709e-05,
+      "loss": 1.4684,
+      "step": 14038
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4536105481202209e-05,
+      "loss": 0.7508,
+      "step": 14039
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453534547898026e-05,
+      "loss": 0.8397,
+      "step": 14040
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4534585443776779e-05,
+      "loss": 1.1411,
+      "step": 14041
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533825375597289e-05,
+      "loss": 0.9182,
+      "step": 14042
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533065274447315e-05,
+      "loss": 1.0487,
+      "step": 14043
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453230514033239e-05,
+      "loss": 0.8838,
+      "step": 14044
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4531544973258036e-05,
+      "loss": 0.9907,
+      "step": 14045
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4530784773229782e-05,
+      "loss": 0.9992,
+      "step": 14046
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4530024540253162e-05,
+      "loss": 0.8777,
+      "step": 14047
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45292642743337e-05,
+      "loss": 0.8148,
+      "step": 14048
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4528503975476925e-05,
+      "loss": 1.1736,
+      "step": 14049
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4527743643688366e-05,
+      "loss": 0.7188,
+      "step": 14050
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526983278973552e-05,
+      "loss": 1.1475,
+      "step": 14051
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526222881338012e-05,
+      "loss": 0.8742,
+      "step": 14052
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4525462450787278e-05,
+      "loss": 1.1826,
+      "step": 14053
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452470198732688e-05,
+      "loss": 1.1693,
+      "step": 14054
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4523941490962342e-05,
+      "loss": 0.6946,
+      "step": 14055
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4523180961699201e-05,
+      "loss": 1.3971,
+      "step": 14056
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4522420399542986e-05,
+      "loss": 1.0899,
+      "step": 14057
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452165980449923e-05,
+      "loss": 0.9101,
+      "step": 14058
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4520899176573461e-05,
+      "loss": 1.0046,
+      "step": 14059
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452013851577121e-05,
+      "loss": 0.9872,
+      "step": 14060
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451937782209801e-05,
+      "loss": 1.186,
+      "step": 14061
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451861709555939e-05,
+      "loss": 1.0017,
+      "step": 14062
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451785633616089e-05,
+      "loss": 1.0117,
+      "step": 14063
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4517095543908033e-05,
+      "loss": 0.7643,
+      "step": 14064
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4516334718806357e-05,
+      "loss": 0.911,
+      "step": 14065
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4515573860861394e-05,
+      "loss": 1.2582,
+      "step": 14066
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514812970078676e-05,
+      "loss": 1.0239,
+      "step": 14067
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514052046463736e-05,
+      "loss": 1.1098,
+      "step": 14068
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451329109002211e-05,
+      "loss": 1.1004,
+      "step": 14069
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451253010075933e-05,
+      "loss": 1.1271,
+      "step": 14070
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4511769078680927e-05,
+      "loss": 1.2572,
+      "step": 14071
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451100802379244e-05,
+      "loss": 1.1657,
+      "step": 14072
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4510246936099401e-05,
+      "loss": 1.0636,
+      "step": 14073
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4509485815607347e-05,
+      "loss": 0.9781,
+      "step": 14074
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4508724662321809e-05,
+      "loss": 1.2651,
+      "step": 14075
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4507963476248328e-05,
+      "loss": 0.85,
+      "step": 14076
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4507202257392433e-05,
+      "loss": 0.6665,
+      "step": 14077
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4506441005759662e-05,
+      "loss": 0.7444,
+      "step": 14078
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505679721355557e-05,
+      "loss": 0.9194,
+      "step": 14079
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4504918404185642e-05,
+      "loss": 0.7422,
+      "step": 14080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4504157054255464e-05,
+      "loss": 0.8761,
+      "step": 14081
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4503395671570555e-05,
+      "loss": 1.173,
+      "step": 14082
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450263425613645e-05,
+      "loss": 0.7921,
+      "step": 14083
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4501872807958695e-05,
+      "loss": 0.9544,
+      "step": 14084
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4501111327042817e-05,
+      "loss": 0.8433,
+      "step": 14085
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4500349813394357e-05,
+      "loss": 0.8183,
+      "step": 14086
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4499588267018854e-05,
+      "loss": 0.931,
+      "step": 14087
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4498826687921845e-05,
+      "loss": 0.9681,
+      "step": 14088
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449806507610887e-05,
+      "loss": 0.7695,
+      "step": 14089
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4497303431585466e-05,
+      "loss": 1.0715,
+      "step": 14090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449654175435717e-05,
+      "loss": 0.9677,
+      "step": 14091
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4495780044429524e-05,
+      "loss": 0.8757,
+      "step": 14092
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4495018301808066e-05,
+      "loss": 1.2505,
+      "step": 14093
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449425652649834e-05,
+      "loss": 0.7504,
+      "step": 14094
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4493494718505877e-05,
+      "loss": 0.9143,
+      "step": 14095
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4492732877836222e-05,
+      "loss": 1.0817,
+      "step": 14096
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4491971004494916e-05,
+      "loss": 0.8892,
+      "step": 14097
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4491209098487495e-05,
+      "loss": 0.9677,
+      "step": 14098
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4490447159819508e-05,
+      "loss": 0.7722,
+      "step": 14099
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4489685188496488e-05,
+      "loss": 0.8874,
+      "step": 14100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4488923184523975e-05,
+      "loss": 0.7392,
+      "step": 14101
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4488161147907517e-05,
+      "loss": 0.7199,
+      "step": 14102
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4487399078652656e-05,
+      "loss": 0.8805,
+      "step": 14103
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4486636976764925e-05,
+      "loss": 0.7342,
+      "step": 14104
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4485874842249876e-05,
+      "loss": 0.7135,
+      "step": 14105
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4485112675113046e-05,
+      "loss": 1.07,
+      "step": 14106
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4484350475359977e-05,
+      "loss": 0.8534,
+      "step": 14107
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483588242996214e-05,
+      "loss": 0.8609,
+      "step": 14108
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4482825978027302e-05,
+      "loss": 0.9773,
+      "step": 14109
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4482063680458777e-05,
+      "loss": 0.8881,
+      "step": 14110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448130135029619e-05,
+      "loss": 0.897,
+      "step": 14111
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4480538987545082e-05,
+      "loss": 1.0723,
+      "step": 14112
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4479776592210997e-05,
+      "loss": 1.1773,
+      "step": 14113
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4479014164299479e-05,
+      "loss": 0.69,
+      "step": 14114
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447825170381607e-05,
+      "loss": 0.5678,
+      "step": 14115
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4477489210766319e-05,
+      "loss": 0.7305,
+      "step": 14116
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447672668515577e-05,
+      "loss": 0.7101,
+      "step": 14117
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4475964126989968e-05,
+      "loss": 1.267,
+      "step": 14118
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4475201536274458e-05,
+      "loss": 0.7768,
+      "step": 14119
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4474438913014783e-05,
+      "loss": 0.9184,
+      "step": 14120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4473676257216494e-05,
+      "loss": 0.7818,
+      "step": 14121
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4472913568885134e-05,
+      "loss": 0.8357,
+      "step": 14122
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4472150848026248e-05,
+      "loss": 1.0523,
+      "step": 14123
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4471388094645387e-05,
+      "loss": 0.9873,
+      "step": 14124
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4470625308748092e-05,
+      "loss": 0.9479,
+      "step": 14125
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4469862490339914e-05,
+      "loss": 0.9245,
+      "step": 14126
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4469099639426402e-05,
+      "loss": 1.1359,
+      "step": 14127
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44683367560131e-05,
+      "loss": 0.7015,
+      "step": 14128
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4467573840105557e-05,
+      "loss": 0.7975,
+      "step": 14129
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446681089170932e-05,
+      "loss": 0.9153,
+      "step": 14130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4466047910829941e-05,
+      "loss": 1.0884,
+      "step": 14131
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.446528489747296e-05,
+      "loss": 0.7655,
+      "step": 14132
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4464521851643936e-05,
+      "loss": 1.1104,
+      "step": 14133
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4463758773348413e-05,
+      "loss": 0.8373,
+      "step": 14134
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4462995662591938e-05,
+      "loss": 0.8662,
+      "step": 14135
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4462232519380065e-05,
+      "loss": 0.9192,
+      "step": 14136
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4461469343718342e-05,
+      "loss": 0.8727,
+      "step": 14137
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4460706135612316e-05,
+      "loss": 0.9086,
+      "step": 14138
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4459942895067545e-05,
+      "loss": 0.9781,
+      "step": 14139
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445917962208957e-05,
+      "loss": 0.984,
+      "step": 14140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4458416316683943e-05,
+      "loss": 0.8526,
+      "step": 14141
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4457652978856223e-05,
+      "loss": 0.936,
+      "step": 14142
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4456889608611953e-05,
+      "loss": 0.9481,
+      "step": 14143
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4456126205956686e-05,
+      "loss": 0.8607,
+      "step": 14144
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4455362770895976e-05,
+      "loss": 0.8802,
+      "step": 14145
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4454599303435373e-05,
+      "loss": 0.8898,
+      "step": 14146
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4453835803580424e-05,
+      "loss": 0.5203,
+      "step": 14147
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4453072271336692e-05,
+      "loss": 0.962,
+      "step": 14148
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4452308706709726e-05,
+      "loss": 1.4552,
+      "step": 14149
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4451545109705072e-05,
+      "loss": 1.1049,
+      "step": 14150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4450781480328288e-05,
+      "loss": 1.1141,
+      "step": 14151
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445001781858493e-05,
+      "loss": 0.9171,
+      "step": 14152
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4449254124480545e-05,
+      "loss": 0.8568,
+      "step": 14153
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4448490398020694e-05,
+      "loss": 0.9428,
+      "step": 14154
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4447726639210923e-05,
+      "loss": 0.9703,
+      "step": 14155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444696284805679e-05,
+      "loss": 0.6606,
+      "step": 14156
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444619902456385e-05,
+      "loss": 1.3991,
+      "step": 14157
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445435168737658e-05,
+      "loss": 1.0158,
+      "step": 14158
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4444671280583767e-05,
+      "loss": 1.1761,
+      "step": 14159
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4443907360107734e-05,
+      "loss": 0.7548,
+      "step": 14160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4443143407315112e-05,
+      "loss": 0.8802,
+      "step": 14161
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4442379422211457e-05,
+      "loss": 0.9604,
+      "step": 14162
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441615404802324e-05,
+      "loss": 0.7882,
+      "step": 14163
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4440851355093278e-05,
+      "loss": 0.8862,
+      "step": 14164
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444008727308986e-05,
+      "loss": 0.5234,
+      "step": 14165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4439323158797638e-05,
+      "loss": 0.7091,
+      "step": 14166
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4438559012222164e-05,
+      "loss": 1.1491,
+      "step": 14167
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4437794833368994e-05,
+      "loss": 0.9138,
+      "step": 14168
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4437030622243691e-05,
+      "loss": 1.0347,
+      "step": 14169
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443626637885181e-05,
+      "loss": 1.2245,
+      "step": 14170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44355021031989e-05,
+      "loss": 0.8547,
+      "step": 14171
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443473779529053e-05,
+      "loss": 0.9689,
+      "step": 14172
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4433973455132255e-05,
+      "loss": 0.7788,
+      "step": 14173
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4433209082729634e-05,
+      "loss": 0.9795,
+      "step": 14174
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4432444678088222e-05,
+      "loss": 0.8737,
+      "step": 14175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4431680241213579e-05,
+      "loss": 0.9174,
+      "step": 14176
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4430915772111267e-05,
+      "loss": 0.9613,
+      "step": 14177
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4430151270786844e-05,
+      "loss": 0.6765,
+      "step": 14178
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4429386737245869e-05,
+      "loss": 1.0167,
+      "step": 14179
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44286221714939e-05,
+      "loss": 1.2811,
+      "step": 14180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4427857573536498e-05,
+      "loss": 1.0573,
+      "step": 14181
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442709294337923e-05,
+      "loss": 0.8856,
+      "step": 14182
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4426328281027644e-05,
+      "loss": 0.6715,
+      "step": 14183
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442556358648731e-05,
+      "loss": 1.3299,
+      "step": 14184
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424798859763787e-05,
+      "loss": 0.826,
+      "step": 14185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424034100862631e-05,
+      "loss": 1.0031,
+      "step": 14186
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4423269309789412e-05,
+      "loss": 1.2303,
+      "step": 14187
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4422504486549685e-05,
+      "loss": 0.9712,
+      "step": 14188
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4421739631149016e-05,
+      "loss": 1.2577,
+      "step": 14189
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4420974743592964e-05,
+      "loss": 1.1426,
+      "step": 14190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4420209823887095e-05,
+      "loss": 1.0463,
+      "step": 14191
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419444872036967e-05,
+      "loss": 0.9467,
+      "step": 14192
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4418679888048143e-05,
+      "loss": 1.1473,
+      "step": 14193
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4417914871926195e-05,
+      "loss": 0.9782,
+      "step": 14194
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4417149823676675e-05,
+      "loss": 0.7204,
+      "step": 14195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4416384743305151e-05,
+      "loss": 1.0842,
+      "step": 14196
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4415619630817186e-05,
+      "loss": 0.9524,
+      "step": 14197
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4414854486218346e-05,
+      "loss": 0.7777,
+      "step": 14198
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4414089309514197e-05,
+      "loss": 1.0838,
+      "step": 14199
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4413324100710295e-05,
+      "loss": 0.9475,
+      "step": 14200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4412558859812213e-05,
+      "loss": 0.9345,
+      "step": 14201
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441179358682551e-05,
+      "loss": 0.7485,
+      "step": 14202
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4411028281755757e-05,
+      "loss": 0.7738,
+      "step": 14203
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4410262944608518e-05,
+      "loss": 1.0936,
+      "step": 14204
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4409497575389352e-05,
+      "loss": 1.0402,
+      "step": 14205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4408732174103835e-05,
+      "loss": 0.5352,
+      "step": 14206
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4407966740757526e-05,
+      "loss": 1.3603,
+      "step": 14207
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440720127535599e-05,
+      "loss": 0.7128,
+      "step": 14208
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4406435777904803e-05,
+      "loss": 0.6584,
+      "step": 14209
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4405670248409525e-05,
+      "loss": 1.0101,
+      "step": 14210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440490468687572e-05,
+      "loss": 1.1029,
+      "step": 14211
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440413909330896e-05,
+      "loss": 1.3068,
+      "step": 14212
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4403373467714812e-05,
+      "loss": 0.7911,
+      "step": 14213
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4402607810098844e-05,
+      "loss": 0.8694,
+      "step": 14214
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4401842120466623e-05,
+      "loss": 0.9259,
+      "step": 14215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4401076398823716e-05,
+      "loss": 0.5991,
+      "step": 14216
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4400310645175693e-05,
+      "loss": 1.016,
+      "step": 14217
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439954485952812e-05,
+      "loss": 1.0847,
+      "step": 14218
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4398779041886575e-05,
+      "loss": 0.8411,
+      "step": 14219
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4398013192256615e-05,
+      "loss": 1.0929,
+      "step": 14220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4397247310643816e-05,
+      "loss": 1.1093,
+      "step": 14221
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4396481397053745e-05,
+      "loss": 1.0482,
+      "step": 14222
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4395715451491974e-05,
+      "loss": 1.238,
+      "step": 14223
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4394949473964074e-05,
+      "loss": 1.0021,
+      "step": 14224
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4394183464475613e-05,
+      "loss": 1.0309,
+      "step": 14225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4393417423032157e-05,
+      "loss": 1.1293,
+      "step": 14226
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4392651349639286e-05,
+      "loss": 0.8831,
+      "step": 14227
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4391885244302567e-05,
+      "loss": 1.1237,
+      "step": 14228
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4391119107027569e-05,
+      "loss": 0.872,
+      "step": 14229
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4390352937819867e-05,
+      "loss": 1.178,
+      "step": 14230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4389586736685029e-05,
+      "loss": 0.8406,
+      "step": 14231
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438882050362863e-05,
+      "loss": 0.9223,
+      "step": 14232
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438805423865624e-05,
+      "loss": 1.0572,
+      "step": 14233
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4387287941773433e-05,
+      "loss": 1.2689,
+      "step": 14234
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438652161298578e-05,
+      "loss": 1.2425,
+      "step": 14235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4385755252298857e-05,
+      "loss": 0.9019,
+      "step": 14236
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4384988859718232e-05,
+      "loss": 0.5786,
+      "step": 14237
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4384222435249481e-05,
+      "loss": 0.908,
+      "step": 14238
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438345597889818e-05,
+      "loss": 1.1559,
+      "step": 14239
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4382689490669898e-05,
+      "loss": 0.821,
+      "step": 14240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4381922970570213e-05,
+      "loss": 1.1607,
+      "step": 14241
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4381156418604696e-05,
+      "loss": 0.7466,
+      "step": 14242
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4380389834778925e-05,
+      "loss": 1.0198,
+      "step": 14243
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4379623219098473e-05,
+      "loss": 0.9987,
+      "step": 14244
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378856571568914e-05,
+      "loss": 0.9193,
+      "step": 14245
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378089892195823e-05,
+      "loss": 0.8811,
+      "step": 14246
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4377323180984774e-05,
+      "loss": 0.9237,
+      "step": 14247
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4376556437941347e-05,
+      "loss": 1.0256,
+      "step": 14248
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4375789663071119e-05,
+      "loss": 0.9608,
+      "step": 14249
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4375022856379657e-05,
+      "loss": 0.7987,
+      "step": 14250
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4374256017872546e-05,
+      "loss": 0.9241,
+      "step": 14251
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4373489147555357e-05,
+      "loss": 0.8353,
+      "step": 14252
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4372722245433672e-05,
+      "loss": 0.8774,
+      "step": 14253
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371955311513064e-05,
+      "loss": 0.8666,
+      "step": 14254
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371188345799111e-05,
+      "loss": 0.9484,
+      "step": 14255
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4370421348297395e-05,
+      "loss": 1.0773,
+      "step": 14256
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4369654319013482e-05,
+      "loss": 0.9817,
+      "step": 14257
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436888725795296e-05,
+      "loss": 0.8562,
+      "step": 14258
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436812016512141e-05,
+      "loss": 1.216,
+      "step": 14259
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4367353040524401e-05,
+      "loss": 1.0294,
+      "step": 14260
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4366585884167517e-05,
+      "loss": 1.1235,
+      "step": 14261
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4365818696056334e-05,
+      "loss": 0.8788,
+      "step": 14262
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4365051476196433e-05,
+      "loss": 0.8033,
+      "step": 14263
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4364284224593395e-05,
+      "loss": 0.9515,
+      "step": 14264
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4363516941252795e-05,
+      "loss": 0.7679,
+      "step": 14265
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4362749626180214e-05,
+      "loss": 0.8415,
+      "step": 14266
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361982279381237e-05,
+      "loss": 0.8841,
+      "step": 14267
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361214900861438e-05,
+      "loss": 0.7585,
+      "step": 14268
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4360447490626402e-05,
+      "loss": 1.0948,
+      "step": 14269
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4359680048681709e-05,
+      "loss": 1.0352,
+      "step": 14270
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4358912575032933e-05,
+      "loss": 0.8489,
+      "step": 14271
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4358145069685664e-05,
+      "loss": 0.7806,
+      "step": 14272
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4357377532645477e-05,
+      "loss": 1.2781,
+      "step": 14273
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4356609963917962e-05,
+      "loss": 1.2535,
+      "step": 14274
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4355842363508693e-05,
+      "loss": 0.6436,
+      "step": 14275
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4355074731423255e-05,
+      "loss": 0.8002,
+      "step": 14276
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4354307067667231e-05,
+      "loss": 0.8285,
+      "step": 14277
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4353539372246199e-05,
+      "loss": 1.0475,
+      "step": 14278
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4352771645165748e-05,
+      "loss": 0.9321,
+      "step": 14279
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4352003886431459e-05,
+      "loss": 1.0337,
+      "step": 14280
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435123609604891e-05,
+      "loss": 0.8071,
+      "step": 14281
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4350468274023691e-05,
+      "loss": 0.9279,
+      "step": 14282
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349700420361385e-05,
+      "loss": 0.9597,
+      "step": 14283
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4348932535067577e-05,
+      "loss": 1.0536,
+      "step": 14284
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4348164618147844e-05,
+      "loss": 1.0872,
+      "step": 14285
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4347396669607778e-05,
+      "loss": 0.6459,
+      "step": 14286
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4346628689452958e-05,
+      "loss": 1.3538,
+      "step": 14287
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4345860677688972e-05,
+      "loss": 1.2393,
+      "step": 14288
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434509263432141e-05,
+      "loss": 1.1014,
+      "step": 14289
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4344324559355842e-05,
+      "loss": 0.6804,
+      "step": 14290
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4343556452797871e-05,
+      "loss": 0.9218,
+      "step": 14291
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4342788314653073e-05,
+      "loss": 0.9903,
+      "step": 14292
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4342020144927033e-05,
+      "loss": 0.891,
+      "step": 14293
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4341251943625344e-05,
+      "loss": 1.1593,
+      "step": 14294
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434048371075359e-05,
+      "loss": 1.1225,
+      "step": 14295
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4339715446317352e-05,
+      "loss": 0.7684,
+      "step": 14296
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4338947150322225e-05,
+      "loss": 0.9246,
+      "step": 14297
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4338178822773792e-05,
+      "loss": 0.9372,
+      "step": 14298
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433741046367764e-05,
+      "loss": 0.8929,
+      "step": 14299
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4336642073039357e-05,
+      "loss": 1.1063,
+      "step": 14300
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4335873650864534e-05,
+      "loss": 1.0075,
+      "step": 14301
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4335105197158751e-05,
+      "loss": 0.7265,
+      "step": 14302
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4334336711927605e-05,
+      "loss": 1.1265,
+      "step": 14303
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4333568195176683e-05,
+      "loss": 1.0212,
+      "step": 14304
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4332799646911571e-05,
+      "loss": 1.1141,
+      "step": 14305
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4332031067137857e-05,
+      "loss": 1.03,
+      "step": 14306
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4331262455861132e-05,
+      "loss": 0.9791,
+      "step": 14307
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4330493813086986e-05,
+      "loss": 1.3427,
+      "step": 14308
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4329725138821008e-05,
+      "loss": 0.9361,
+      "step": 14309
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328956433068789e-05,
+      "loss": 1.1323,
+      "step": 14310
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328187695835915e-05,
+      "loss": 0.9063,
+      "step": 14311
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4327418927127983e-05,
+      "loss": 1.2811,
+      "step": 14312
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4326650126950579e-05,
+      "loss": 1.0322,
+      "step": 14313
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4325881295309295e-05,
+      "loss": 0.8995,
+      "step": 14314
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4325112432209722e-05,
+      "loss": 0.9907,
+      "step": 14315
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.432434353765745e-05,
+      "loss": 1.06,
+      "step": 14316
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4323574611658072e-05,
+      "loss": 0.9803,
+      "step": 14317
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4322805654217177e-05,
+      "loss": 0.9579,
+      "step": 14318
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4322036665340365e-05,
+      "loss": 0.547,
+      "step": 14319
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4321267645033217e-05,
+      "loss": 0.9186,
+      "step": 14320
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4320498593301333e-05,
+      "loss": 0.801,
+      "step": 14321
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4319729510150304e-05,
+      "loss": 0.8604,
+      "step": 14322
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318960395585722e-05,
+      "loss": 0.672,
+      "step": 14323
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318191249613181e-05,
+      "loss": 0.7598,
+      "step": 14324
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4317422072238271e-05,
+      "loss": 0.7517,
+      "step": 14325
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4316652863466591e-05,
+      "loss": 0.6664,
+      "step": 14326
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431588362330373e-05,
+      "loss": 1.0396,
+      "step": 14327
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4315114351755286e-05,
+      "loss": 0.9054,
+      "step": 14328
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431434504882685e-05,
+      "loss": 1.1161,
+      "step": 14329
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4313575714524017e-05,
+      "loss": 0.9783,
+      "step": 14330
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4312806348852384e-05,
+      "loss": 0.6295,
+      "step": 14331
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4312036951817545e-05,
+      "loss": 0.9027,
+      "step": 14332
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431126752342509e-05,
+      "loss": 0.865,
+      "step": 14333
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4310498063680624e-05,
+      "loss": 1.0496,
+      "step": 14334
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4309728572589737e-05,
+      "loss": 1.0868,
+      "step": 14335
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430895905015802e-05,
+      "loss": 0.8548,
+      "step": 14336
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4308189496391078e-05,
+      "loss": 1.1502,
+      "step": 14337
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4307419911294504e-05,
+      "loss": 1.0469,
+      "step": 14338
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4306650294873893e-05,
+      "loss": 0.9556,
+      "step": 14339
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4305880647134847e-05,
+      "loss": 1.1129,
+      "step": 14340
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4305110968082953e-05,
+      "loss": 1.1526,
+      "step": 14341
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4304341257723815e-05,
+      "loss": 0.8964,
+      "step": 14342
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4303571516063031e-05,
+      "loss": 0.8028,
+      "step": 14343
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4302801743106198e-05,
+      "loss": 1.3316,
+      "step": 14344
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4302031938858911e-05,
+      "loss": 0.8801,
+      "step": 14345
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430126210332677e-05,
+      "loss": 1.076,
+      "step": 14346
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4300492236515373e-05,
+      "loss": 1.0366,
+      "step": 14347
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429972233843032e-05,
+      "loss": 0.9979,
+      "step": 14348
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429895240907721e-05,
+      "loss": 1.4874,
+      "step": 14349
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429818244846164e-05,
+      "loss": 1.137,
+      "step": 14350
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4297412456589206e-05,
+      "loss": 1.2689,
+      "step": 14351
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4296642433465515e-05,
+      "loss": 0.9882,
+      "step": 14352
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4295872379096163e-05,
+      "loss": 1.2479,
+      "step": 14353
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4295102293486749e-05,
+      "loss": 0.9087,
+      "step": 14354
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294332176642875e-05,
+      "loss": 1.171,
+      "step": 14355
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4293562028570141e-05,
+      "loss": 0.8663,
+      "step": 14356
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4292791849274144e-05,
+      "loss": 1.0252,
+      "step": 14357
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4292021638760489e-05,
+      "loss": 1.1181,
+      "step": 14358
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4291251397034782e-05,
+      "loss": 0.7367,
+      "step": 14359
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4290481124102612e-05,
+      "loss": 0.8138,
+      "step": 14360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428971081996959e-05,
+      "loss": 0.978,
+      "step": 14361
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4288940484641311e-05,
+      "loss": 0.7966,
+      "step": 14362
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428817011812338e-05,
+      "loss": 0.8597,
+      "step": 14363
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4287399720421404e-05,
+      "loss": 0.9225,
+      "step": 14364
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428662929154098e-05,
+      "loss": 0.8391,
+      "step": 14365
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428585883148771e-05,
+      "loss": 1.4067,
+      "step": 14366
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4285088340267201e-05,
+      "loss": 1.1839,
+      "step": 14367
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4284317817885052e-05,
+      "loss": 0.725,
+      "step": 14368
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4283547264346869e-05,
+      "loss": 1.0122,
+      "step": 14369
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4282776679658255e-05,
+      "loss": 0.7306,
+      "step": 14370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4282006063824811e-05,
+      "loss": 1.3142,
+      "step": 14371
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4281235416852144e-05,
+      "loss": 0.7864,
+      "step": 14372
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4280464738745861e-05,
+      "loss": 0.874,
+      "step": 14373
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4279694029511561e-05,
+      "loss": 0.8927,
+      "step": 14374
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4278923289154849e-05,
+      "loss": 0.608,
+      "step": 14375
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4278152517681333e-05,
+      "loss": 0.8591,
+      "step": 14376
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277381715096618e-05,
+      "loss": 1.0021,
+      "step": 14377
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4276610881406307e-05,
+      "loss": 1.2303,
+      "step": 14378
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4275840016616007e-05,
+      "loss": 1.0041,
+      "step": 14379
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4275069120731325e-05,
+      "loss": 0.7716,
+      "step": 14380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4274298193757861e-05,
+      "loss": 0.9886,
+      "step": 14381
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427352723570123e-05,
+      "loss": 0.9701,
+      "step": 14382
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4272756246567032e-05,
+      "loss": 1.0311,
+      "step": 14383
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4271985226360879e-05,
+      "loss": 0.7177,
+      "step": 14384
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4271214175088374e-05,
+      "loss": 1.3417,
+      "step": 14385
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4270443092755125e-05,
+      "loss": 0.9933,
+      "step": 14386
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426967197936674e-05,
+      "loss": 0.7301,
+      "step": 14387
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4268900834928823e-05,
+      "loss": 0.8954,
+      "step": 14388
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426812965944699e-05,
+      "loss": 0.8467,
+      "step": 14389
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426735845292684e-05,
+      "loss": 0.8378,
+      "step": 14390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4266587215373988e-05,
+      "loss": 1.1973,
+      "step": 14391
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4265815946794039e-05,
+      "loss": 1.1089,
+      "step": 14392
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4265044647192603e-05,
+      "loss": 0.5657,
+      "step": 14393
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426427331657529e-05,
+      "loss": 0.7215,
+      "step": 14394
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263501954947705e-05,
+      "loss": 0.8496,
+      "step": 14395
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4262730562315463e-05,
+      "loss": 1.1317,
+      "step": 14396
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261959138684169e-05,
+      "loss": 0.9483,
+      "step": 14397
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261187684059437e-05,
+      "loss": 0.8943,
+      "step": 14398
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4260416198446872e-05,
+      "loss": 0.9242,
+      "step": 14399
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425964468185209e-05,
+      "loss": 1.0141,
+      "step": 14400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4258873134280697e-05,
+      "loss": 0.9848,
+      "step": 14401
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4258101555738306e-05,
+      "loss": 1.0133,
+      "step": 14402
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4257329946230527e-05,
+      "loss": 1.0263,
+      "step": 14403
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4256558305762975e-05,
+      "loss": 0.9946,
+      "step": 14404
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255786634341258e-05,
+      "loss": 0.8442,
+      "step": 14405
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255014931970985e-05,
+      "loss": 0.8183,
+      "step": 14406
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4254243198657772e-05,
+      "loss": 1.1436,
+      "step": 14407
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.425347143440723e-05,
+      "loss": 0.7395,
+      "step": 14408
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4252699639224971e-05,
+      "loss": 1.2257,
+      "step": 14409
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4251927813116608e-05,
+      "loss": 1.4679,
+      "step": 14410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4251155956087757e-05,
+      "loss": 1.1709,
+      "step": 14411
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4250384068144024e-05,
+      "loss": 0.7537,
+      "step": 14412
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4249612149291025e-05,
+      "loss": 0.6913,
+      "step": 14413
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4248840199534382e-05,
+      "loss": 0.915,
+      "step": 14414
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4248068218879691e-05,
+      "loss": 0.9698,
+      "step": 14415
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4247296207332584e-05,
+      "loss": 0.9186,
+      "step": 14416
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4246524164898664e-05,
+      "loss": 1.0013,
+      "step": 14417
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4245752091583547e-05,
+      "loss": 0.8043,
+      "step": 14418
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244979987392855e-05,
+      "loss": 1.0574,
+      "step": 14419
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244207852332193e-05,
+      "loss": 0.6134,
+      "step": 14420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4243435686407178e-05,
+      "loss": 1.0657,
+      "step": 14421
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4242663489623432e-05,
+      "loss": 0.8915,
+      "step": 14422
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4241891261986564e-05,
+      "loss": 1.4125,
+      "step": 14423
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4241119003502193e-05,
+      "loss": 1.1321,
+      "step": 14424
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4240346714175932e-05,
+      "loss": 0.807,
+      "step": 14425
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.42395743940134e-05,
+      "loss": 1.0388,
+      "step": 14426
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4238802043020211e-05,
+      "loss": 0.8114,
+      "step": 14427
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4238029661201985e-05,
+      "loss": 1.0855,
+      "step": 14428
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4237257248564335e-05,
+      "loss": 1.1153,
+      "step": 14429
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4236484805112878e-05,
+      "loss": 1.2649,
+      "step": 14430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4235712330853235e-05,
+      "loss": 1.1338,
+      "step": 14431
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4234939825791023e-05,
+      "loss": 1.1391,
+      "step": 14432
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4234167289931854e-05,
+      "loss": 0.8766,
+      "step": 14433
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4233394723281358e-05,
+      "loss": 0.899,
+      "step": 14434
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4232622125845139e-05,
+      "loss": 0.885,
+      "step": 14435
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4231849497628823e-05,
+      "loss": 0.9415,
+      "step": 14436
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4231076838638027e-05,
+      "loss": 1.2452,
+      "step": 14437
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4230304148878372e-05,
+      "loss": 1.0166,
+      "step": 14438
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4229531428355476e-05,
+      "loss": 0.5972,
+      "step": 14439
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4228758677074956e-05,
+      "loss": 0.9989,
+      "step": 14440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4227985895042435e-05,
+      "loss": 0.7375,
+      "step": 14441
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4227213082263527e-05,
+      "loss": 1.1709,
+      "step": 14442
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4226440238743859e-05,
+      "loss": 0.9344,
+      "step": 14443
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4225667364489051e-05,
+      "loss": 1.24,
+      "step": 14444
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4224894459504717e-05,
+      "loss": 0.9563,
+      "step": 14445
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4224121523796481e-05,
+      "loss": 1.0483,
+      "step": 14446
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4223348557369968e-05,
+      "loss": 0.9853,
+      "step": 14447
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4222575560230791e-05,
+      "loss": 1.0138,
+      "step": 14448
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4221802532384579e-05,
+      "loss": 0.9029,
+      "step": 14449
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4221029473836949e-05,
+      "loss": 1.1221,
+      "step": 14450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4220256384593522e-05,
+      "loss": 0.8951,
+      "step": 14451
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4219483264659924e-05,
+      "loss": 0.7573,
+      "step": 14452
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4218710114041774e-05,
+      "loss": 0.6819,
+      "step": 14453
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4217936932744696e-05,
+      "loss": 1.1001,
+      "step": 14454
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4217163720774311e-05,
+      "loss": 0.8087,
+      "step": 14455
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4216390478136246e-05,
+      "loss": 1.1602,
+      "step": 14456
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421561720483612e-05,
+      "loss": 0.8478,
+      "step": 14457
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4214843900879556e-05,
+      "loss": 1.3582,
+      "step": 14458
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4214070566272182e-05,
+      "loss": 1.097,
+      "step": 14459
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4213297201019618e-05,
+      "loss": 1.0924,
+      "step": 14460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4212523805127487e-05,
+      "loss": 0.6602,
+      "step": 14461
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4211750378601416e-05,
+      "loss": 1.1857,
+      "step": 14462
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421097692144703e-05,
+      "loss": 0.73,
+      "step": 14463
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.421020343366995e-05,
+      "loss": 0.7631,
+      "step": 14464
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4209429915275806e-05,
+      "loss": 0.7261,
+      "step": 14465
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4208656366270217e-05,
+      "loss": 0.9642,
+      "step": 14466
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4207882786658811e-05,
+      "loss": 0.9606,
+      "step": 14467
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4207109176447218e-05,
+      "loss": 0.8921,
+      "step": 14468
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4206335535641058e-05,
+      "loss": 0.8885,
+      "step": 14469
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4205561864245958e-05,
+      "loss": 0.8821,
+      "step": 14470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204788162267547e-05,
+      "loss": 1.3413,
+      "step": 14471
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204014429711447e-05,
+      "loss": 1.0385,
+      "step": 14472
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4203240666583287e-05,
+      "loss": 0.9443,
+      "step": 14473
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4202466872888698e-05,
+      "loss": 0.9611,
+      "step": 14474
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4201693048633302e-05,
+      "loss": 1.1149,
+      "step": 14475
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200919193822724e-05,
+      "loss": 1.0261,
+      "step": 14476
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200145308462599e-05,
+      "loss": 0.5013,
+      "step": 14477
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419937139255855e-05,
+      "loss": 0.9735,
+      "step": 14478
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198597446116205e-05,
+      "loss": 1.1321,
+      "step": 14479
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4197823469141195e-05,
+      "loss": 0.99,
+      "step": 14480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4197049461639143e-05,
+      "loss": 1.0924,
+      "step": 14481
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4196275423615681e-05,
+      "loss": 0.7332,
+      "step": 14482
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4195501355076443e-05,
+      "loss": 1.0533,
+      "step": 14483
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4194727256027052e-05,
+      "loss": 1.0257,
+      "step": 14484
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4193953126473137e-05,
+      "loss": 0.6722,
+      "step": 14485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4193178966420328e-05,
+      "loss": 0.6746,
+      "step": 14486
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4192404775874257e-05,
+      "loss": 0.971,
+      "step": 14487
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4191630554840553e-05,
+      "loss": 1.1619,
+      "step": 14488
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419085630332485e-05,
+      "loss": 1.1164,
+      "step": 14489
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419008202133277e-05,
+      "loss": 1.2379,
+      "step": 14490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4189307708869948e-05,
+      "loss": 1.077,
+      "step": 14491
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4188533365942016e-05,
+      "loss": 0.8856,
+      "step": 14492
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4187758992554605e-05,
+      "loss": 1.0928,
+      "step": 14493
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4186984588713346e-05,
+      "loss": 0.9979,
+      "step": 14494
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418621015442387e-05,
+      "loss": 1.2473,
+      "step": 14495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4185435689691809e-05,
+      "loss": 1.2301,
+      "step": 14496
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418466119452279e-05,
+      "loss": 0.6992,
+      "step": 14497
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4183886668922457e-05,
+      "loss": 0.9657,
+      "step": 14498
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4183112112896433e-05,
+      "loss": 0.9487,
+      "step": 14499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4182337526450351e-05,
+      "loss": 0.8762,
+      "step": 14500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4181562909589848e-05,
+      "loss": 1.1986,
+      "step": 14501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4180788262320558e-05,
+      "loss": 0.4749,
+      "step": 14502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4180013584648107e-05,
+      "loss": 0.7364,
+      "step": 14503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4179238876578134e-05,
+      "loss": 0.9986,
+      "step": 14504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4178464138116272e-05,
+      "loss": 0.8638,
+      "step": 14505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4177689369268155e-05,
+      "loss": 1.1875,
+      "step": 14506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176914570039417e-05,
+      "loss": 0.945,
+      "step": 14507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176139740435695e-05,
+      "loss": 1.1263,
+      "step": 14508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4175364880462618e-05,
+      "loss": 0.8908,
+      "step": 14509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4174589990125824e-05,
+      "loss": 0.7557,
+      "step": 14510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.417381506943095e-05,
+      "loss": 0.8983,
+      "step": 14511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4173040118383628e-05,
+      "loss": 0.9922,
+      "step": 14512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4172265136989496e-05,
+      "loss": 1.1872,
+      "step": 14513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4171490125254186e-05,
+      "loss": 0.8789,
+      "step": 14514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4170715083183337e-05,
+      "loss": 1.08,
+      "step": 14515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169940010782586e-05,
+      "loss": 1.1859,
+      "step": 14516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169164908057566e-05,
+      "loss": 0.8717,
+      "step": 14517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.416838977501392e-05,
+      "loss": 0.9174,
+      "step": 14518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4167614611657278e-05,
+      "loss": 0.9865,
+      "step": 14519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166839417993281e-05,
+      "loss": 1.0485,
+      "step": 14520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166064194027565e-05,
+      "loss": 1.2418,
+      "step": 14521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4165288939765764e-05,
+      "loss": 0.9593,
+      "step": 14522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4164513655213526e-05,
+      "loss": 1.0417,
+      "step": 14523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4163738340376477e-05,
+      "loss": 0.7247,
+      "step": 14524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4162962995260261e-05,
+      "loss": 1.4425,
+      "step": 14525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4162187619870516e-05,
+      "loss": 1.1115,
+      "step": 14526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4161412214212882e-05,
+      "loss": 0.7829,
+      "step": 14527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4160636778292993e-05,
+      "loss": 0.9457,
+      "step": 14528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159861312116494e-05,
+      "loss": 1.0592,
+      "step": 14529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159085815689022e-05,
+      "loss": 0.8767,
+      "step": 14530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4158310289016216e-05,
+      "loss": 1.0147,
+      "step": 14531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157534732103715e-05,
+      "loss": 0.8518,
+      "step": 14532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4156759144957159e-05,
+      "loss": 0.9754,
+      "step": 14533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.415598352758219e-05,
+      "loss": 0.8458,
+      "step": 14534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4155207879984447e-05,
+      "loss": 0.9438,
+      "step": 14535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.415443220216957e-05,
+      "loss": 1.0638,
+      "step": 14536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41536564941432e-05,
+      "loss": 1.0296,
+      "step": 14537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4152880755910982e-05,
+      "loss": 0.8839,
+      "step": 14538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4152104987478552e-05,
+      "loss": 1.33,
+      "step": 14539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4151329188851554e-05,
+      "loss": 0.9951,
+      "step": 14540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4150553360035628e-05,
+      "loss": 1.0034,
+      "step": 14541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149777501036419e-05,
+      "loss": 0.7904,
+      "step": 14542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4149001611859566e-05,
+      "loss": 0.8987,
+      "step": 14543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4148225692510713e-05,
+      "loss": 0.8574,
+      "step": 14544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4147449742995502e-05,
+      "loss": 1.3889,
+      "step": 14545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4146673763319574e-05,
+      "loss": 1.2204,
+      "step": 14546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4145897753488575e-05,
+      "loss": 0.9337,
+      "step": 14547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414512171350815e-05,
+      "loss": 1.0463,
+      "step": 14548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4144345643383934e-05,
+      "loss": 1.0627,
+      "step": 14549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414356954312158e-05,
+      "loss": 0.8071,
+      "step": 14550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414279341272673e-05,
+      "loss": 0.8514,
+      "step": 14551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4142017252205023e-05,
+      "loss": 1.2899,
+      "step": 14552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4141241061562107e-05,
+      "loss": 1.2592,
+      "step": 14553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414046484080363e-05,
+      "loss": 1.0635,
+      "step": 14554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4139688589935226e-05,
+      "loss": 0.9639,
+      "step": 14555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4138912308962551e-05,
+      "loss": 0.9391,
+      "step": 14556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4138135997891245e-05,
+      "loss": 0.7482,
+      "step": 14557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4137359656726957e-05,
+      "loss": 0.7838,
+      "step": 14558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4136583285475327e-05,
+      "loss": 1.1278,
+      "step": 14559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135806884142005e-05,
+      "loss": 0.8935,
+      "step": 14560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135030452732634e-05,
+      "loss": 0.9049,
+      "step": 14561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4134253991252867e-05,
+      "loss": 1.2228,
+      "step": 14562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4133477499708344e-05,
+      "loss": 0.5283,
+      "step": 14563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4132700978104712e-05,
+      "loss": 0.952,
+      "step": 14564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4131924426447621e-05,
+      "loss": 1.0943,
+      "step": 14565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4131147844742717e-05,
+      "loss": 0.8296,
+      "step": 14566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4130371232995643e-05,
+      "loss": 0.6961,
+      "step": 14567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4129594591212056e-05,
+      "loss": 0.69,
+      "step": 14568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4128817919397597e-05,
+      "loss": 0.8936,
+      "step": 14569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4128041217557915e-05,
+      "loss": 0.8919,
+      "step": 14570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4127264485698659e-05,
+      "loss": 0.7884,
+      "step": 14571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4126487723825477e-05,
+      "loss": 0.6522,
+      "step": 14572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4125710931944019e-05,
+      "loss": 0.9712,
+      "step": 14573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4124934110059933e-05,
+      "loss": 0.9871,
+      "step": 14574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4124157258178868e-05,
+      "loss": 0.9263,
+      "step": 14575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4123380376306472e-05,
+      "loss": 0.9971,
+      "step": 14576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4122603464448396e-05,
+      "loss": 1.086,
+      "step": 14577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4121826522610294e-05,
+      "loss": 0.8477,
+      "step": 14578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4121049550797807e-05,
+      "loss": 0.7562,
+      "step": 14579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4120272549016591e-05,
+      "loss": 0.7563,
+      "step": 14580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4119495517272296e-05,
+      "loss": 1.2352,
+      "step": 14581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4118718455570576e-05,
+      "loss": 0.9182,
+      "step": 14582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4117941363917072e-05,
+      "loss": 1.0742,
+      "step": 14583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4117164242317443e-05,
+      "loss": 0.9228,
+      "step": 14584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411638709077734e-05,
+      "loss": 0.9594,
+      "step": 14585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411560990930241e-05,
+      "loss": 1.0071,
+      "step": 14586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4114832697898311e-05,
+      "loss": 1.1697,
+      "step": 14587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411405545657069e-05,
+      "loss": 1.1473,
+      "step": 14588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41132781853252e-05,
+      "loss": 1.015,
+      "step": 14589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4112500884167498e-05,
+      "loss": 0.8696,
+      "step": 14590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411172355310323e-05,
+      "loss": 0.9723,
+      "step": 14591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4110946192138048e-05,
+      "loss": 0.8029,
+      "step": 14592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4110168801277615e-05,
+      "loss": 1.089,
+      "step": 14593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4109391380527576e-05,
+      "loss": 0.9276,
+      "step": 14594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4108613929893586e-05,
+      "loss": 1.0833,
+      "step": 14595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.41078364493813e-05,
+      "loss": 0.873,
+      "step": 14596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4107058938996369e-05,
+      "loss": 0.6937,
+      "step": 14597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4106281398744452e-05,
+      "loss": 0.7109,
+      "step": 14598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4105503828631201e-05,
+      "loss": 0.8564,
+      "step": 14599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4104726228662268e-05,
+      "loss": 0.9559,
+      "step": 14600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410394859884331e-05,
+      "loss": 1.3372,
+      "step": 14601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4103170939179985e-05,
+      "loss": 0.8944,
+      "step": 14602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4102393249677943e-05,
+      "loss": 0.9026,
+      "step": 14603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4101615530342843e-05,
+      "loss": 0.6926,
+      "step": 14604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4100837781180339e-05,
+      "loss": 0.8396,
+      "step": 14605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4100060002196085e-05,
+      "loss": 0.798,
+      "step": 14606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409928219339574e-05,
+      "loss": 0.9667,
+      "step": 14607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4098504354784962e-05,
+      "loss": 0.7602,
+      "step": 14608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4097726486369405e-05,
+      "loss": 0.8056,
+      "step": 14609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4096948588154723e-05,
+      "loss": 0.8494,
+      "step": 14610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4096170660146578e-05,
+      "loss": 0.7971,
+      "step": 14611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4095392702350625e-05,
+      "loss": 0.9574,
+      "step": 14612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4094614714772522e-05,
+      "loss": 0.8844,
+      "step": 14613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093836697417927e-05,
+      "loss": 1.0517,
+      "step": 14614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093058650292495e-05,
+      "loss": 1.0437,
+      "step": 14615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4092280573401884e-05,
+      "loss": 1.032,
+      "step": 14616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4091502466751759e-05,
+      "loss": 0.7408,
+      "step": 14617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4090724330347772e-05,
+      "loss": 0.8792,
+      "step": 14618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.408994616419558e-05,
+      "loss": 0.8206,
+      "step": 14619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.408916796830085e-05,
+      "loss": 0.8121,
+      "step": 14620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4088389742669236e-05,
+      "loss": 0.98,
+      "step": 14621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4087611487306393e-05,
+      "loss": 0.9258,
+      "step": 14622
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408683320221799e-05,
+      "loss": 0.7848,
+      "step": 14623
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4086054887409683e-05,
+      "loss": 0.9241,
+      "step": 14624
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4085276542887128e-05,
+      "loss": 0.7806,
+      "step": 14625
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408449816865599e-05,
+      "loss": 0.9309,
+      "step": 14626
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4083719764721927e-05,
+      "loss": 0.8736,
+      "step": 14627
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40829413310906e-05,
+      "loss": 0.93,
+      "step": 14628
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408216286776767e-05,
+      "loss": 0.7937,
+      "step": 14629
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4081384374758799e-05,
+      "loss": 1.0666,
+      "step": 14630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4080605852069646e-05,
+      "loss": 0.7562,
+      "step": 14631
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4079827299705877e-05,
+      "loss": 1.0079,
+      "step": 14632
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4079048717673152e-05,
+      "loss": 0.8636,
+      "step": 14633
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4078270105977125e-05,
+      "loss": 1.1033,
+      "step": 14634
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4077491464623471e-05,
+      "loss": 1.2761,
+      "step": 14635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4076712793617843e-05,
+      "loss": 1.1441,
+      "step": 14636
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4075934092965908e-05,
+      "loss": 0.9774,
+      "step": 14637
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407515536267333e-05,
+      "loss": 1.119,
+      "step": 14638
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4074376602745767e-05,
+      "loss": 1.1256,
+      "step": 14639
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4073597813188884e-05,
+      "loss": 0.9414,
+      "step": 14640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072818994008349e-05,
+      "loss": 1.0065,
+      "step": 14641
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072040145209817e-05,
+      "loss": 0.6574,
+      "step": 14642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4071261266798961e-05,
+      "loss": 0.8679,
+      "step": 14643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4070482358781441e-05,
+      "loss": 0.7769,
+      "step": 14644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406970342116292e-05,
+      "loss": 1.0472,
+      "step": 14645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4068924453949062e-05,
+      "loss": 0.7563,
+      "step": 14646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4068145457145535e-05,
+      "loss": 0.895,
+      "step": 14647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4067366430758004e-05,
+      "loss": 1.0117,
+      "step": 14648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4066587374792131e-05,
+      "loss": 0.7442,
+      "step": 14649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4065808289253581e-05,
+      "loss": 1.3346,
+      "step": 14650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4065029174148027e-05,
+      "loss": 0.7179,
+      "step": 14651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4064250029481126e-05,
+      "loss": 0.9245,
+      "step": 14652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4063470855258548e-05,
+      "loss": 1.1788,
+      "step": 14653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4062691651485958e-05,
+      "loss": 0.9831,
+      "step": 14654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4061912418169024e-05,
+      "loss": 0.8849,
+      "step": 14655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4061133155313412e-05,
+      "loss": 0.9371,
+      "step": 14656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4060353862924788e-05,
+      "loss": 0.8465,
+      "step": 14657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4059574541008821e-05,
+      "loss": 0.8707,
+      "step": 14658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4058795189571176e-05,
+      "loss": 0.7665,
+      "step": 14659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4058015808617522e-05,
+      "loss": 1.0024,
+      "step": 14660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4057236398153527e-05,
+      "loss": 0.812,
+      "step": 14661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4056456958184854e-05,
+      "loss": 0.9373,
+      "step": 14662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4055677488717182e-05,
+      "loss": 1.2,
+      "step": 14663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405489798975617e-05,
+      "loss": 0.9744,
+      "step": 14664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4054118461307488e-05,
+      "loss": 0.8615,
+      "step": 14665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405333890337681e-05,
+      "loss": 0.9506,
+      "step": 14666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4052559315969801e-05,
+      "loss": 0.7939,
+      "step": 14667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4051779699092127e-05,
+      "loss": 0.7983,
+      "step": 14668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4051000052749464e-05,
+      "loss": 0.8686,
+      "step": 14669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405022037694748e-05,
+      "loss": 0.816,
+      "step": 14670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4049440671691839e-05,
+      "loss": 0.9236,
+      "step": 14671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404866093698822e-05,
+      "loss": 0.9456,
+      "step": 14672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4047881172842287e-05,
+      "loss": 1.071,
+      "step": 14673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4047101379259711e-05,
+      "loss": 1.2156,
+      "step": 14674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4046321556246168e-05,
+      "loss": 1.0739,
+      "step": 14675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4045541703807325e-05,
+      "loss": 1.0203,
+      "step": 14676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4044761821948849e-05,
+      "loss": 0.8181,
+      "step": 14677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404398191067642e-05,
+      "loss": 0.6508,
+      "step": 14678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4043201969995706e-05,
+      "loss": 1.3602,
+      "step": 14679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4042421999912374e-05,
+      "loss": 0.9379,
+      "step": 14680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4041642000432103e-05,
+      "loss": 1.2588,
+      "step": 14681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4040861971560565e-05,
+      "loss": 0.8639,
+      "step": 14682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4040081913303427e-05,
+      "loss": 1.0115,
+      "step": 14683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4039301825666366e-05,
+      "loss": 1.0833,
+      "step": 14684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4038521708655054e-05,
+      "loss": 1.0376,
+      "step": 14685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.403774156227516e-05,
+      "loss": 0.7379,
+      "step": 14686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4036961386532364e-05,
+      "loss": 0.8348,
+      "step": 14687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4036181181432337e-05,
+      "loss": 1.2456,
+      "step": 14688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.403540094698075e-05,
+      "loss": 0.9687,
+      "step": 14689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4034620683183283e-05,
+      "loss": 0.9897,
+      "step": 14690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4033840390045601e-05,
+      "loss": 0.9504,
+      "step": 14691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4033060067573385e-05,
+      "loss": 0.9267,
+      "step": 14692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4032279715772314e-05,
+      "loss": 1.171,
+      "step": 14693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4031499334648054e-05,
+      "loss": 0.9319,
+      "step": 14694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4030718924206279e-05,
+      "loss": 0.9527,
+      "step": 14695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4029938484452673e-05,
+      "loss": 1.1448,
+      "step": 14696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4029158015392906e-05,
+      "loss": 1.3369,
+      "step": 14697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4028377517032651e-05,
+      "loss": 1.0924,
+      "step": 14698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4027596989377591e-05,
+      "loss": 0.9479,
+      "step": 14699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.40268164324334e-05,
+      "loss": 0.869,
+      "step": 14700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4026035846205748e-05,
+      "loss": 0.9913,
+      "step": 14701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4025255230700319e-05,
+      "loss": 0.9966,
+      "step": 14702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402447458592279e-05,
+      "loss": 0.8525,
+      "step": 14703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4023693911878829e-05,
+      "loss": 1.0352,
+      "step": 14704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4022913208574123e-05,
+      "loss": 1.0824,
+      "step": 14705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4022132476014342e-05,
+      "loss": 0.7803,
+      "step": 14706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402135171420517e-05,
+      "loss": 1.087,
+      "step": 14707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4020570923152282e-05,
+      "loss": 1.0825,
+      "step": 14708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4019790102861355e-05,
+      "loss": 0.9693,
+      "step": 14709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4019009253338065e-05,
+      "loss": 1.1011,
+      "step": 14710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4018228374588097e-05,
+      "loss": 0.8639,
+      "step": 14711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4017447466617124e-05,
+      "loss": 0.9371,
+      "step": 14712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401666652943083e-05,
+      "loss": 0.7533,
+      "step": 14713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4015885563034889e-05,
+      "loss": 0.9297,
+      "step": 14714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4015104567434981e-05,
+      "loss": 1.0204,
+      "step": 14715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4014323542636787e-05,
+      "loss": 0.9879,
+      "step": 14716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4013542488645986e-05,
+      "loss": 0.8762,
+      "step": 14717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4012761405468262e-05,
+      "loss": 0.8825,
+      "step": 14718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4011980293109288e-05,
+      "loss": 0.9836,
+      "step": 14719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4011199151574748e-05,
+      "loss": 1.1139,
+      "step": 14720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4010417980870323e-05,
+      "loss": 0.9271,
+      "step": 14721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4009636781001695e-05,
+      "loss": 0.9868,
+      "step": 14722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.400885555197454e-05,
+      "loss": 0.9731,
+      "step": 14723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4008074293794544e-05,
+      "loss": 1.0291,
+      "step": 14724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4007293006467388e-05,
+      "loss": 0.9629,
+      "step": 14725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4006511689998747e-05,
+      "loss": 0.78,
+      "step": 14726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4005730344394312e-05,
+      "loss": 0.8936,
+      "step": 14727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004948969659762e-05,
+      "loss": 0.8933,
+      "step": 14728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004167565800777e-05,
+      "loss": 1.3051,
+      "step": 14729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.400338613282304e-05,
+      "loss": 0.7433,
+      "step": 14730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4002604670732236e-05,
+      "loss": 0.9332,
+      "step": 14731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001823179534044e-05,
+      "loss": 0.8793,
+      "step": 14732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001041659234153e-05,
+      "loss": 0.7722,
+      "step": 14733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4000260109838241e-05,
+      "loss": 0.9456,
+      "step": 14734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3999478531351991e-05,
+      "loss": 1.0576,
+      "step": 14735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3998696923781093e-05,
+      "loss": 0.8779,
+      "step": 14736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3997915287131224e-05,
+      "loss": 1.1664,
+      "step": 14737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3997133621408072e-05,
+      "loss": 0.7328,
+      "step": 14738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3996351926617322e-05,
+      "loss": 1.1453,
+      "step": 14739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3995570202764656e-05,
+      "loss": 0.797,
+      "step": 14740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3994788449855757e-05,
+      "loss": 0.9491,
+      "step": 14741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3994006667896316e-05,
+      "loss": 0.9567,
+      "step": 14742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3993224856892015e-05,
+      "loss": 0.9513,
+      "step": 14743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3992443016848538e-05,
+      "loss": 0.865,
+      "step": 14744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3991661147771574e-05,
+      "loss": 1.1353,
+      "step": 14745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3990879249666804e-05,
+      "loss": 0.774,
+      "step": 14746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3990097322539918e-05,
+      "loss": 0.7918,
+      "step": 14747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989315366396603e-05,
+      "loss": 0.8383,
+      "step": 14748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3988533381242542e-05,
+      "loss": 0.8806,
+      "step": 14749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3987751367083421e-05,
+      "loss": 1.1924,
+      "step": 14750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3986969323924932e-05,
+      "loss": 0.9222,
+      "step": 14751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398618725177276e-05,
+      "loss": 0.937,
+      "step": 14752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3985405150632589e-05,
+      "loss": 0.9477,
+      "step": 14753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3984623020510113e-05,
+      "loss": 0.9412,
+      "step": 14754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3983840861411014e-05,
+      "loss": 1.0091,
+      "step": 14755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398305867334098e-05,
+      "loss": 0.5629,
+      "step": 14756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3982276456305703e-05,
+      "loss": 1.1265,
+      "step": 14757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3981494210310871e-05,
+      "loss": 1.0084,
+      "step": 14758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3980711935362168e-05,
+      "loss": 0.8317,
+      "step": 14759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3979929631465286e-05,
+      "loss": 1.2013,
+      "step": 14760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3979147298625914e-05,
+      "loss": 0.9813,
+      "step": 14761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.397836493684974e-05,
+      "loss": 0.7556,
+      "step": 14762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3977582546142459e-05,
+      "loss": 0.9125,
+      "step": 14763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3976800126509752e-05,
+      "loss": 0.9179,
+      "step": 14764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3976017677957313e-05,
+      "loss": 1.2154,
+      "step": 14765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3975235200490834e-05,
+      "loss": 0.697,
+      "step": 14766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3974452694116e-05,
+      "loss": 0.9363,
+      "step": 14767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3973670158838508e-05,
+      "loss": 0.7516,
+      "step": 14768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3972887594664044e-05,
+      "loss": 0.9371,
+      "step": 14769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39721050015983e-05,
+      "loss": 0.9907,
+      "step": 14770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3971322379646965e-05,
+      "loss": 0.8272,
+      "step": 14771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3970539728815738e-05,
+      "loss": 0.9647,
+      "step": 14772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969757049110305e-05,
+      "loss": 1.2171,
+      "step": 14773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3968974340536352e-05,
+      "loss": 1.1522,
+      "step": 14774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396819160309958e-05,
+      "loss": 1.1141,
+      "step": 14775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3967408836805678e-05,
+      "loss": 1.0334,
+      "step": 14776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3966626041660339e-05,
+      "loss": 1.0192,
+      "step": 14777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3965843217669255e-05,
+      "loss": 0.853,
+      "step": 14778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3965060364838117e-05,
+      "loss": 0.9489,
+      "step": 14779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396427748317262e-05,
+      "loss": 0.9426,
+      "step": 14780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3963494572678457e-05,
+      "loss": 1.0898,
+      "step": 14781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3962711633361323e-05,
+      "loss": 0.7945,
+      "step": 14782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396192866522691e-05,
+      "loss": 1.4124,
+      "step": 14783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3961145668280912e-05,
+      "loss": 0.7707,
+      "step": 14784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396036264252902e-05,
+      "loss": 0.8801,
+      "step": 14785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3959579587976933e-05,
+      "loss": 1.0651,
+      "step": 14786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3958796504630341e-05,
+      "loss": 0.9717,
+      "step": 14787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3958013392494947e-05,
+      "loss": 1.2258,
+      "step": 14788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3957230251576435e-05,
+      "loss": 0.5724,
+      "step": 14789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3956447081880506e-05,
+      "loss": 1.363,
+      "step": 14790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3955663883412856e-05,
+      "loss": 1.2915,
+      "step": 14791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3954880656179176e-05,
+      "loss": 1.4738,
+      "step": 14792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3954097400185169e-05,
+      "loss": 1.1345,
+      "step": 14793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3953314115436526e-05,
+      "loss": 1.069,
+      "step": 14794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3952530801938942e-05,
+      "loss": 0.912,
+      "step": 14795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3951747459698114e-05,
+      "loss": 1.119,
+      "step": 14796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395096408871974e-05,
+      "loss": 0.7899,
+      "step": 14797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950180689009516e-05,
+      "loss": 1.0902,
+      "step": 14798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3949397260573143e-05,
+      "loss": 0.8439,
+      "step": 14799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3948613803416312e-05,
+      "loss": 0.8653,
+      "step": 14800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3947830317544725e-05,
+      "loss": 0.9783,
+      "step": 14801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3947046802964072e-05,
+      "loss": 0.49,
+      "step": 14802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3946263259680062e-05,
+      "loss": 0.9805,
+      "step": 14803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3945479687698384e-05,
+      "loss": 0.8958,
+      "step": 14804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.394469608702474e-05,
+      "loss": 0.6399,
+      "step": 14805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.394391245766483e-05,
+      "loss": 0.7824,
+      "step": 14806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3943128799624351e-05,
+      "loss": 1.0879,
+      "step": 14807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3942345112908999e-05,
+      "loss": 0.9161,
+      "step": 14808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3941561397524477e-05,
+      "loss": 0.9504,
+      "step": 14809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3940777653476482e-05,
+      "loss": 1.0807,
+      "step": 14810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939993880770715e-05,
+      "loss": 0.8086,
+      "step": 14811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939210079412875e-05,
+      "loss": 1.0152,
+      "step": 14812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3938426249408663e-05,
+      "loss": 0.5836,
+      "step": 14813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3937642390763775e-05,
+      "loss": 0.8956,
+      "step": 14814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3936858503483916e-05,
+      "loss": 0.7919,
+      "step": 14815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3936074587574786e-05,
+      "loss": 0.9096,
+      "step": 14816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3935290643042082e-05,
+      "loss": 0.718,
+      "step": 14817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3934506669891507e-05,
+      "loss": 0.7661,
+      "step": 14818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3933722668128766e-05,
+      "loss": 0.7247,
+      "step": 14819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3932938637759555e-05,
+      "loss": 1.1844,
+      "step": 14820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3932154578789576e-05,
+      "loss": 1.0726,
+      "step": 14821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3931370491224535e-05,
+      "loss": 0.8207,
+      "step": 14822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.393058637507013e-05,
+      "loss": 0.5531,
+      "step": 14823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3929802230332064e-05,
+      "loss": 0.7771,
+      "step": 14824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3929018057016042e-05,
+      "loss": 1.2672,
+      "step": 14825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392823385512776e-05,
+      "loss": 1.2087,
+      "step": 14826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3927449624672928e-05,
+      "loss": 0.9706,
+      "step": 14827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3926665365657248e-05,
+      "loss": 1.219,
+      "step": 14828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925881078086419e-05,
+      "loss": 0.9526,
+      "step": 14829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925096761966146e-05,
+      "loss": 1.2387,
+      "step": 14830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3924312417302134e-05,
+      "loss": 0.8418,
+      "step": 14831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3923528044100087e-05,
+      "loss": 0.7087,
+      "step": 14832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.392274364236571e-05,
+      "loss": 0.7344,
+      "step": 14833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921959212104703e-05,
+      "loss": 0.9412,
+      "step": 14834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921174753322775e-05,
+      "loss": 1.0867,
+      "step": 14835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3920390266025628e-05,
+      "loss": 1.0202,
+      "step": 14836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3919605750218967e-05,
+      "loss": 1.0489,
+      "step": 14837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3918821205908497e-05,
+      "loss": 0.4828,
+      "step": 14838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3918036633099928e-05,
+      "loss": 0.9076,
+      "step": 14839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3917252031798959e-05,
+      "loss": 1.0754,
+      "step": 14840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3916467402011296e-05,
+      "loss": 0.7601,
+      "step": 14841
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3915682743742648e-05,
+      "loss": 1.3764,
+      "step": 14842
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3914898056998725e-05,
+      "loss": 1.2301,
+      "step": 14843
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3914113341785223e-05,
+      "loss": 0.7448,
+      "step": 14844
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3913328598107856e-05,
+      "loss": 0.8057,
+      "step": 14845
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3912543825972329e-05,
+      "loss": 1.319,
+      "step": 14846
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3911759025384348e-05,
+      "loss": 1.0988,
+      "step": 14847
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3910974196349623e-05,
+      "loss": 0.8689,
+      "step": 14848
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3910189338873858e-05,
+      "loss": 0.8829,
+      "step": 14849
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390940445296276e-05,
+      "loss": 0.8651,
+      "step": 14850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3908619538622042e-05,
+      "loss": 0.9949,
+      "step": 14851
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3907834595857406e-05,
+      "loss": 0.6995,
+      "step": 14852
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3907049624674564e-05,
+      "loss": 1.3318,
+      "step": 14853
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3906264625079225e-05,
+      "loss": 0.9199,
+      "step": 14854
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3905479597077096e-05,
+      "loss": 1.1682,
+      "step": 14855
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3904694540673883e-05,
+      "loss": 0.721,
+      "step": 14856
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903909455875298e-05,
+      "loss": 1.0908,
+      "step": 14857
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3903124342687053e-05,
+      "loss": 0.9703,
+      "step": 14858
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390233920111485e-05,
+      "loss": 1.018,
+      "step": 14859
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3901554031164404e-05,
+      "loss": 0.8307,
+      "step": 14860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3900768832841425e-05,
+      "loss": 0.8453,
+      "step": 14861
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899983606151621e-05,
+      "loss": 1.2334,
+      "step": 14862
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899198351100705e-05,
+      "loss": 1.0407,
+      "step": 14863
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3898413067694385e-05,
+      "loss": 1.1339,
+      "step": 14864
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3897627755938372e-05,
+      "loss": 0.9769,
+      "step": 14865
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3896842415838374e-05,
+      "loss": 0.9558,
+      "step": 14866
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3896057047400109e-05,
+      "loss": 0.8924,
+      "step": 14867
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3895271650629285e-05,
+      "loss": 0.8727,
+      "step": 14868
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3894486225531611e-05,
+      "loss": 1.1163,
+      "step": 14869
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3893700772112802e-05,
+      "loss": 1.2164,
+      "step": 14870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3892915290378568e-05,
+      "loss": 1.3988,
+      "step": 14871
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389212978033462e-05,
+      "loss": 1.2252,
+      "step": 14872
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3891344241986675e-05,
+      "loss": 1.0144,
+      "step": 14873
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3890558675340442e-05,
+      "loss": 0.8893,
+      "step": 14874
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3889773080401631e-05,
+      "loss": 1.2192,
+      "step": 14875
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3888987457175962e-05,
+      "loss": 0.9024,
+      "step": 14876
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3888201805669143e-05,
+      "loss": 0.7483,
+      "step": 14877
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3887416125886888e-05,
+      "loss": 1.0257,
+      "step": 14878
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3886630417834913e-05,
+      "loss": 0.8388,
+      "step": 14879
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388584468151893e-05,
+      "loss": 1.3484,
+      "step": 14880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388505891694465e-05,
+      "loss": 0.8974,
+      "step": 14881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388427312411779e-05,
+      "loss": 1.0221,
+      "step": 14882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388348730304407e-05,
+      "loss": 0.9955,
+      "step": 14883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3882701453729195e-05,
+      "loss": 1.1408,
+      "step": 14884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3881915576178883e-05,
+      "loss": 0.8419,
+      "step": 14885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3881129670398852e-05,
+      "loss": 0.9438,
+      "step": 14886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3880343736394812e-05,
+      "loss": 1.1445,
+      "step": 14887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3879557774172487e-05,
+      "loss": 0.8396,
+      "step": 14888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3878771783737584e-05,
+      "loss": 1.4771,
+      "step": 14889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387798576509582e-05,
+      "loss": 1.179,
+      "step": 14890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3877199718252915e-05,
+      "loss": 1.1037,
+      "step": 14891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3876413643214584e-05,
+      "loss": 0.6018,
+      "step": 14892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387562753998654e-05,
+      "loss": 1.1626,
+      "step": 14893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3874841408574504e-05,
+      "loss": 0.9969,
+      "step": 14894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3874055248984191e-05,
+      "loss": 0.9503,
+      "step": 14895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3873269061221315e-05,
+      "loss": 1.1117,
+      "step": 14896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3872482845291599e-05,
+      "loss": 1.1313,
+      "step": 14897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3871696601200759e-05,
+      "loss": 1.2449,
+      "step": 14898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3870910328954509e-05,
+      "loss": 0.9232,
+      "step": 14899
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387012402855857e-05,
+      "loss": 0.9672,
+      "step": 14900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3869337700018659e-05,
+      "loss": 0.93,
+      "step": 14901
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3868551343340493e-05,
+      "loss": 0.6437,
+      "step": 14902
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3867764958529798e-05,
+      "loss": 0.9916,
+      "step": 14903
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3866978545592283e-05,
+      "loss": 0.8528,
+      "step": 14904
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3866192104533667e-05,
+      "loss": 0.8689,
+      "step": 14905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3865405635359674e-05,
+      "loss": 0.8627,
+      "step": 14906
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3864619138076025e-05,
+      "loss": 0.7064,
+      "step": 14907
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863832612688436e-05,
+      "loss": 1.0686,
+      "step": 14908
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863046059202628e-05,
+      "loss": 0.884,
+      "step": 14909
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3862259477624317e-05,
+      "loss": 1.2884,
+      "step": 14910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3861472867959228e-05,
+      "loss": 1.0144,
+      "step": 14911
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3860686230213078e-05,
+      "loss": 1.2223,
+      "step": 14912
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3859899564391592e-05,
+      "loss": 0.7528,
+      "step": 14913
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3859112870500483e-05,
+      "loss": 1.2032,
+      "step": 14914
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3858326148545483e-05,
+      "loss": 1.0474,
+      "step": 14915
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3857539398532302e-05,
+      "loss": 1.1494,
+      "step": 14916
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385675262046667e-05,
+      "loss": 0.7413,
+      "step": 14917
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855965814354303e-05,
+      "loss": 0.6673,
+      "step": 14918
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855178980200923e-05,
+      "loss": 1.0622,
+      "step": 14919
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385439211801225e-05,
+      "loss": 1.3399,
+      "step": 14920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3853605227794017e-05,
+      "loss": 0.7696,
+      "step": 14921
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852818309551933e-05,
+      "loss": 1.2402,
+      "step": 14922
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852031363291728e-05,
+      "loss": 1.0612,
+      "step": 14923
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3851244389019125e-05,
+      "loss": 0.8772,
+      "step": 14924
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3850457386739846e-05,
+      "loss": 1.1683,
+      "step": 14925
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384967035645961e-05,
+      "loss": 0.8953,
+      "step": 14926
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3848883298184143e-05,
+      "loss": 0.9937,
+      "step": 14927
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3848096211919173e-05,
+      "loss": 0.7134,
+      "step": 14928
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3847309097670419e-05,
+      "loss": 1.0596,
+      "step": 14929
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3846521955443604e-05,
+      "loss": 1.0611,
+      "step": 14930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3845734785244455e-05,
+      "loss": 1.1487,
+      "step": 14931
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3844947587078697e-05,
+      "loss": 0.6831,
+      "step": 14932
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3844160360952053e-05,
+      "loss": 0.8343,
+      "step": 14933
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3843373106870247e-05,
+      "loss": 0.8348,
+      "step": 14934
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3842585824839006e-05,
+      "loss": 1.014,
+      "step": 14935
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384179851486405e-05,
+      "loss": 0.8218,
+      "step": 14936
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3841011176951114e-05,
+      "loss": 1.0382,
+      "step": 14937
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3840223811105916e-05,
+      "loss": 0.9955,
+      "step": 14938
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3839436417334184e-05,
+      "loss": 1.2585,
+      "step": 14939
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3838648995641645e-05,
+      "loss": 1.1046,
+      "step": 14940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3837861546034023e-05,
+      "loss": 0.7953,
+      "step": 14941
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3837074068517045e-05,
+      "loss": 0.8703,
+      "step": 14942
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3836286563096441e-05,
+      "loss": 0.7512,
+      "step": 14943
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3835499029777935e-05,
+      "loss": 0.8105,
+      "step": 14944
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383471146856725e-05,
+      "loss": 1.1688,
+      "step": 14945
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3833923879470119e-05,
+      "loss": 1.3901,
+      "step": 14946
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3833136262492271e-05,
+      "loss": 0.9269,
+      "step": 14947
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3832348617639428e-05,
+      "loss": 0.926,
+      "step": 14948
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383156094491732e-05,
+      "loss": 1.275,
+      "step": 14949
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3830773244331677e-05,
+      "loss": 0.832,
+      "step": 14950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382998551588822e-05,
+      "loss": 0.9517,
+      "step": 14951
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3829197759592689e-05,
+      "loss": 0.7783,
+      "step": 14952
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3828409975450807e-05,
+      "loss": 0.7826,
+      "step": 14953
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3827622163468299e-05,
+      "loss": 1.1014,
+      "step": 14954
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3826834323650899e-05,
+      "loss": 0.9742,
+      "step": 14955
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3826046456004336e-05,
+      "loss": 0.561,
+      "step": 14956
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3825258560534334e-05,
+      "loss": 1.0096,
+      "step": 14957
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3824470637246634e-05,
+      "loss": 1.16,
+      "step": 14958
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3823682686146956e-05,
+      "loss": 0.7287,
+      "step": 14959
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382289470724103e-05,
+      "loss": 0.8845,
+      "step": 14960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382210670053459e-05,
+      "loss": 0.9515,
+      "step": 14961
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3821318666033367e-05,
+      "loss": 0.9471,
+      "step": 14962
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382053060374309e-05,
+      "loss": 1.2984,
+      "step": 14963
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3819742513669491e-05,
+      "loss": 0.9646,
+      "step": 14964
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38189543958183e-05,
+      "loss": 0.905,
+      "step": 14965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3818166250195246e-05,
+      "loss": 1.0114,
+      "step": 14966
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3817378076806065e-05,
+      "loss": 1.0626,
+      "step": 14967
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381658987565649e-05,
+      "loss": 1.0017,
+      "step": 14968
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815801646752244e-05,
+      "loss": 0.8561,
+      "step": 14969
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815013390099068e-05,
+      "loss": 0.8533,
+      "step": 14970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381422510570269e-05,
+      "loss": 0.7599,
+      "step": 14971
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3813436793568843e-05,
+      "loss": 1.0365,
+      "step": 14972
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3812648453703261e-05,
+      "loss": 0.8905,
+      "step": 14973
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3811860086111676e-05,
+      "loss": 1.3808,
+      "step": 14974
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381107169079982e-05,
+      "loss": 0.7972,
+      "step": 14975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3810283267773428e-05,
+      "loss": 0.9202,
+      "step": 14976
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3809494817038234e-05,
+      "loss": 0.8907,
+      "step": 14977
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380870633859997e-05,
+      "loss": 0.5203,
+      "step": 14978
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380791783246437e-05,
+      "loss": 1.2096,
+      "step": 14979
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3807129298637173e-05,
+      "loss": 1.0041,
+      "step": 14980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3806340737124101e-05,
+      "loss": 1.0246,
+      "step": 14981
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.38055521479309e-05,
+      "loss": 0.8756,
+      "step": 14982
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3804763531063306e-05,
+      "loss": 1.026,
+      "step": 14983
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803974886527043e-05,
+      "loss": 0.9059,
+      "step": 14984
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803186214327852e-05,
+      "loss": 0.7791,
+      "step": 14985
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3802397514471473e-05,
+      "loss": 0.5978,
+      "step": 14986
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3801608786963634e-05,
+      "loss": 1.0907,
+      "step": 14987
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3800820031810073e-05,
+      "loss": 1.5549,
+      "step": 14988
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.380003124901653e-05,
+      "loss": 0.7852,
+      "step": 14989
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3799242438588735e-05,
+      "loss": 1.0557,
+      "step": 14990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.379845360053243e-05,
+      "loss": 1.4458,
+      "step": 14991
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3797664734853344e-05,
+      "loss": 1.2256,
+      "step": 14992
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3796875841557223e-05,
+      "loss": 1.116,
+      "step": 14993
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3796086920649795e-05,
+      "loss": 0.8187,
+      "step": 14994
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3795297972136805e-05,
+      "loss": 0.8258,
+      "step": 14995
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3794508996023986e-05,
+      "loss": 0.7781,
+      "step": 14996
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3793719992317072e-05,
+      "loss": 1.2138,
+      "step": 14997
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3792930961021812e-05,
+      "loss": 1.0386,
+      "step": 14998
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3792141902143935e-05,
+      "loss": 1.3103,
+      "step": 14999
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3791352815689174e-05,
+      "loss": 0.9499,
+      "step": 15000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3790563701663283e-05,
+      "loss": 0.9655,
+      "step": 15001
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3789774560071987e-05,
+      "loss": 1.0634,
+      "step": 15002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378898539092103e-05,
+      "loss": 0.9243,
+      "step": 15003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3788196194216153e-05,
+      "loss": 0.8077,
+      "step": 15004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378740696996309e-05,
+      "loss": 0.9613,
+      "step": 15005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3786617718167583e-05,
+      "loss": 0.9256,
+      "step": 15006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3785828438835372e-05,
+      "loss": 0.7859,
+      "step": 15007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.37850391319722e-05,
+      "loss": 1.2029,
+      "step": 15008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3784249797583797e-05,
+      "loss": 0.7826,
+      "step": 15009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3783460435675913e-05,
+      "loss": 0.8451,
+      "step": 15010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3782671046254284e-05,
+      "loss": 0.968,
+      "step": 15011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3781881629324647e-05,
+      "loss": 0.8587,
+      "step": 15012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3781092184892753e-05,
+      "loss": 0.8203,
+      "step": 15013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3780302712964335e-05,
+      "loss": 0.711,
+      "step": 15014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3779513213545132e-05,
+      "loss": 1.1085,
+      "step": 15015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3778723686640893e-05,
+      "loss": 1.0827,
+      "step": 15016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3777934132257356e-05,
+      "loss": 0.791,
+      "step": 15017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3777144550400262e-05,
+      "loss": 1.0139,
+      "step": 15018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3776354941075351e-05,
+      "loss": 1.018,
+      "step": 15019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3775565304288372e-05,
+      "loss": 0.7618,
+      "step": 15020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3774775640045057e-05,
+      "loss": 0.925,
+      "step": 15021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3773985948351159e-05,
+      "loss": 1.069,
+      "step": 15022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3773196229212416e-05,
+      "loss": 0.8295,
+      "step": 15023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3772406482634568e-05,
+      "loss": 0.7538,
+      "step": 15024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3771616708623363e-05,
+      "loss": 0.7078,
+      "step": 15025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3770826907184543e-05,
+      "loss": 0.8252,
+      "step": 15026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3770037078323847e-05,
+      "loss": 0.996,
+      "step": 15027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3769247222047028e-05,
+      "loss": 1.1782,
+      "step": 15028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3768457338359824e-05,
+      "loss": 0.9536,
+      "step": 15029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3767667427267976e-05,
+      "loss": 1.0973,
+      "step": 15030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3766877488777236e-05,
+      "loss": 1.1679,
+      "step": 15031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3766087522893342e-05,
+      "loss": 0.5781,
+      "step": 15032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3765297529622043e-05,
+      "loss": 0.9976,
+      "step": 15033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376450750896908e-05,
+      "loss": 0.9754,
+      "step": 15034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3763717460940202e-05,
+      "loss": 0.9471,
+      "step": 15035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762927385541151e-05,
+      "loss": 0.9102,
+      "step": 15036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762137282777676e-05,
+      "loss": 1.0667,
+      "step": 15037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3761347152655524e-05,
+      "loss": 0.9519,
+      "step": 15038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3760556995180432e-05,
+      "loss": 0.4942,
+      "step": 15039
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3759766810358156e-05,
+      "loss": 0.9297,
+      "step": 15040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758976598194438e-05,
+      "loss": 1.0955,
+      "step": 15041
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758186358695022e-05,
+      "loss": 1.1072,
+      "step": 15042
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375739609186566e-05,
+      "loss": 0.647,
+      "step": 15043
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3756605797712094e-05,
+      "loss": 0.8712,
+      "step": 15044
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3755815476240076e-05,
+      "loss": 1.0516,
+      "step": 15045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3755025127455348e-05,
+      "loss": 0.8392,
+      "step": 15046
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3754234751363662e-05,
+      "loss": 0.7691,
+      "step": 15047
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3753444347970764e-05,
+      "loss": 0.9601,
+      "step": 15048
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3752653917282402e-05,
+      "loss": 1.0115,
+      "step": 15049
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3751863459304324e-05,
+      "loss": 0.746,
+      "step": 15050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.375107297404228e-05,
+      "loss": 0.8128,
+      "step": 15051
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3750282461502012e-05,
+      "loss": 0.7826,
+      "step": 15052
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3749491921689277e-05,
+      "loss": 1.0859,
+      "step": 15053
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748701354609819e-05,
+      "loss": 0.9142,
+      "step": 15054
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3747910760269389e-05,
+      "loss": 1.038,
+      "step": 15055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3747120138673736e-05,
+      "loss": 1.0322,
+      "step": 15056
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374632948982861e-05,
+      "loss": 1.2241,
+      "step": 15057
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3745538813739758e-05,
+      "loss": 0.9369,
+      "step": 15058
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3744748110412934e-05,
+      "loss": 0.6929,
+      "step": 15059
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3743957379853885e-05,
+      "loss": 0.7746,
+      "step": 15060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374316662206836e-05,
+      "loss": 1.004,
+      "step": 15061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3742375837062115e-05,
+      "loss": 0.6815,
+      "step": 15062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3741585024840896e-05,
+      "loss": 1.1492,
+      "step": 15063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3740794185410453e-05,
+      "loss": 1.0192,
+      "step": 15064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3740003318776542e-05,
+      "loss": 1.2509,
+      "step": 15065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3739212424944912e-05,
+      "loss": 0.8181,
+      "step": 15066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3738421503921311e-05,
+      "loss": 0.9399,
+      "step": 15067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3737630555711499e-05,
+      "loss": 0.7678,
+      "step": 15068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3736839580321219e-05,
+      "loss": 1.0194,
+      "step": 15069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3736048577756225e-05,
+      "loss": 0.9188,
+      "step": 15070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3735257548022271e-05,
+      "loss": 0.8018,
+      "step": 15071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3734466491125113e-05,
+      "loss": 0.9296,
+      "step": 15072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3733675407070497e-05,
+      "loss": 0.8656,
+      "step": 15073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3732884295864182e-05,
+      "loss": 1.1963,
+      "step": 15074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3732093157511914e-05,
+      "loss": 0.667,
+      "step": 15075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.373130199201945e-05,
+      "loss": 0.9991,
+      "step": 15076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3730510799392547e-05,
+      "loss": 0.7449,
+      "step": 15077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3729719579636955e-05,
+      "loss": 0.7465,
+      "step": 15078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3728928332758425e-05,
+      "loss": 0.8037,
+      "step": 15079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3728137058762715e-05,
+      "loss": 1.0001,
+      "step": 15080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3727345757655579e-05,
+      "loss": 0.9131,
+      "step": 15081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372655442944277e-05,
+      "loss": 0.9489,
+      "step": 15082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3725763074130045e-05,
+      "loss": 0.69,
+      "step": 15083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3724971691723157e-05,
+      "loss": 1.3484,
+      "step": 15084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372418028222786e-05,
+      "loss": 1.0555,
+      "step": 15085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372338884564991e-05,
+      "loss": 0.8601,
+      "step": 15086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3722597381995064e-05,
+      "loss": 0.9472,
+      "step": 15087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3721805891269075e-05,
+      "loss": 0.8444,
+      "step": 15088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3721014373477704e-05,
+      "loss": 0.9689,
+      "step": 15089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3720222828626699e-05,
+      "loss": 0.9752,
+      "step": 15090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371943125672182e-05,
+      "loss": 0.974,
+      "step": 15091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3718639657768829e-05,
+      "loss": 1.4034,
+      "step": 15092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3717848031773475e-05,
+      "loss": 1.0877,
+      "step": 15093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3717056378741515e-05,
+      "loss": 0.8539,
+      "step": 15094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3716264698678709e-05,
+      "loss": 0.8025,
+      "step": 15095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3715472991590817e-05,
+      "loss": 1.1091,
+      "step": 15096
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3714681257483587e-05,
+      "loss": 0.8072,
+      "step": 15097
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3713889496362789e-05,
+      "loss": 0.7322,
+      "step": 15098
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371309770823417e-05,
+      "loss": 0.8009,
+      "step": 15099
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3712305893103492e-05,
+      "loss": 0.8692,
+      "step": 15100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3711514050976514e-05,
+      "loss": 0.6848,
+      "step": 15101
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3710722181858995e-05,
+      "loss": 1.1509,
+      "step": 15102
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370993028575669e-05,
+      "loss": 0.8131,
+      "step": 15103
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3709138362675362e-05,
+      "loss": 0.6991,
+      "step": 15104
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3708346412620768e-05,
+      "loss": 1.0216,
+      "step": 15105
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3707554435598665e-05,
+      "loss": 1.0764,
+      "step": 15106
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3706762431614815e-05,
+      "loss": 0.9684,
+      "step": 15107
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370597040067498e-05,
+      "loss": 0.8769,
+      "step": 15108
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3705178342784913e-05,
+      "loss": 1.1485,
+      "step": 15109
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3704386257950378e-05,
+      "loss": 0.9866,
+      "step": 15110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3703594146177137e-05,
+      "loss": 0.9024,
+      "step": 15111
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3702802007470945e-05,
+      "loss": 0.9033,
+      "step": 15112
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3702009841837569e-05,
+      "loss": 0.781,
+      "step": 15113
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3701217649282765e-05,
+      "loss": 0.9032,
+      "step": 15114
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3700425429812292e-05,
+      "loss": 0.9854,
+      "step": 15115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3699633183431917e-05,
+      "loss": 0.9989,
+      "step": 15116
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36988409101474e-05,
+      "loss": 1.0118,
+      "step": 15117
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36980486099645e-05,
+      "loss": 0.8126,
+      "step": 15118
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3697256282888982e-05,
+      "loss": 0.7867,
+      "step": 15119
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3696463928926602e-05,
+      "loss": 0.8328,
+      "step": 15120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369567154808313e-05,
+      "loss": 0.8422,
+      "step": 15121
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3694879140364319e-05,
+      "loss": 0.9377,
+      "step": 15122
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369408670577594e-05,
+      "loss": 0.9482,
+      "step": 15123
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3693294244323752e-05,
+      "loss": 1.127,
+      "step": 15124
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3692501756013517e-05,
+      "loss": 0.954,
+      "step": 15125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3691709240850999e-05,
+      "loss": 1.0459,
+      "step": 15126
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3690916698841963e-05,
+      "loss": 0.931,
+      "step": 15127
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3690124129992171e-05,
+      "loss": 0.9572,
+      "step": 15128
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3689331534307386e-05,
+      "loss": 0.9089,
+      "step": 15129
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3688538911793371e-05,
+      "loss": 0.9508,
+      "step": 15130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368774626245589e-05,
+      "loss": 0.9753,
+      "step": 15131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686953586300713e-05,
+      "loss": 0.9736,
+      "step": 15132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686160883333596e-05,
+      "loss": 0.9262,
+      "step": 15133
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3685368153560313e-05,
+      "loss": 0.861,
+      "step": 15134
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3684575396986622e-05,
+      "loss": 1.4741,
+      "step": 15135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3683782613618285e-05,
+      "loss": 1.1154,
+      "step": 15136
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3682989803461074e-05,
+      "loss": 0.9875,
+      "step": 15137
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3682196966520756e-05,
+      "loss": 0.9186,
+      "step": 15138
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3681404102803088e-05,
+      "loss": 1.1936,
+      "step": 15139
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368061121231384e-05,
+      "loss": 0.8202,
+      "step": 15140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3679818295058782e-05,
+      "loss": 1.0268,
+      "step": 15141
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3679025351043678e-05,
+      "loss": 1.0791,
+      "step": 15142
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367823238027429e-05,
+      "loss": 0.7288,
+      "step": 15143
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3677439382756388e-05,
+      "loss": 0.8485,
+      "step": 15144
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676646358495739e-05,
+      "loss": 1.1208,
+      "step": 15145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3675853307498109e-05,
+      "loss": 0.9042,
+      "step": 15146
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3675060229769264e-05,
+      "loss": 0.939,
+      "step": 15147
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3674267125314977e-05,
+      "loss": 1.1188,
+      "step": 15148
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3673473994141008e-05,
+      "loss": 1.0321,
+      "step": 15149
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3672680836253129e-05,
+      "loss": 1.0077,
+      "step": 15150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3671887651657107e-05,
+      "loss": 1.053,
+      "step": 15151
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367109444035871e-05,
+      "loss": 0.853,
+      "step": 15152
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.367030120236371e-05,
+      "loss": 0.9777,
+      "step": 15153
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3669507937677869e-05,
+      "loss": 0.7528,
+      "step": 15154
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3668714646306957e-05,
+      "loss": 0.781,
+      "step": 15155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3667921328256749e-05,
+      "loss": 0.6872,
+      "step": 15156
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3667127983533007e-05,
+      "loss": 1.0293,
+      "step": 15157
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3666334612141503e-05,
+      "loss": 0.9777,
+      "step": 15158
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3665541214088008e-05,
+      "loss": 0.8352,
+      "step": 15159
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3664747789378288e-05,
+      "loss": 0.8458,
+      "step": 15160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3663954338018116e-05,
+      "loss": 1.4283,
+      "step": 15161
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.366316086001326e-05,
+      "loss": 1.115,
+      "step": 15162
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3662367355369496e-05,
+      "loss": 1.0198,
+      "step": 15163
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3661573824092587e-05,
+      "loss": 0.8124,
+      "step": 15164
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3660780266188306e-05,
+      "loss": 1.1211,
+      "step": 15165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3659986681662425e-05,
+      "loss": 0.9337,
+      "step": 15166
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3659193070520713e-05,
+      "loss": 0.9711,
+      "step": 15167
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3658399432768947e-05,
+      "loss": 0.8629,
+      "step": 15168
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3657605768412892e-05,
+      "loss": 0.842,
+      "step": 15169
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3656812077458318e-05,
+      "loss": 0.8763,
+      "step": 15170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3656018359911003e-05,
+      "loss": 1.3077,
+      "step": 15171
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3655224615776717e-05,
+      "loss": 0.9668,
+      "step": 15172
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3654430845061232e-05,
+      "loss": 1.0042,
+      "step": 15173
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365363704777032e-05,
+      "loss": 1.0664,
+      "step": 15174
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3652843223909753e-05,
+      "loss": 0.7437,
+      "step": 15175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3652049373485303e-05,
+      "loss": 1.0633,
+      "step": 15176
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3651255496502746e-05,
+      "loss": 1.1268,
+      "step": 15177
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3650461592967857e-05,
+      "loss": 1.1168,
+      "step": 15178
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3649667662886401e-05,
+      "loss": 0.7811,
+      "step": 15179
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648873706264159e-05,
+      "loss": 1.3539,
+      "step": 15180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648079723106899e-05,
+      "loss": 0.9474,
+      "step": 15181
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3647285713420401e-05,
+      "loss": 0.6233,
+      "step": 15182
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3646491677210437e-05,
+      "loss": 0.9927,
+      "step": 15183
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3645697614482778e-05,
+      "loss": 0.9515,
+      "step": 15184
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.36449035252432e-05,
+      "loss": 0.8654,
+      "step": 15185
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3644109409497484e-05,
+      "loss": 0.8756,
+      "step": 15186
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3643315267251395e-05,
+      "loss": 1.0192,
+      "step": 15187
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3642521098510715e-05,
+      "loss": 0.6399,
+      "step": 15188
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3641726903281218e-05,
+      "loss": 0.9315,
+      "step": 15189
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3640932681568677e-05,
+      "loss": 1.1159,
+      "step": 15190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3640138433378872e-05,
+      "loss": 1.121,
+      "step": 15191
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3639344158717571e-05,
+      "loss": 0.8503,
+      "step": 15192
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363854985759056e-05,
+      "loss": 0.8916,
+      "step": 15193
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3637755530003604e-05,
+      "loss": 1.0708,
+      "step": 15194
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363696117596249e-05,
+      "loss": 1.1824,
+      "step": 15195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3636166795472992e-05,
+      "loss": 0.8338,
+      "step": 15196
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3635372388540884e-05,
+      "loss": 0.8486,
+      "step": 15197
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3634577955171943e-05,
+      "loss": 1.1016,
+      "step": 15198
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363378349537195e-05,
+      "loss": 0.9274,
+      "step": 15199
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632989009146678e-05,
+      "loss": 0.8918,
+      "step": 15200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632194496501904e-05,
+      "loss": 0.7449,
+      "step": 15201
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3631399957443413e-05,
+      "loss": 0.9373,
+      "step": 15202
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3630605391976976e-05,
+      "loss": 1.0,
+      "step": 15203
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3629810800108374e-05,
+      "loss": 0.7183,
+      "step": 15204
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3629016181843384e-05,
+      "loss": 1.0738,
+      "step": 15205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3628221537187787e-05,
+      "loss": 0.6954,
+      "step": 15206
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3627426866147357e-05,
+      "loss": 0.9727,
+      "step": 15207
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362663216872788e-05,
+      "loss": 1.1046,
+      "step": 15208
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362583744493513e-05,
+      "loss": 0.6916,
+      "step": 15209
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3625042694774886e-05,
+      "loss": 0.9725,
+      "step": 15210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3624247918252929e-05,
+      "loss": 1.0083,
+      "step": 15211
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3623453115375038e-05,
+      "loss": 0.8385,
+      "step": 15212
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3622658286146997e-05,
+      "loss": 0.6297,
+      "step": 15213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362186343057458e-05,
+      "loss": 0.8715,
+      "step": 15214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3621068548663572e-05,
+      "loss": 0.8697,
+      "step": 15215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3620273640419748e-05,
+      "loss": 0.8642,
+      "step": 15216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3619478705848895e-05,
+      "loss": 0.9478,
+      "step": 15217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3618683744956789e-05,
+      "loss": 1.1413,
+      "step": 15218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3617888757749213e-05,
+      "loss": 1.0767,
+      "step": 15219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3617093744231951e-05,
+      "loss": 0.9719,
+      "step": 15220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3616298704410778e-05,
+      "loss": 1.0879,
+      "step": 15221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615503638291478e-05,
+      "loss": 0.9259,
+      "step": 15222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361470854587984e-05,
+      "loss": 0.7123,
+      "step": 15223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3613913427181637e-05,
+      "loss": 0.9346,
+      "step": 15224
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3613118282202653e-05,
+      "loss": 0.8605,
+      "step": 15225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3612323110948673e-05,
+      "loss": 1.0253,
+      "step": 15226
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3611527913425477e-05,
+      "loss": 0.817,
+      "step": 15227
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3610732689638853e-05,
+      "loss": 1.0969,
+      "step": 15228
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3609937439594576e-05,
+      "loss": 1.0928,
+      "step": 15229
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3609142163298433e-05,
+      "loss": 1.1673,
+      "step": 15230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3608346860756208e-05,
+      "loss": 1.0651,
+      "step": 15231
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3607551531973683e-05,
+      "loss": 1.1079,
+      "step": 15232
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3606756176956646e-05,
+      "loss": 0.8549,
+      "step": 15233
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3605960795710876e-05,
+      "loss": 1.337,
+      "step": 15234
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3605165388242157e-05,
+      "loss": 0.9876,
+      "step": 15235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3604369954556275e-05,
+      "loss": 0.7405,
+      "step": 15236
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3603574494659018e-05,
+      "loss": 0.9995,
+      "step": 15237
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3602779008556163e-05,
+      "loss": 0.9496,
+      "step": 15238
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3601983496253502e-05,
+      "loss": 1.2782,
+      "step": 15239
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3601187957756814e-05,
+      "loss": 1.1126,
+      "step": 15240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3600392393071888e-05,
+      "loss": 1.0219,
+      "step": 15241
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3599596802204511e-05,
+      "loss": 0.9427,
+      "step": 15242
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3598801185160464e-05,
+      "loss": 0.704,
+      "step": 15243
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3598005541945538e-05,
+      "loss": 0.823,
+      "step": 15244
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3597209872565512e-05,
+      "loss": 0.7143,
+      "step": 15245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359641417702618e-05,
+      "loss": 1.0334,
+      "step": 15246
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3595618455333321e-05,
+      "loss": 0.9405,
+      "step": 15247
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3594822707492732e-05,
+      "loss": 1.1105,
+      "step": 15248
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3594026933510185e-05,
+      "loss": 1.088,
+      "step": 15249
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359323113339148e-05,
+      "loss": 0.923,
+      "step": 15250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3592435307142398e-05,
+      "loss": 1.2921,
+      "step": 15251
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3591639454768729e-05,
+      "loss": 1.021,
+      "step": 15252
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3590843576276256e-05,
+      "loss": 0.7371,
+      "step": 15253
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3590047671670771e-05,
+      "loss": 0.7702,
+      "step": 15254
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358925174095806e-05,
+      "loss": 0.9692,
+      "step": 15255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3588455784143913e-05,
+      "loss": 1.0431,
+      "step": 15256
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587659801234116e-05,
+      "loss": 0.7098,
+      "step": 15257
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358686379223446e-05,
+      "loss": 0.885,
+      "step": 15258
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3586067757150732e-05,
+      "loss": 0.9981,
+      "step": 15259
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358527169598872e-05,
+      "loss": 0.808,
+      "step": 15260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3584475608754216e-05,
+      "loss": 1.2761,
+      "step": 15261
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3583679495453e-05,
+      "loss": 0.6587,
+      "step": 15262
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358288335609088e-05,
+      "loss": 0.9433,
+      "step": 15263
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358208719067363e-05,
+      "loss": 1.0811,
+      "step": 15264
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.358129099920704e-05,
+      "loss": 0.999,
+      "step": 15265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3580494781696908e-05,
+      "loss": 0.8897,
+      "step": 15266
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357969853814902e-05,
+      "loss": 1.0481,
+      "step": 15267
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578902268569167e-05,
+      "loss": 0.7691,
+      "step": 15268
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578105972963138e-05,
+      "loss": 0.925,
+      "step": 15269
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3577309651336728e-05,
+      "loss": 0.8349,
+      "step": 15270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.357651330369572e-05,
+      "loss": 0.73,
+      "step": 15271
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3575716930045913e-05,
+      "loss": 0.681,
+      "step": 15272
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3574920530393096e-05,
+      "loss": 0.709,
+      "step": 15273
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3574124104743057e-05,
+      "loss": 1.288,
+      "step": 15274
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3573327653101593e-05,
+      "loss": 1.0581,
+      "step": 15275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3572531175474495e-05,
+      "loss": 0.9452,
+      "step": 15276
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3571734671867549e-05,
+      "loss": 0.9611,
+      "step": 15277
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3570938142286556e-05,
+      "loss": 0.722,
+      "step": 15278
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3570141586737305e-05,
+      "loss": 0.9584,
+      "step": 15279
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3569345005225582e-05,
+      "loss": 1.1704,
+      "step": 15280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356854839775719e-05,
+      "loss": 0.7843,
+      "step": 15281
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356775176433792e-05,
+      "loss": 1.0471,
+      "step": 15282
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3566955104973558e-05,
+      "loss": 1.5564,
+      "step": 15283
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3566158419669906e-05,
+      "loss": 1.2331,
+      "step": 15284
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565361708432754e-05,
+      "loss": 1.2808,
+      "step": 15285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356456497126789e-05,
+      "loss": 1.2217,
+      "step": 15286
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3563768208181118e-05,
+      "loss": 1.1485,
+      "step": 15287
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3562971419178233e-05,
+      "loss": 1.1399,
+      "step": 15288
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3562174604265015e-05,
+      "loss": 0.7589,
+      "step": 15289
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3561377763447273e-05,
+      "loss": 1.4404,
+      "step": 15290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3560580896730795e-05,
+      "loss": 1.1377,
+      "step": 15291
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3559784004121377e-05,
+      "loss": 0.7785,
+      "step": 15292
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3558987085624817e-05,
+      "loss": 0.8288,
+      "step": 15293
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3558190141246906e-05,
+      "loss": 0.8836,
+      "step": 15294
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3557393170993439e-05,
+      "loss": 1.3584,
+      "step": 15295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3556596174870215e-05,
+      "loss": 1.0507,
+      "step": 15296
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3555799152883029e-05,
+      "loss": 0.9182,
+      "step": 15297
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3555002105037678e-05,
+      "loss": 1.1264,
+      "step": 15298
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3554205031339955e-05,
+      "loss": 1.3612,
+      "step": 15299
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3553407931795662e-05,
+      "loss": 1.2657,
+      "step": 15300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3552610806410586e-05,
+      "loss": 1.1361,
+      "step": 15301
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3551813655190534e-05,
+      "loss": 0.896,
+      "step": 15302
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35510164781413e-05,
+      "loss": 0.7856,
+      "step": 15303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3550219275268674e-05,
+      "loss": 0.7859,
+      "step": 15304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3549422046578465e-05,
+      "loss": 0.9862,
+      "step": 15305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3548624792076462e-05,
+      "loss": 1.1558,
+      "step": 15306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3547827511768465e-05,
+      "loss": 0.9675,
+      "step": 15307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3547030205660273e-05,
+      "loss": 0.9519,
+      "step": 15308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3546232873757684e-05,
+      "loss": 0.8329,
+      "step": 15309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3545435516066493e-05,
+      "loss": 0.9628,
+      "step": 15310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3544638132592504e-05,
+      "loss": 0.8062,
+      "step": 15311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3543840723341512e-05,
+      "loss": 1.1165,
+      "step": 15312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354304328831932e-05,
+      "loss": 0.8561,
+      "step": 15313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354224582753172e-05,
+      "loss": 1.032,
+      "step": 15314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3541448340984516e-05,
+      "loss": 0.8149,
+      "step": 15315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3540650828683505e-05,
+      "loss": 1.0659,
+      "step": 15316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3539853290634489e-05,
+      "loss": 1.2266,
+      "step": 15317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3539055726843269e-05,
+      "loss": 0.7272,
+      "step": 15318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3538258137315639e-05,
+      "loss": 0.8151,
+      "step": 15319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3537460522057404e-05,
+      "loss": 0.8888,
+      "step": 15320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3536662881074364e-05,
+      "loss": 1.1013,
+      "step": 15321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353586521437232e-05,
+      "loss": 1.2656,
+      "step": 15322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353506752195707e-05,
+      "loss": 1.0474,
+      "step": 15323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3534269803834418e-05,
+      "loss": 0.6751,
+      "step": 15324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3533472060010162e-05,
+      "loss": 0.7958,
+      "step": 15325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3532674290490104e-05,
+      "loss": 1.0613,
+      "step": 15326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3531876495280047e-05,
+      "loss": 0.7252,
+      "step": 15327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3531078674385794e-05,
+      "loss": 1.1162,
+      "step": 15328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353028082781314e-05,
+      "loss": 1.1471,
+      "step": 15329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3529482955567896e-05,
+      "loss": 0.8448,
+      "step": 15330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352868505765586e-05,
+      "loss": 1.0205,
+      "step": 15331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3527887134082831e-05,
+      "loss": 0.9153,
+      "step": 15332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3527089184854618e-05,
+      "loss": 1.1025,
+      "step": 15333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3526291209977021e-05,
+      "loss": 0.9196,
+      "step": 15334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352549320945584e-05,
+      "loss": 0.9941,
+      "step": 15335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3524695183296884e-05,
+      "loss": 1.1243,
+      "step": 15336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3523897131505952e-05,
+      "loss": 0.8367,
+      "step": 15337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3523099054088849e-05,
+      "loss": 1.2333,
+      "step": 15338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3522300951051377e-05,
+      "loss": 1.0875,
+      "step": 15339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3521502822399342e-05,
+      "loss": 1.0239,
+      "step": 15340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3520704668138547e-05,
+      "loss": 0.9785,
+      "step": 15341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3519906488274796e-05,
+      "loss": 0.7027,
+      "step": 15342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35191082828139e-05,
+      "loss": 0.9515,
+      "step": 15343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3518310051761651e-05,
+      "loss": 0.8203,
+      "step": 15344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3517511795123864e-05,
+      "loss": 0.9718,
+      "step": 15345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3516713512906338e-05,
+      "loss": 0.923,
+      "step": 15346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3515915205114883e-05,
+      "loss": 0.9694,
+      "step": 15347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3515116871755302e-05,
+      "loss": 1.0219,
+      "step": 15348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3514318512833402e-05,
+      "loss": 0.8788,
+      "step": 15349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3513520128354982e-05,
+      "loss": 1.1912,
+      "step": 15350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3512721718325858e-05,
+      "loss": 0.992,
+      "step": 15351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.351192328275183e-05,
+      "loss": 0.7899,
+      "step": 15352
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3511124821638709e-05,
+      "loss": 0.8444,
+      "step": 15353
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510326334992296e-05,
+      "loss": 0.8414,
+      "step": 15354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.35095278228184e-05,
+      "loss": 1.1386,
+      "step": 15355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3508729285122826e-05,
+      "loss": 0.7708,
+      "step": 15356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3507930721911387e-05,
+      "loss": 1.122,
+      "step": 15357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3507132133189886e-05,
+      "loss": 1.0432,
+      "step": 15358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3506333518964127e-05,
+      "loss": 1.3314,
+      "step": 15359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3505534879239923e-05,
+      "loss": 0.4972,
+      "step": 15360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3504736214023082e-05,
+      "loss": 0.8098,
+      "step": 15361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3503937523319405e-05,
+      "loss": 1.0999,
+      "step": 15362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3503138807134712e-05,
+      "loss": 1.1552,
+      "step": 15363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3502340065474802e-05,
+      "loss": 0.9783,
+      "step": 15364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3501541298345483e-05,
+      "loss": 0.8851,
+      "step": 15365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350074250575257e-05,
+      "loss": 0.9464,
+      "step": 15366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349994368770187e-05,
+      "loss": 1.1461,
+      "step": 15367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3499144844199188e-05,
+      "loss": 0.9947,
+      "step": 15368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3498345975250338e-05,
+      "loss": 0.992,
+      "step": 15369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3497547080861127e-05,
+      "loss": 0.7029,
+      "step": 15370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349674816103736e-05,
+      "loss": 0.9625,
+      "step": 15371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3495949215784858e-05,
+      "loss": 0.7095,
+      "step": 15372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3495150245109428e-05,
+      "loss": 1.1877,
+      "step": 15373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349435124901687e-05,
+      "loss": 0.829,
+      "step": 15374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3493552227513007e-05,
+      "loss": 0.9861,
+      "step": 15375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3492753180603642e-05,
+      "loss": 1.2412,
+      "step": 15376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3491954108294588e-05,
+      "loss": 0.9599,
+      "step": 15377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3491155010591656e-05,
+      "loss": 1.0656,
+      "step": 15378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3490355887500659e-05,
+      "loss": 0.6221,
+      "step": 15379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.34895567390274e-05,
+      "loss": 1.0302,
+      "step": 15380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3488757565177704e-05,
+      "loss": 0.9932,
+      "step": 15381
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3487958365957372e-05,
+      "loss": 1.0597,
+      "step": 15382
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3487159141372221e-05,
+      "loss": 1.0557,
+      "step": 15383
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3486359891428063e-05,
+      "loss": 1.0132,
+      "step": 15384
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3485560616130706e-05,
+      "loss": 0.9052,
+      "step": 15385
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3484761315485966e-05,
+      "loss": 0.772,
+      "step": 15386
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3483961989499652e-05,
+      "loss": 1.0481,
+      "step": 15387
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3483162638177585e-05,
+      "loss": 0.6809,
+      "step": 15388
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3482363261525568e-05,
+      "loss": 0.6998,
+      "step": 15389
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.348156385954942e-05,
+      "loss": 1.072,
+      "step": 15390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3480764432254953e-05,
+      "loss": 0.9041,
+      "step": 15391
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3479964979647981e-05,
+      "loss": 0.7843,
+      "step": 15392
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3479165501734319e-05,
+      "loss": 1.002,
+      "step": 15393
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3478365998519776e-05,
+      "loss": 1.1236,
+      "step": 15394
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3477566470010173e-05,
+      "loss": 0.9937,
+      "step": 15395
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3476766916211315e-05,
+      "loss": 0.9222,
+      "step": 15396
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475967337129024e-05,
+      "loss": 1.0763,
+      "step": 15397
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475167732769115e-05,
+      "loss": 0.8525,
+      "step": 15398
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3474368103137398e-05,
+      "loss": 1.0645,
+      "step": 15399
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473568448239692e-05,
+      "loss": 1.1772,
+      "step": 15400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3472768768081807e-05,
+      "loss": 0.9679,
+      "step": 15401
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3471969062669565e-05,
+      "loss": 0.7931,
+      "step": 15402
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.347116933200878e-05,
+      "loss": 0.8209,
+      "step": 15403
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3470369576105262e-05,
+      "loss": 1.0352,
+      "step": 15404
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3469569794964832e-05,
+      "loss": 1.0042,
+      "step": 15405
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3468769988593307e-05,
+      "loss": 0.9788,
+      "step": 15406
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3467970156996501e-05,
+      "loss": 0.9443,
+      "step": 15407
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.346717030018023e-05,
+      "loss": 1.014,
+      "step": 15408
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3466370418150315e-05,
+      "loss": 0.8476,
+      "step": 15409
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3465570510912564e-05,
+      "loss": 0.705,
+      "step": 15410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3464770578472801e-05,
+      "loss": 0.9772,
+      "step": 15411
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3463970620836842e-05,
+      "loss": 0.9871,
+      "step": 15412
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3463170638010508e-05,
+      "loss": 0.7657,
+      "step": 15413
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3462370629999606e-05,
+      "loss": 0.8886,
+      "step": 15414
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3461570596809963e-05,
+      "loss": 0.7783,
+      "step": 15415
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3460770538447396e-05,
+      "loss": 1.1028,
+      "step": 15416
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345997045491772e-05,
+      "loss": 1.0059,
+      "step": 15417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3459170346226753e-05,
+      "loss": 1.5775,
+      "step": 15418
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345837021238032e-05,
+      "loss": 1.4172,
+      "step": 15419
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3457570053384225e-05,
+      "loss": 1.3175,
+      "step": 15420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3456769869244304e-05,
+      "loss": 0.7899,
+      "step": 15421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345596965996637e-05,
+      "loss": 0.9224,
+      "step": 15422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3455169425556239e-05,
+      "loss": 1.1108,
+      "step": 15423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345436916601973e-05,
+      "loss": 0.9061,
+      "step": 15424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3453568881362666e-05,
+      "loss": 1.1139,
+      "step": 15425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3452768571590865e-05,
+      "loss": 1.0486,
+      "step": 15426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345196823671015e-05,
+      "loss": 1.0553,
+      "step": 15427
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3451167876726338e-05,
+      "loss": 1.0444,
+      "step": 15428
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3450367491645248e-05,
+      "loss": 1.0473,
+      "step": 15429
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3449567081472705e-05,
+      "loss": 1.2042,
+      "step": 15430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3448766646214527e-05,
+      "loss": 1.1134,
+      "step": 15431
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447966185876533e-05,
+      "loss": 1.0564,
+      "step": 15432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447165700464546e-05,
+      "loss": 1.003,
+      "step": 15433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3446365189984389e-05,
+      "loss": 0.9552,
+      "step": 15434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3445564654441879e-05,
+      "loss": 1.1147,
+      "step": 15435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3444764093842841e-05,
+      "loss": 0.9657,
+      "step": 15436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443963508193098e-05,
+      "loss": 0.9471,
+      "step": 15437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344316289749847e-05,
+      "loss": 0.8128,
+      "step": 15438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3442362261764776e-05,
+      "loss": 0.9432,
+      "step": 15439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3441561600997844e-05,
+      "loss": 0.7002,
+      "step": 15440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344076091520349e-05,
+      "loss": 0.9664,
+      "step": 15441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3439960204387548e-05,
+      "loss": 0.9997,
+      "step": 15442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3439159468555829e-05,
+      "loss": 0.7798,
+      "step": 15443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3438358707714159e-05,
+      "loss": 1.1491,
+      "step": 15444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3437557921868363e-05,
+      "loss": 0.747,
+      "step": 15445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3436757111024266e-05,
+      "loss": 0.9386,
+      "step": 15446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435956275187687e-05,
+      "loss": 1.0061,
+      "step": 15447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435155414364455e-05,
+      "loss": 0.8769,
+      "step": 15448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3434354528560391e-05,
+      "loss": 0.8198,
+      "step": 15449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3433553617781318e-05,
+      "loss": 0.7821,
+      "step": 15450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3432752682033064e-05,
+      "loss": 0.9011,
+      "step": 15451
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3431951721321452e-05,
+      "loss": 1.0061,
+      "step": 15452
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34311507356523e-05,
+      "loss": 0.7295,
+      "step": 15453
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3430349725031444e-05,
+      "loss": 0.9264,
+      "step": 15454
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3429548689464704e-05,
+      "loss": 0.9463,
+      "step": 15455
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3428747628957904e-05,
+      "loss": 0.7611,
+      "step": 15456
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342794654351687e-05,
+      "loss": 0.8342,
+      "step": 15457
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342714543314743e-05,
+      "loss": 1.0648,
+      "step": 15458
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3426344297855405e-05,
+      "loss": 0.8192,
+      "step": 15459
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3425543137646625e-05,
+      "loss": 0.95,
+      "step": 15460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3424741952526914e-05,
+      "loss": 1.1753,
+      "step": 15461
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3423940742502102e-05,
+      "loss": 1.005,
+      "step": 15462
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3423139507578009e-05,
+      "loss": 1.1999,
+      "step": 15463
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3422338247760468e-05,
+      "loss": 0.7338,
+      "step": 15464
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3421536963055304e-05,
+      "loss": 0.9625,
+      "step": 15465
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.342073565346834e-05,
+      "loss": 1.0805,
+      "step": 15466
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341993431900541e-05,
+      "loss": 1.24,
+      "step": 15467
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3419132959672336e-05,
+      "loss": 0.6187,
+      "step": 15468
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3418331575474945e-05,
+      "loss": 0.8088,
+      "step": 15469
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341753016641907e-05,
+      "loss": 1.5632,
+      "step": 15470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3416728732510535e-05,
+      "loss": 1.0432,
+      "step": 15471
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3415927273755172e-05,
+      "loss": 1.1281,
+      "step": 15472
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3415125790158805e-05,
+      "loss": 1.0747,
+      "step": 15473
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3414324281727262e-05,
+      "loss": 0.9813,
+      "step": 15474
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3413522748466374e-05,
+      "loss": 1.2593,
+      "step": 15475
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3412721190381972e-05,
+      "loss": 0.9211,
+      "step": 15476
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3411919607479881e-05,
+      "loss": 0.7058,
+      "step": 15477
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3411117999765933e-05,
+      "loss": 0.5823,
+      "step": 15478
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3410316367245955e-05,
+      "loss": 1.1205,
+      "step": 15479
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3409514709925777e-05,
+      "loss": 1.2272,
+      "step": 15480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3408713027811229e-05,
+      "loss": 0.7543,
+      "step": 15481
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3407911320908145e-05,
+      "loss": 0.8902,
+      "step": 15482
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340710958922235e-05,
+      "loss": 0.9393,
+      "step": 15483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3406307832759673e-05,
+      "loss": 1.0344,
+      "step": 15484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3405506051525948e-05,
+      "loss": 1.1804,
+      "step": 15485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3404704245527004e-05,
+      "loss": 0.8666,
+      "step": 15486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3403902414768678e-05,
+      "loss": 0.8724,
+      "step": 15487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.340310055925679e-05,
+      "loss": 0.9116,
+      "step": 15488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3402298678997179e-05,
+      "loss": 1.3209,
+      "step": 15489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3401496773995672e-05,
+      "loss": 0.7446,
+      "step": 15490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3400694844258107e-05,
+      "loss": 0.9445,
+      "step": 15491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3399892889790312e-05,
+      "loss": 0.8719,
+      "step": 15492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339909091059811e-05,
+      "loss": 0.8079,
+      "step": 15493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3398288906687349e-05,
+      "loss": 0.8421,
+      "step": 15494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3397486878063852e-05,
+      "loss": 0.8568,
+      "step": 15495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3396684824733451e-05,
+      "loss": 0.9039,
+      "step": 15496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3395882746701985e-05,
+      "loss": 1.1958,
+      "step": 15497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339508064397528e-05,
+      "loss": 1.0517,
+      "step": 15498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3394278516559172e-05,
+      "loss": 0.7905,
+      "step": 15499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3393476364459492e-05,
+      "loss": 0.8527,
+      "step": 15500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3392674187682078e-05,
+      "loss": 0.7963,
+      "step": 15501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3391871986232758e-05,
+      "loss": 0.7698,
+      "step": 15502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.339106976011737e-05,
+      "loss": 0.8585,
+      "step": 15503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3390267509341746e-05,
+      "loss": 1.0564,
+      "step": 15504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3389465233911716e-05,
+      "loss": 1.1935,
+      "step": 15505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3388662933833123e-05,
+      "loss": 0.742,
+      "step": 15506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3387860609111797e-05,
+      "loss": 0.6436,
+      "step": 15507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3387058259753571e-05,
+      "loss": 1.1988,
+      "step": 15508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.338625588576428e-05,
+      "loss": 1.0384,
+      "step": 15509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3385453487149765e-05,
+      "loss": 0.8706,
+      "step": 15510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3384651063915853e-05,
+      "loss": 1.0788,
+      "step": 15511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.338384861606838e-05,
+      "loss": 0.7663,
+      "step": 15512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3383046143613187e-05,
+      "loss": 1.0454,
+      "step": 15513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3382243646556106e-05,
+      "loss": 1.115,
+      "step": 15514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3381441124902974e-05,
+      "loss": 0.9813,
+      "step": 15515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3380638578659625e-05,
+      "loss": 1.1069,
+      "step": 15516
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3379836007831899e-05,
+      "loss": 1.0469,
+      "step": 15517
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3379033412425631e-05,
+      "loss": 1.0716,
+      "step": 15518
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3378230792446657e-05,
+      "loss": 1.0089,
+      "step": 15519
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337742814790081e-05,
+      "loss": 0.7315,
+      "step": 15520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3376625478793932e-05,
+      "loss": 0.9243,
+      "step": 15521
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375822785131863e-05,
+      "loss": 0.793,
+      "step": 15522
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375020066920427e-05,
+      "loss": 0.8756,
+      "step": 15523
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3374217324165477e-05,
+      "loss": 1.1266,
+      "step": 15524
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3373414556872844e-05,
+      "loss": 0.7481,
+      "step": 15525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3372611765048362e-05,
+      "loss": 0.9981,
+      "step": 15526
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3371808948697875e-05,
+      "loss": 0.8455,
+      "step": 15527
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3371006107827217e-05,
+      "loss": 0.9114,
+      "step": 15528
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3370203242442232e-05,
+      "loss": 1.1293,
+      "step": 15529
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336940035254875e-05,
+      "loss": 0.7818,
+      "step": 15530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3368597438152618e-05,
+      "loss": 0.9978,
+      "step": 15531
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336779449925967e-05,
+      "loss": 0.7464,
+      "step": 15532
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3366991535875746e-05,
+      "loss": 0.7112,
+      "step": 15533
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3366188548006686e-05,
+      "loss": 0.9361,
+      "step": 15534
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336538553565833e-05,
+      "loss": 0.9803,
+      "step": 15535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3364582498836515e-05,
+      "loss": 0.7059,
+      "step": 15536
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3363779437547086e-05,
+      "loss": 1.1032,
+      "step": 15537
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3362976351795878e-05,
+      "loss": 0.8144,
+      "step": 15538
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336217324158873e-05,
+      "loss": 1.3139,
+      "step": 15539
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3361370106931486e-05,
+      "loss": 0.7396,
+      "step": 15540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3360566947829986e-05,
+      "loss": 0.9284,
+      "step": 15541
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335976376429007e-05,
+      "loss": 1.0465,
+      "step": 15542
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335896055631758e-05,
+      "loss": 0.7694,
+      "step": 15543
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3358157323918353e-05,
+      "loss": 0.7381,
+      "step": 15544
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3357354067098231e-05,
+      "loss": 0.7737,
+      "step": 15545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3356550785863062e-05,
+      "loss": 0.7847,
+      "step": 15546
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3355747480218684e-05,
+      "loss": 0.7536,
+      "step": 15547
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3354944150170934e-05,
+      "loss": 1.3189,
+      "step": 15548
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3354140795725659e-05,
+      "loss": 0.8634,
+      "step": 15549
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33533374168887e-05,
+      "loss": 0.7354,
+      "step": 15550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3352534013665897e-05,
+      "loss": 1.0321,
+      "step": 15551
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3351730586063096e-05,
+      "loss": 0.9227,
+      "step": 15552
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.335092713408614e-05,
+      "loss": 0.8814,
+      "step": 15553
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3350123657740866e-05,
+      "loss": 0.9443,
+      "step": 15554
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3349320157033121e-05,
+      "loss": 0.7768,
+      "step": 15555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3348516631968749e-05,
+      "loss": 0.7652,
+      "step": 15556
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3347713082553593e-05,
+      "loss": 0.8729,
+      "step": 15557
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3346909508793495e-05,
+      "loss": 0.7964,
+      "step": 15558
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33461059106943e-05,
+      "loss": 1.0468,
+      "step": 15559
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334530228826185e-05,
+      "loss": 0.827,
+      "step": 15560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334449864150199e-05,
+      "loss": 0.8162,
+      "step": 15561
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3343694970420568e-05,
+      "loss": 0.7102,
+      "step": 15562
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3342891275023422e-05,
+      "loss": 0.9746,
+      "step": 15563
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33420875553164e-05,
+      "loss": 1.2917,
+      "step": 15564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3341283811305346e-05,
+      "loss": 0.7558,
+      "step": 15565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3340480042996103e-05,
+      "loss": 1.3536,
+      "step": 15566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3339676250394521e-05,
+      "loss": 1.0014,
+      "step": 15567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3338872433506442e-05,
+      "loss": 0.9151,
+      "step": 15568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333806859233771e-05,
+      "loss": 1.0703,
+      "step": 15569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3337264726894175e-05,
+      "loss": 0.8392,
+      "step": 15570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3336460837181679e-05,
+      "loss": 1.0256,
+      "step": 15571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3335656923206071e-05,
+      "loss": 0.9705,
+      "step": 15572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3334852984973195e-05,
+      "loss": 0.8746,
+      "step": 15573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3334049022488895e-05,
+      "loss": 0.8746,
+      "step": 15574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3333245035759022e-05,
+      "loss": 1.3387,
+      "step": 15575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333244102478942e-05,
+      "loss": 1.1095,
+      "step": 15576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333163698958594e-05,
+      "loss": 0.7203,
+      "step": 15577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3330832930154422e-05,
+      "loss": 0.7474,
+      "step": 15578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333002884650072e-05,
+      "loss": 1.175,
+      "step": 15579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3329224738630677e-05,
+      "loss": 0.9705,
+      "step": 15580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3328420606550143e-05,
+      "loss": 1.5098,
+      "step": 15581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3327616450264964e-05,
+      "loss": 0.6074,
+      "step": 15582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3326812269780988e-05,
+      "loss": 0.9123,
+      "step": 15583
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3326008065104063e-05,
+      "loss": 0.8667,
+      "step": 15584
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3325203836240039e-05,
+      "loss": 0.9041,
+      "step": 15585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3324399583194766e-05,
+      "loss": 1.0272,
+      "step": 15586
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3323595305974086e-05,
+      "loss": 0.7572,
+      "step": 15587
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3322791004583855e-05,
+      "loss": 0.7254,
+      "step": 15588
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3321986679029917e-05,
+      "loss": 1.0678,
+      "step": 15589
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3321182329318124e-05,
+      "loss": 1.1051,
+      "step": 15590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320377955454322e-05,
+      "loss": 0.7026,
+      "step": 15591
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3319573557444368e-05,
+      "loss": 1.2294,
+      "step": 15592
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3318769135294098e-05,
+      "loss": 0.9629,
+      "step": 15593
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3317964689009378e-05,
+      "loss": 1.2528,
+      "step": 15594
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3317160218596048e-05,
+      "loss": 1.1105,
+      "step": 15595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3316355724059959e-05,
+      "loss": 0.8572,
+      "step": 15596
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3315551205406965e-05,
+      "loss": 0.8212,
+      "step": 15597
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3314746662642915e-05,
+      "loss": 0.9353,
+      "step": 15598
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3313942095773657e-05,
+      "loss": 0.841,
+      "step": 15599
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3313137504805042e-05,
+      "loss": 0.8851,
+      "step": 15600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3312332889742927e-05,
+      "loss": 0.9567,
+      "step": 15601
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331152825059316e-05,
+      "loss": 0.8378,
+      "step": 15602
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331072358736159e-05,
+      "loss": 0.6924,
+      "step": 15603
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3309918900054071e-05,
+      "loss": 1.1537,
+      "step": 15604
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3309114188676453e-05,
+      "loss": 0.8864,
+      "step": 15605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330830945323459e-05,
+      "loss": 1.5533,
+      "step": 15606
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3307504693734334e-05,
+      "loss": 0.7772,
+      "step": 15607
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3306699910181535e-05,
+      "loss": 1.2205,
+      "step": 15608
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3305895102582046e-05,
+      "loss": 0.9527,
+      "step": 15609
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3305090270941725e-05,
+      "loss": 0.8657,
+      "step": 15610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3304285415266418e-05,
+      "loss": 1.0351,
+      "step": 15611
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3303480535561981e-05,
+      "loss": 0.8491,
+      "step": 15612
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3302675631834265e-05,
+      "loss": 0.9766,
+      "step": 15613
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3301870704089126e-05,
+      "loss": 0.8414,
+      "step": 15614
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3301065752332415e-05,
+      "loss": 0.9697,
+      "step": 15615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3300260776569992e-05,
+      "loss": 1.2328,
+      "step": 15616
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3299455776807704e-05,
+      "loss": 0.8591,
+      "step": 15617
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298650753051405e-05,
+      "loss": 0.6609,
+      "step": 15618
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3297845705306954e-05,
+      "loss": 1.035,
+      "step": 15619
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3297040633580203e-05,
+      "loss": 0.7915,
+      "step": 15620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3296235537877003e-05,
+      "loss": 0.9965,
+      "step": 15621
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3295430418203217e-05,
+      "loss": 1.0379,
+      "step": 15622
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3294625274564691e-05,
+      "loss": 0.9705,
+      "step": 15623
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3293820106967284e-05,
+      "loss": 1.0616,
+      "step": 15624
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3293014915416852e-05,
+      "loss": 0.7883,
+      "step": 15625
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3292209699919251e-05,
+      "loss": 1.0457,
+      "step": 15626
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3291404460480338e-05,
+      "loss": 0.7397,
+      "step": 15627
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3290599197105963e-05,
+      "loss": 1.0794,
+      "step": 15628
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3289793909801985e-05,
+      "loss": 0.8382,
+      "step": 15629
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328898859857426e-05,
+      "loss": 0.6742,
+      "step": 15630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3288183263428645e-05,
+      "loss": 0.756,
+      "step": 15631
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287377904370999e-05,
+      "loss": 0.8709,
+      "step": 15632
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3286572521407173e-05,
+      "loss": 1.1002,
+      "step": 15633
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3285767114543026e-05,
+      "loss": 0.8528,
+      "step": 15634
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3284961683784418e-05,
+      "loss": 0.9016,
+      "step": 15635
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3284156229137202e-05,
+      "loss": 1.1418,
+      "step": 15636
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328335075060724e-05,
+      "loss": 0.8376,
+      "step": 15637
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3282545248200384e-05,
+      "loss": 0.9167,
+      "step": 15638
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3281739721922495e-05,
+      "loss": 1.3001,
+      "step": 15639
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3280934171779432e-05,
+      "loss": 0.9569,
+      "step": 15640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3280128597777048e-05,
+      "loss": 0.818,
+      "step": 15641
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3279322999921207e-05,
+      "loss": 0.951,
+      "step": 15642
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3278517378217766e-05,
+      "loss": 1.0397,
+      "step": 15643
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3277711732672583e-05,
+      "loss": 0.8213,
+      "step": 15644
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3276906063291511e-05,
+      "loss": 0.9139,
+      "step": 15645
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3276100370080417e-05,
+      "loss": 0.7948,
+      "step": 15646
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327529465304516e-05,
+      "loss": 0.8373,
+      "step": 15647
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3274488912191594e-05,
+      "loss": 0.88,
+      "step": 15648
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3273683147525582e-05,
+      "loss": 0.9264,
+      "step": 15649
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3272877359052982e-05,
+      "loss": 0.7115,
+      "step": 15650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3272071546779654e-05,
+      "loss": 0.9687,
+      "step": 15651
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.327126571071146e-05,
+      "loss": 0.8059,
+      "step": 15652
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3270459850854257e-05,
+      "loss": 0.9834,
+      "step": 15653
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3269653967213906e-05,
+      "loss": 1.2808,
+      "step": 15654
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326884805979627e-05,
+      "loss": 1.0538,
+      "step": 15655
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3268042128607208e-05,
+      "loss": 1.009,
+      "step": 15656
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326723617365258e-05,
+      "loss": 1.0237,
+      "step": 15657
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3266430194938247e-05,
+      "loss": 1.0454,
+      "step": 15658
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3265624192470073e-05,
+      "loss": 0.8533,
+      "step": 15659
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3264818166253917e-05,
+      "loss": 1.2248,
+      "step": 15660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3264012116295636e-05,
+      "loss": 1.2389,
+      "step": 15661
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3263206042601104e-05,
+      "loss": 1.0305,
+      "step": 15662
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326239994517617e-05,
+      "loss": 0.8464,
+      "step": 15663
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3261593824026702e-05,
+      "loss": 0.9454,
+      "step": 15664
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3260787679158562e-05,
+      "loss": 0.9898,
+      "step": 15665
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3259981510577613e-05,
+      "loss": 0.5497,
+      "step": 15666
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3259175318289716e-05,
+      "loss": 0.7691,
+      "step": 15667
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3258369102300735e-05,
+      "loss": 0.7693,
+      "step": 15668
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3257562862616528e-05,
+      "loss": 1.0347,
+      "step": 15669
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3256756599242965e-05,
+      "loss": 0.9105,
+      "step": 15670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255950312185905e-05,
+      "loss": 1.1478,
+      "step": 15671
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255144001451216e-05,
+      "loss": 0.9869,
+      "step": 15672
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3254337667044757e-05,
+      "loss": 0.9184,
+      "step": 15673
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3253531308972392e-05,
+      "loss": 1.721,
+      "step": 15674
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3252724927239986e-05,
+      "loss": 1.1025,
+      "step": 15675
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3251918521853403e-05,
+      "loss": 1.112,
+      "step": 15676
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.325111209281851e-05,
+      "loss": 0.9965,
+      "step": 15677
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3250305640141168e-05,
+      "loss": 1.1615,
+      "step": 15678
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3249499163827237e-05,
+      "loss": 1.04,
+      "step": 15679
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3248692663882591e-05,
+      "loss": 0.8942,
+      "step": 15680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3247886140313094e-05,
+      "loss": 0.8874,
+      "step": 15681
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3247079593124608e-05,
+      "loss": 1.0409,
+      "step": 15682
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3246273022322996e-05,
+      "loss": 0.8761,
+      "step": 15683
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3245466427914127e-05,
+      "loss": 1.2439,
+      "step": 15684
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3244659809903864e-05,
+      "loss": 0.8782,
+      "step": 15685
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3243853168298076e-05,
+      "loss": 1.1107,
+      "step": 15686
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324304650310263e-05,
+      "loss": 1.0551,
+      "step": 15687
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3242239814323386e-05,
+      "loss": 1.039,
+      "step": 15688
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3241433101966214e-05,
+      "loss": 0.9186,
+      "step": 15689
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3240626366036982e-05,
+      "loss": 0.6924,
+      "step": 15690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3239819606541552e-05,
+      "loss": 0.7237,
+      "step": 15691
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3239012823485802e-05,
+      "loss": 1.1396,
+      "step": 15692
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3238206016875586e-05,
+      "loss": 1.0678,
+      "step": 15693
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3237399186716774e-05,
+      "loss": 1.0665,
+      "step": 15694
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3236592333015236e-05,
+      "loss": 0.8587,
+      "step": 15695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3235785455776843e-05,
+      "loss": 0.7789,
+      "step": 15696
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3234978555007459e-05,
+      "loss": 0.6838,
+      "step": 15697
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3234171630712949e-05,
+      "loss": 0.8332,
+      "step": 15698
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3233364682899183e-05,
+      "loss": 0.5891,
+      "step": 15699
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3232557711572032e-05,
+      "loss": 0.8306,
+      "step": 15700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3231750716737361e-05,
+      "loss": 0.856,
+      "step": 15701
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3230943698401042e-05,
+      "loss": 1.097,
+      "step": 15702
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.323013665656894e-05,
+      "loss": 0.7659,
+      "step": 15703
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3229329591246927e-05,
+      "loss": 1.0973,
+      "step": 15704
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3228522502440868e-05,
+      "loss": 0.712,
+      "step": 15705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3227715390156635e-05,
+      "loss": 0.9189,
+      "step": 15706
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3226908254400102e-05,
+      "loss": 0.8486,
+      "step": 15707
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3226101095177128e-05,
+      "loss": 0.692,
+      "step": 15708
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.322529391249359e-05,
+      "loss": 0.7837,
+      "step": 15709
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3224486706355357e-05,
+      "loss": 0.7161,
+      "step": 15710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3223679476768301e-05,
+      "loss": 0.8677,
+      "step": 15711
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3222872223738288e-05,
+      "loss": 0.7965,
+      "step": 15712
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3222064947271189e-05,
+      "loss": 1.0166,
+      "step": 15713
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3221257647372878e-05,
+      "loss": 0.8646,
+      "step": 15714
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3220450324049222e-05,
+      "loss": 0.7896,
+      "step": 15715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3219642977306096e-05,
+      "loss": 0.8677,
+      "step": 15716
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321883560714937e-05,
+      "loss": 0.9248,
+      "step": 15717
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3218028213584909e-05,
+      "loss": 0.8928,
+      "step": 15718
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321722079661859e-05,
+      "loss": 0.8789,
+      "step": 15719
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3216413356256286e-05,
+      "loss": 1.0908,
+      "step": 15720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3215605892503866e-05,
+      "loss": 1.17,
+      "step": 15721
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3214798405367204e-05,
+      "loss": 1.0535,
+      "step": 15722
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321399089485217e-05,
+      "loss": 0.7804,
+      "step": 15723
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3213183360964642e-05,
+      "loss": 1.0522,
+      "step": 15724
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321237580371048e-05,
+      "loss": 0.9896,
+      "step": 15725
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3211568223095568e-05,
+      "loss": 0.8001,
+      "step": 15726
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3210760619125776e-05,
+      "loss": 0.7982,
+      "step": 15727
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209952991806976e-05,
+      "loss": 1.109,
+      "step": 15728
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.320914534114504e-05,
+      "loss": 0.5536,
+      "step": 15729
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3208337667145844e-05,
+      "loss": 0.9618,
+      "step": 15730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3207529969815258e-05,
+      "loss": 1.0389,
+      "step": 15731
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3206722249159163e-05,
+      "loss": 0.728,
+      "step": 15732
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3205914505183426e-05,
+      "loss": 0.857,
+      "step": 15733
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3205106737893917e-05,
+      "loss": 1.1107,
+      "step": 15734
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3204298947296521e-05,
+      "loss": 0.7489,
+      "step": 15735
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3203491133397107e-05,
+      "loss": 0.9796,
+      "step": 15736
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3202683296201548e-05,
+      "loss": 1.0104,
+      "step": 15737
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3201875435715723e-05,
+      "loss": 1.2277,
+      "step": 15738
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3201067551945504e-05,
+      "loss": 0.8903,
+      "step": 15739
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3200259644896762e-05,
+      "loss": 1.0924,
+      "step": 15740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.319945171457538e-05,
+      "loss": 0.8938,
+      "step": 15741
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3198643760987231e-05,
+      "loss": 1.2482,
+      "step": 15742
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197835784138186e-05,
+      "loss": 0.7382,
+      "step": 15743
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197027784034126e-05,
+      "loss": 0.8585,
+      "step": 15744
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3196219760680924e-05,
+      "loss": 1.0666,
+      "step": 15745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3195411714084456e-05,
+      "loss": 0.8299,
+      "step": 15746
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3194603644250604e-05,
+      "loss": 1.0497,
+      "step": 15747
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3193795551185237e-05,
+      "loss": 1.1848,
+      "step": 15748
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3192987434894232e-05,
+      "loss": 0.82,
+      "step": 15749
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.319217929538347e-05,
+      "loss": 0.8064,
+      "step": 15750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3191371132658823e-05,
+      "loss": 0.9037,
+      "step": 15751
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3190562946726174e-05,
+      "loss": 1.0012,
+      "step": 15752
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3189754737591396e-05,
+      "loss": 1.0531,
+      "step": 15753
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3188946505260367e-05,
+      "loss": 0.9263,
+      "step": 15754
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3188138249738963e-05,
+      "loss": 0.7896,
+      "step": 15755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3187329971033064e-05,
+      "loss": 0.9333,
+      "step": 15756
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3186521669148553e-05,
+      "loss": 0.9258,
+      "step": 15757
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3185713344091297e-05,
+      "loss": 1.2319,
+      "step": 15758
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318490499586718e-05,
+      "loss": 0.6705,
+      "step": 15759
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3184096624482083e-05,
+      "loss": 1.0984,
+      "step": 15760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3183288229941878e-05,
+      "loss": 0.8198,
+      "step": 15761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3182479812252452e-05,
+      "loss": 1.0509,
+      "step": 15762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3181671371419677e-05,
+      "loss": 1.0974,
+      "step": 15763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3180862907449435e-05,
+      "loss": 0.8039,
+      "step": 15764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3180054420347603e-05,
+      "loss": 1.0612,
+      "step": 15765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3179245910120065e-05,
+      "loss": 0.6679,
+      "step": 15766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3178437376772695e-05,
+      "loss": 0.9356,
+      "step": 15767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3177628820311376e-05,
+      "loss": 0.8662,
+      "step": 15768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317682024074199e-05,
+      "loss": 0.7933,
+      "step": 15769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317601163807041e-05,
+      "loss": 1.0505,
+      "step": 15770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3175203012302524e-05,
+      "loss": 0.9296,
+      "step": 15771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317439436344421e-05,
+      "loss": 0.9215,
+      "step": 15772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3173585691501343e-05,
+      "loss": 1.0242,
+      "step": 15773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3172776996479813e-05,
+      "loss": 0.6339,
+      "step": 15774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3171968278385494e-05,
+      "loss": 0.9989,
+      "step": 15775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317115953722427e-05,
+      "loss": 1.1867,
+      "step": 15776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3170350773002024e-05,
+      "loss": 0.9186,
+      "step": 15777
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3169541985724632e-05,
+      "loss": 1.0365,
+      "step": 15778
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3168733175397979e-05,
+      "loss": 1.0512,
+      "step": 15779
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3167924342027947e-05,
+      "loss": 0.9239,
+      "step": 15780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3167115485620417e-05,
+      "loss": 0.9714,
+      "step": 15781
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3166306606181271e-05,
+      "loss": 0.7651,
+      "step": 15782
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3165497703716392e-05,
+      "loss": 1.1602,
+      "step": 15783
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3164688778231664e-05,
+      "loss": 0.9947,
+      "step": 15784
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3163879829732969e-05,
+      "loss": 0.8706,
+      "step": 15785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3163070858226183e-05,
+      "loss": 1.0815,
+      "step": 15786
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.31622618637172e-05,
+      "loss": 0.8611,
+      "step": 15787
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3161452846211892e-05,
+      "loss": 1.1133,
+      "step": 15788
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.316064380571615e-05,
+      "loss": 0.9889,
+      "step": 15789
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315983474223586e-05,
+      "loss": 0.5368,
+      "step": 15790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3159025655776896e-05,
+      "loss": 1.0295,
+      "step": 15791
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3158216546345149e-05,
+      "loss": 0.8833,
+      "step": 15792
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3157407413946498e-05,
+      "loss": 0.7497,
+      "step": 15793
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3156598258586833e-05,
+      "loss": 0.9612,
+      "step": 15794
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315578908027203e-05,
+      "loss": 1.0465,
+      "step": 15795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154979879007983e-05,
+      "loss": 0.8158,
+      "step": 15796
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154170654800571e-05,
+      "loss": 0.8882,
+      "step": 15797
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3153361407655678e-05,
+      "loss": 1.0372,
+      "step": 15798
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3152552137579191e-05,
+      "loss": 1.2155,
+      "step": 15799
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3151742844576995e-05,
+      "loss": 0.662,
+      "step": 15800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3150933528654975e-05,
+      "loss": 0.998,
+      "step": 15801
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315012418981902e-05,
+      "loss": 1.0445,
+      "step": 15802
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3149314828075008e-05,
+      "loss": 1.1622,
+      "step": 15803
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3148505443428829e-05,
+      "loss": 0.7177,
+      "step": 15804
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314769603588637e-05,
+      "loss": 0.621,
+      "step": 15805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3146886605453517e-05,
+      "loss": 0.8681,
+      "step": 15806
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3146077152136152e-05,
+      "loss": 1.0848,
+      "step": 15807
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3145267675940168e-05,
+      "loss": 0.6604,
+      "step": 15808
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3144458176871446e-05,
+      "loss": 0.8625,
+      "step": 15809
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3143648654935875e-05,
+      "loss": 0.8049,
+      "step": 15810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3142839110139342e-05,
+      "loss": 1.3218,
+      "step": 15811
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3142029542487738e-05,
+      "loss": 0.8868,
+      "step": 15812
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3141219951986939e-05,
+      "loss": 1.115,
+      "step": 15813
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3140410338642843e-05,
+      "loss": 0.8383,
+      "step": 15814
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3139600702461335e-05,
+      "loss": 1.0128,
+      "step": 15815
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.31387910434483e-05,
+      "loss": 1.246,
+      "step": 15816
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3137981361609632e-05,
+      "loss": 1.2048,
+      "step": 15817
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3137171656951213e-05,
+      "loss": 0.9659,
+      "step": 15818
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3136361929478931e-05,
+      "loss": 0.746,
+      "step": 15819
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3135552179198679e-05,
+      "loss": 0.6705,
+      "step": 15820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3134742406116343e-05,
+      "loss": 0.8936,
+      "step": 15821
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3133932610237815e-05,
+      "loss": 0.6426,
+      "step": 15822
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3133122791568976e-05,
+      "loss": 1.0607,
+      "step": 15823
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3132312950115723e-05,
+      "loss": 0.9869,
+      "step": 15824
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313150308588394e-05,
+      "loss": 0.8307,
+      "step": 15825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3130693198879521e-05,
+      "loss": 0.9635,
+      "step": 15826
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3129883289108357e-05,
+      "loss": 1.2038,
+      "step": 15827
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3129073356576328e-05,
+      "loss": 0.6184,
+      "step": 15828
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3128263401289331e-05,
+      "loss": 1.0973,
+      "step": 15829
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3127453423253255e-05,
+      "loss": 0.9492,
+      "step": 15830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3126643422473993e-05,
+      "loss": 1.1045,
+      "step": 15831
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3125833398957434e-05,
+      "loss": 0.6081,
+      "step": 15832
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3125023352709464e-05,
+      "loss": 1.1675,
+      "step": 15833
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3124213283735976e-05,
+      "loss": 1.1691,
+      "step": 15834
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3123403192042864e-05,
+      "loss": 0.7511,
+      "step": 15835
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3122593077636019e-05,
+      "loss": 1.0367,
+      "step": 15836
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3121782940521328e-05,
+      "loss": 0.6928,
+      "step": 15837
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3120972780704686e-05,
+      "loss": 0.8272,
+      "step": 15838
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3120162598191984e-05,
+      "loss": 0.8172,
+      "step": 15839
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311935239298911e-05,
+      "loss": 0.7953,
+      "step": 15840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3118542165101961e-05,
+      "loss": 0.7816,
+      "step": 15841
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311773191453643e-05,
+      "loss": 0.951,
+      "step": 15842
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116921641298401e-05,
+      "loss": 1.0698,
+      "step": 15843
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116111345393776e-05,
+      "loss": 0.4896,
+      "step": 15844
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311530102682844e-05,
+      "loss": 0.7587,
+      "step": 15845
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3114490685608292e-05,
+      "loss": 1.2352,
+      "step": 15846
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3113680321739219e-05,
+      "loss": 0.8451,
+      "step": 15847
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311286993522712e-05,
+      "loss": 0.823,
+      "step": 15848
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.311205952607788e-05,
+      "loss": 0.7738,
+      "step": 15849
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3111249094297403e-05,
+      "loss": 1.0511,
+      "step": 15850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3110438639891573e-05,
+      "loss": 0.9187,
+      "step": 15851
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.310962816286629e-05,
+      "loss": 0.5215,
+      "step": 15852
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3108817663227447e-05,
+      "loss": 1.0785,
+      "step": 15853
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3108007140980936e-05,
+      "loss": 0.8984,
+      "step": 15854
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.310719659613265e-05,
+      "loss": 1.1776,
+      "step": 15855
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3106386028688485e-05,
+      "loss": 1.1497,
+      "step": 15856
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3105575438654338e-05,
+      "loss": 0.9238,
+      "step": 15857
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.31047648260361e-05,
+      "loss": 0.8973,
+      "step": 15858
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.310395419083967e-05,
+      "loss": 1.1181,
+      "step": 15859
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3103143533070938e-05,
+      "loss": 0.9068,
+      "step": 15860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3102332852735802e-05,
+      "loss": 1.0022,
+      "step": 15861
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3101522149840157e-05,
+      "loss": 0.7332,
+      "step": 15862
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.31007114243899e-05,
+      "loss": 1.0585,
+      "step": 15863
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3099900676390924e-05,
+      "loss": 1.1526,
+      "step": 15864
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3099089905849126e-05,
+      "loss": 0.7811,
+      "step": 15865
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30982791127704e-05,
+      "loss": 1.0455,
+      "step": 15866
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3097468297160647e-05,
+      "loss": 0.7045,
+      "step": 15867
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3096657459025762e-05,
+      "loss": 1.0093,
+      "step": 15868
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3095846598371638e-05,
+      "loss": 0.9782,
+      "step": 15869
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3095035715204171e-05,
+      "loss": 0.9088,
+      "step": 15870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3094224809529264e-05,
+      "loss": 0.8883,
+      "step": 15871
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3093413881352813e-05,
+      "loss": 0.603,
+      "step": 15872
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.309260293068071e-05,
+      "loss": 0.927,
+      "step": 15873
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3091791957518852e-05,
+      "loss": 1.0162,
+      "step": 15874
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3090980961873143e-05,
+      "loss": 0.7012,
+      "step": 15875
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3090169943749475e-05,
+      "loss": 0.983,
+      "step": 15876
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3089358903153751e-05,
+      "loss": 0.7683,
+      "step": 15877
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088547840091863e-05,
+      "loss": 0.8033,
+      "step": 15878
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3087736754569715e-05,
+      "loss": 0.8271,
+      "step": 15879
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.30869256465932e-05,
+      "loss": 1.113,
+      "step": 15880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308611451616822e-05,
+      "loss": 0.8789,
+      "step": 15881
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3085303363300674e-05,
+      "loss": 1.2776,
+      "step": 15882
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3084492187996457e-05,
+      "loss": 0.9826,
+      "step": 15883
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3083680990261472e-05,
+      "loss": 1.073,
+      "step": 15884
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3082869770101613e-05,
+      "loss": 0.8555,
+      "step": 15885
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3082058527522785e-05,
+      "loss": 1.0138,
+      "step": 15886
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3081247262530887e-05,
+      "loss": 0.8824,
+      "step": 15887
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3080435975131815e-05,
+      "loss": 1.0692,
+      "step": 15888
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.307962466533147e-05,
+      "loss": 1.0174,
+      "step": 15889
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.307881333313575e-05,
+      "loss": 1.0252,
+      "step": 15890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3078001978550562e-05,
+      "loss": 1.0097,
+      "step": 15891
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3077190601581802e-05,
+      "loss": 1.0559,
+      "step": 15892
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3076379202235367e-05,
+      "loss": 0.9182,
+      "step": 15893
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3075567780517165e-05,
+      "loss": 0.8626,
+      "step": 15894
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3074756336433088e-05,
+      "loss": 0.7866,
+      "step": 15895
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073944869989043e-05,
+      "loss": 1.1229,
+      "step": 15896
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073133381190932e-05,
+      "loss": 0.9512,
+      "step": 15897
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3072321870044651e-05,
+      "loss": 0.6925,
+      "step": 15898
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3071510336556106e-05,
+      "loss": 0.8225,
+      "step": 15899
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3070698780731194e-05,
+      "loss": 0.9274,
+      "step": 15900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306988720257582e-05,
+      "loss": 1.3469,
+      "step": 15901
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306907560209589e-05,
+      "loss": 0.6837,
+      "step": 15902
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3068263979297298e-05,
+      "loss": 0.8741,
+      "step": 15903
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3067452334185947e-05,
+      "loss": 1.0126,
+      "step": 15904
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3066640666767745e-05,
+      "loss": 1.0038,
+      "step": 15905
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306582897704859e-05,
+      "loss": 1.4563,
+      "step": 15906
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3065017265034388e-05,
+      "loss": 0.7029,
+      "step": 15907
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306420553073104e-05,
+      "loss": 0.8683,
+      "step": 15908
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3063393774144451e-05,
+      "loss": 0.9506,
+      "step": 15909
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3062581995280517e-05,
+      "loss": 1.0919,
+      "step": 15910
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306177019414515e-05,
+      "loss": 0.6976,
+      "step": 15911
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306095837074425e-05,
+      "loss": 0.9879,
+      "step": 15912
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306014652508372e-05,
+      "loss": 1.0071,
+      "step": 15913
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3059334657169465e-05,
+      "loss": 0.9909,
+      "step": 15914
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305852276700739e-05,
+      "loss": 1.2013,
+      "step": 15915
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3057710854603397e-05,
+      "loss": 0.8833,
+      "step": 15916
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305689891996339e-05,
+      "loss": 0.8991,
+      "step": 15917
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3056086963093275e-05,
+      "loss": 0.9571,
+      "step": 15918
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3055274983998953e-05,
+      "loss": 0.8389,
+      "step": 15919
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3054462982686337e-05,
+      "loss": 0.7982,
+      "step": 15920
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3053650959161326e-05,
+      "loss": 0.5612,
+      "step": 15921
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3052838913429826e-05,
+      "loss": 1.388,
+      "step": 15922
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3052026845497743e-05,
+      "loss": 0.9691,
+      "step": 15923
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305121475537098e-05,
+      "loss": 1.0499,
+      "step": 15924
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3050402643055447e-05,
+      "loss": 0.7262,
+      "step": 15925
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3049590508557044e-05,
+      "loss": 1.2367,
+      "step": 15926
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3048778351881684e-05,
+      "loss": 0.6985,
+      "step": 15927
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3047966173035265e-05,
+      "loss": 0.872,
+      "step": 15928
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3047153972023698e-05,
+      "loss": 0.7791,
+      "step": 15929
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.304634174885289e-05,
+      "loss": 0.9808,
+      "step": 15930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045529503528749e-05,
+      "loss": 1.0242,
+      "step": 15931
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3044717236057178e-05,
+      "loss": 0.7845,
+      "step": 15932
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043904946444083e-05,
+      "loss": 1.162,
+      "step": 15933
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043092634695376e-05,
+      "loss": 1.002,
+      "step": 15934
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3042280300816959e-05,
+      "loss": 1.2934,
+      "step": 15935
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3041467944814743e-05,
+      "loss": 1.0056,
+      "step": 15936
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3040655566694636e-05,
+      "loss": 0.6177,
+      "step": 15937
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303984316646254e-05,
+      "loss": 0.9897,
+      "step": 15938
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303903074412437e-05,
+      "loss": 0.6587,
+      "step": 15939
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303821829968603e-05,
+      "loss": 1.026,
+      "step": 15940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3037405833153427e-05,
+      "loss": 1.0527,
+      "step": 15941
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3036593344532478e-05,
+      "loss": 0.8143,
+      "step": 15942
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3035780833829081e-05,
+      "loss": 0.8212,
+      "step": 15943
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3034968301049147e-05,
+      "loss": 0.9172,
+      "step": 15944
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3034155746198588e-05,
+      "loss": 0.9941,
+      "step": 15945
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3033343169283311e-05,
+      "loss": 0.5545,
+      "step": 15946
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3032530570309225e-05,
+      "loss": 0.8181,
+      "step": 15947
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303171794928224e-05,
+      "loss": 1.0497,
+      "step": 15948
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303090530620827e-05,
+      "loss": 1.2031,
+      "step": 15949
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3030092641093214e-05,
+      "loss": 0.9065,
+      "step": 15950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3029279953942991e-05,
+      "loss": 1.3958,
+      "step": 15951
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3028467244763508e-05,
+      "loss": 0.8583,
+      "step": 15952
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3027654513560672e-05,
+      "loss": 0.7964,
+      "step": 15953
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3026841760340398e-05,
+      "loss": 0.7627,
+      "step": 15954
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3026028985108596e-05,
+      "loss": 0.8979,
+      "step": 15955
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3025216187871173e-05,
+      "loss": 1.1135,
+      "step": 15956
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3024403368634045e-05,
+      "loss": 0.9203,
+      "step": 15957
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3023590527403118e-05,
+      "loss": 0.9072,
+      "step": 15958
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3022777664184303e-05,
+      "loss": 1.0983,
+      "step": 15959
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3021964778983513e-05,
+      "loss": 0.8508,
+      "step": 15960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3021151871806663e-05,
+      "loss": 1.1488,
+      "step": 15961
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302033894265966e-05,
+      "loss": 0.9795,
+      "step": 15962
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3019525991548416e-05,
+      "loss": 1.0284,
+      "step": 15963
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3018713018478844e-05,
+      "loss": 0.7035,
+      "step": 15964
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3017900023456852e-05,
+      "loss": 0.7931,
+      "step": 15965
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3017087006488361e-05,
+      "loss": 1.3973,
+      "step": 15966
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3016273967579279e-05,
+      "loss": 0.8317,
+      "step": 15967
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3015460906735516e-05,
+      "loss": 0.8258,
+      "step": 15968
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3014647823962984e-05,
+      "loss": 1.0438,
+      "step": 15969
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3013834719267599e-05,
+      "loss": 0.9305,
+      "step": 15970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3013021592655275e-05,
+      "loss": 1.3167,
+      "step": 15971
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3012208444131925e-05,
+      "loss": 0.9637,
+      "step": 15972
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3011395273703457e-05,
+      "loss": 0.5978,
+      "step": 15973
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301058208137579e-05,
+      "loss": 0.7083,
+      "step": 15974
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3009768867154834e-05,
+      "loss": 0.8701,
+      "step": 15975
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3008955631046507e-05,
+      "loss": 0.6253,
+      "step": 15976
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300814237305672e-05,
+      "loss": 0.9042,
+      "step": 15977
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3007329093191387e-05,
+      "loss": 1.0399,
+      "step": 15978
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3006515791456425e-05,
+      "loss": 0.9958,
+      "step": 15979
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3005702467857744e-05,
+      "loss": 1.3744,
+      "step": 15980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3004889122401258e-05,
+      "loss": 1.1703,
+      "step": 15981
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300407575509289e-05,
+      "loss": 0.8602,
+      "step": 15982
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3003262365938547e-05,
+      "loss": 0.9702,
+      "step": 15983
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3002448954944148e-05,
+      "loss": 1.1034,
+      "step": 15984
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3001635522115605e-05,
+      "loss": 0.8027,
+      "step": 15985
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3000822067458837e-05,
+      "loss": 1.102,
+      "step": 15986
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3000008590979755e-05,
+      "loss": 0.9289,
+      "step": 15987
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299919509268428e-05,
+      "loss": 0.9738,
+      "step": 15988
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2998381572578325e-05,
+      "loss": 0.9398,
+      "step": 15989
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2997568030667802e-05,
+      "loss": 0.8581,
+      "step": 15990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2996754466958637e-05,
+      "loss": 0.8278,
+      "step": 15991
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2995940881456737e-05,
+      "loss": 0.6614,
+      "step": 15992
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2995127274168025e-05,
+      "loss": 0.8644,
+      "step": 15993
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2994313645098413e-05,
+      "loss": 1.0166,
+      "step": 15994
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299349999425382e-05,
+      "loss": 1.0802,
+      "step": 15995
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2992686321640162e-05,
+      "loss": 0.8305,
+      "step": 15996
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299187262726336e-05,
+      "loss": 0.8737,
+      "step": 15997
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2991058911129326e-05,
+      "loss": 0.8708,
+      "step": 15998
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2990245173243979e-05,
+      "loss": 1.0832,
+      "step": 15999
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2989431413613237e-05,
+      "loss": 1.0023,
+      "step": 16000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2988617632243016e-05,
+      "loss": 0.9002,
+      "step": 16001
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987803829139238e-05,
+      "loss": 0.8798,
+      "step": 16002
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.298699000430782e-05,
+      "loss": 0.9717,
+      "step": 16003
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2986176157754678e-05,
+      "loss": 0.902,
+      "step": 16004
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2985362289485728e-05,
+      "loss": 1.1733,
+      "step": 16005
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2984548399506895e-05,
+      "loss": 0.5231,
+      "step": 16006
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2983734487824098e-05,
+      "loss": 0.7091,
+      "step": 16007
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2982920554443246e-05,
+      "loss": 1.0873,
+      "step": 16008
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2982106599370268e-05,
+      "loss": 0.6964,
+      "step": 16009
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2981292622611079e-05,
+      "loss": 1.1764,
+      "step": 16010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2980478624171595e-05,
+      "loss": 1.3732,
+      "step": 16011
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2979664604057747e-05,
+      "loss": 0.7583,
+      "step": 16012
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2978850562275444e-05,
+      "loss": 0.7598,
+      "step": 16013
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2978036498830607e-05,
+      "loss": 0.9084,
+      "step": 16014
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2977222413729157e-05,
+      "loss": 0.8525,
+      "step": 16015
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2976408306977019e-05,
+      "loss": 1.0075,
+      "step": 16016
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2975594178580109e-05,
+      "loss": 1.29,
+      "step": 16017
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2974780028544345e-05,
+      "loss": 0.8637,
+      "step": 16018
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2973965856875652e-05,
+      "loss": 0.9288,
+      "step": 16019
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2973151663579948e-05,
+      "loss": 1.1652,
+      "step": 16020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2972337448663155e-05,
+      "loss": 0.8535,
+      "step": 16021
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2971523212131198e-05,
+      "loss": 0.8687,
+      "step": 16022
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2970708953989987e-05,
+      "loss": 1.2172,
+      "step": 16023
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2969894674245456e-05,
+      "loss": 1.017,
+      "step": 16024
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296908037290352e-05,
+      "loss": 1.2772,
+      "step": 16025
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29682660499701e-05,
+      "loss": 0.8915,
+      "step": 16026
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2967451705451123e-05,
+      "loss": 0.9292,
+      "step": 16027
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2966637339352505e-05,
+      "loss": 1.1752,
+      "step": 16028
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2965822951680169e-05,
+      "loss": 0.996,
+      "step": 16029
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2965008542440042e-05,
+      "loss": 0.747,
+      "step": 16030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2964194111638043e-05,
+      "loss": 0.9334,
+      "step": 16031
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2963379659280098e-05,
+      "loss": 1.2794,
+      "step": 16032
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2962565185372125e-05,
+      "loss": 1.3894,
+      "step": 16033
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2961750689920047e-05,
+      "loss": 1.1628,
+      "step": 16034
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296093617292979e-05,
+      "loss": 1.3112,
+      "step": 16035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2960121634407276e-05,
+      "loss": 0.7793,
+      "step": 16036
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2959307074358431e-05,
+      "loss": 1.2813,
+      "step": 16037
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2958492492789174e-05,
+      "loss": 1.2046,
+      "step": 16038
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2957677889705434e-05,
+      "loss": 0.6362,
+      "step": 16039
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956863265113128e-05,
+      "loss": 0.9174,
+      "step": 16040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956048619018189e-05,
+      "loss": 0.8912,
+      "step": 16041
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2955233951426532e-05,
+      "loss": 1.1337,
+      "step": 16042
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2954419262344087e-05,
+      "loss": 0.6793,
+      "step": 16043
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2953604551776776e-05,
+      "loss": 0.9111,
+      "step": 16044
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2952789819730525e-05,
+      "loss": 1.132,
+      "step": 16045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.295197506621126e-05,
+      "loss": 0.9782,
+      "step": 16046
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2951160291224902e-05,
+      "loss": 0.907,
+      "step": 16047
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2950345494777381e-05,
+      "loss": 1.1836,
+      "step": 16048
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294953067687462e-05,
+      "loss": 1.0501,
+      "step": 16049
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2948715837522542e-05,
+      "loss": 1.2711,
+      "step": 16050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2947900976727076e-05,
+      "loss": 0.7265,
+      "step": 16051
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294708609449415e-05,
+      "loss": 0.9518,
+      "step": 16052
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2946271190829683e-05,
+      "loss": 0.9571,
+      "step": 16053
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2945456265739604e-05,
+      "loss": 0.8181,
+      "step": 16054
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2944641319229842e-05,
+      "loss": 1.1458,
+      "step": 16055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2943826351306321e-05,
+      "loss": 0.8861,
+      "step": 16056
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294301136197497e-05,
+      "loss": 0.672,
+      "step": 16057
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2942196351241712e-05,
+      "loss": 1.0862,
+      "step": 16058
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2941381319112474e-05,
+      "loss": 0.8952,
+      "step": 16059
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2940566265593182e-05,
+      "loss": 0.7157,
+      "step": 16060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2939751190689768e-05,
+      "loss": 1.0436,
+      "step": 16061
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2938936094408159e-05,
+      "loss": 1.0149,
+      "step": 16062
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293812097675428e-05,
+      "loss": 0.4721,
+      "step": 16063
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2937305837734058e-05,
+      "loss": 1.1303,
+      "step": 16064
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2936490677353422e-05,
+      "loss": 0.9401,
+      "step": 16065
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2935675495618299e-05,
+      "loss": 0.8514,
+      "step": 16066
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2934860292534618e-05,
+      "loss": 0.7544,
+      "step": 16067
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293404506810831e-05,
+      "loss": 1.017,
+      "step": 16068
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2933229822345294e-05,
+      "loss": 0.9086,
+      "step": 16069
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2932414555251508e-05,
+      "loss": 0.8471,
+      "step": 16070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2931599266832879e-05,
+      "loss": 0.9954,
+      "step": 16071
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2930783957095336e-05,
+      "loss": 0.7564,
+      "step": 16072
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2929968626044805e-05,
+      "loss": 0.8325,
+      "step": 16073
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2929153273687216e-05,
+      "loss": 0.9257,
+      "step": 16074
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2928337900028498e-05,
+      "loss": 1.1936,
+      "step": 16075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2927522505074584e-05,
+      "loss": 1.1251,
+      "step": 16076
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29267070888314e-05,
+      "loss": 1.0969,
+      "step": 16077
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2925891651304876e-05,
+      "loss": 1.209,
+      "step": 16078
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2925076192500945e-05,
+      "loss": 0.7142,
+      "step": 16079
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2924260712425536e-05,
+      "loss": 0.8993,
+      "step": 16080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2923445211084575e-05,
+      "loss": 0.9199,
+      "step": 16081
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2922629688484e-05,
+      "loss": 1.2018,
+      "step": 16082
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2921814144629736e-05,
+      "loss": 1.0048,
+      "step": 16083
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292099857952771e-05,
+      "loss": 0.8809,
+      "step": 16084
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2920182993183864e-05,
+      "loss": 0.7428,
+      "step": 16085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2919367385604123e-05,
+      "loss": 0.8254,
+      "step": 16086
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2918551756794415e-05,
+      "loss": 1.0356,
+      "step": 16087
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2917736106760677e-05,
+      "loss": 1.1596,
+      "step": 16088
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2916920435508835e-05,
+      "loss": 0.9433,
+      "step": 16089
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2916104743044824e-05,
+      "loss": 1.0304,
+      "step": 16090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2915289029374579e-05,
+      "loss": 1.0872,
+      "step": 16091
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291447329450403e-05,
+      "loss": 0.9404,
+      "step": 16092
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2913657538439099e-05,
+      "loss": 0.9965,
+      "step": 16093
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2912841761185733e-05,
+      "loss": 0.9463,
+      "step": 16094
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2912025962749856e-05,
+      "loss": 1.0953,
+      "step": 16095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2911210143137403e-05,
+      "loss": 1.0563,
+      "step": 16096
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291039430235431e-05,
+      "loss": 0.9749,
+      "step": 16097
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2909578440406504e-05,
+      "loss": 0.6219,
+      "step": 16098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908762557299919e-05,
+      "loss": 0.9331,
+      "step": 16099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2907946653040491e-05,
+      "loss": 0.8951,
+      "step": 16100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2907130727634152e-05,
+      "loss": 0.7754,
+      "step": 16101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2906314781086837e-05,
+      "loss": 0.787,
+      "step": 16102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2905498813404475e-05,
+      "loss": 0.7789,
+      "step": 16103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2904682824593006e-05,
+      "loss": 1.0572,
+      "step": 16104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2903866814658358e-05,
+      "loss": 0.8815,
+      "step": 16105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290305078360647e-05,
+      "loss": 0.9165,
+      "step": 16106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2902234731443275e-05,
+      "loss": 1.3891,
+      "step": 16107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2901418658174704e-05,
+      "loss": 0.7009,
+      "step": 16108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2900602563806697e-05,
+      "loss": 0.6484,
+      "step": 16109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2899786448345186e-05,
+      "loss": 0.8505,
+      "step": 16110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2898970311796103e-05,
+      "loss": 1.0441,
+      "step": 16111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2898154154165389e-05,
+      "loss": 0.6573,
+      "step": 16112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2897337975458977e-05,
+      "loss": 0.9412,
+      "step": 16113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2896521775682798e-05,
+      "loss": 0.9552,
+      "step": 16114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2895705554842792e-05,
+      "loss": 0.627,
+      "step": 16115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2894889312944897e-05,
+      "loss": 0.851,
+      "step": 16116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2894073049995043e-05,
+      "loss": 0.8344,
+      "step": 16117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2893256765999167e-05,
+      "loss": 0.8166,
+      "step": 16118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2892440460963207e-05,
+      "loss": 0.9321,
+      "step": 16119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2891624134893101e-05,
+      "loss": 0.5991,
+      "step": 16120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.289080778779478e-05,
+      "loss": 0.9466,
+      "step": 16121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2889991419674189e-05,
+      "loss": 0.975,
+      "step": 16122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2889175030537254e-05,
+      "loss": 0.935,
+      "step": 16123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2888358620389918e-05,
+      "loss": 0.7368,
+      "step": 16124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288754218923812e-05,
+      "loss": 0.8021,
+      "step": 16125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2886725737087793e-05,
+      "loss": 1.0478,
+      "step": 16126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2885909263944876e-05,
+      "loss": 0.6355,
+      "step": 16127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2885092769815306e-05,
+      "loss": 0.785,
+      "step": 16128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2884276254705022e-05,
+      "loss": 0.804,
+      "step": 16129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288345971861996e-05,
+      "loss": 1.2567,
+      "step": 16130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2882643161566058e-05,
+      "loss": 1.0103,
+      "step": 16131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2881826583549255e-05,
+      "loss": 1.143,
+      "step": 16132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288100998457549e-05,
+      "loss": 0.8383,
+      "step": 16133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2880193364650701e-05,
+      "loss": 0.7642,
+      "step": 16134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2879376723780826e-05,
+      "loss": 0.9885,
+      "step": 16135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2878560061971801e-05,
+      "loss": 1.0597,
+      "step": 16136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2877743379229572e-05,
+      "loss": 1.0248,
+      "step": 16137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2876926675560072e-05,
+      "loss": 0.7935,
+      "step": 16138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2876109950969238e-05,
+      "loss": 0.6482,
+      "step": 16139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2875293205463018e-05,
+      "loss": 0.6263,
+      "step": 16140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2874476439047342e-05,
+      "loss": 0.6199,
+      "step": 16141
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2873659651728159e-05,
+      "loss": 0.9922,
+      "step": 16142
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2872842843511402e-05,
+      "loss": 0.8181,
+      "step": 16143
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2872026014403012e-05,
+      "loss": 0.8312,
+      "step": 16144
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2871209164408928e-05,
+      "loss": 1.1366,
+      "step": 16145
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2870392293535095e-05,
+      "loss": 0.9137,
+      "step": 16146
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286957540178745e-05,
+      "loss": 0.8868,
+      "step": 16147
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2868758489171933e-05,
+      "loss": 0.7167,
+      "step": 16148
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2867941555694488e-05,
+      "loss": 1.2384,
+      "step": 16149
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2867124601361051e-05,
+      "loss": 1.1974,
+      "step": 16150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2866307626177565e-05,
+      "loss": 0.8398,
+      "step": 16151
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2865490630149975e-05,
+      "loss": 1.0977,
+      "step": 16152
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2864673613284217e-05,
+      "loss": 1.0581,
+      "step": 16153
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2863856575586234e-05,
+      "loss": 0.911,
+      "step": 16154
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2863039517061968e-05,
+      "loss": 0.8488,
+      "step": 16155
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2862222437717363e-05,
+      "loss": 1.0485,
+      "step": 16156
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2861405337558358e-05,
+      "loss": 1.0535,
+      "step": 16157
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2860588216590892e-05,
+      "loss": 0.9683,
+      "step": 16158
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2859771074820913e-05,
+      "loss": 0.9091,
+      "step": 16159
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2858953912254362e-05,
+      "loss": 1.1669,
+      "step": 16160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2858136728897179e-05,
+      "loss": 1.1084,
+      "step": 16161
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2857319524755312e-05,
+      "loss": 0.9073,
+      "step": 16162
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2856502299834695e-05,
+      "loss": 1.0743,
+      "step": 16163
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285568505414128e-05,
+      "loss": 1.1207,
+      "step": 16164
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2854867787681004e-05,
+      "loss": 0.902,
+      "step": 16165
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2854050500459811e-05,
+      "loss": 0.7773,
+      "step": 16166
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2853233192483653e-05,
+      "loss": 1.3562,
+      "step": 16167
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2852415863758458e-05,
+      "loss": 1.0627,
+      "step": 16168
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2851598514290183e-05,
+      "loss": 0.9593,
+      "step": 16169
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2850781144084763e-05,
+      "loss": 0.862,
+      "step": 16170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2849963753148149e-05,
+      "loss": 0.9099,
+      "step": 16171
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284914634148628e-05,
+      "loss": 1.1121,
+      "step": 16172
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2848328909105106e-05,
+      "loss": 1.0901,
+      "step": 16173
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2847511456010567e-05,
+      "loss": 0.7708,
+      "step": 16174
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2846693982208605e-05,
+      "loss": 0.959,
+      "step": 16175
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2845876487705172e-05,
+      "loss": 0.7824,
+      "step": 16176
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2845058972506208e-05,
+      "loss": 0.9468,
+      "step": 16177
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2844241436617658e-05,
+      "loss": 0.8107,
+      "step": 16178
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284342388004547e-05,
+      "loss": 1.1462,
+      "step": 16179
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2842606302795586e-05,
+      "loss": 1.043,
+      "step": 16180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2841788704873954e-05,
+      "loss": 0.813,
+      "step": 16181
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284097108628652e-05,
+      "loss": 0.8453,
+      "step": 16182
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284015344703923e-05,
+      "loss": 0.8355,
+      "step": 16183
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2839335787138023e-05,
+      "loss": 1.0291,
+      "step": 16184
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2838518106588856e-05,
+      "loss": 0.4965,
+      "step": 16185
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2837700405397669e-05,
+      "loss": 1.2923,
+      "step": 16186
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836882683570408e-05,
+      "loss": 0.9925,
+      "step": 16187
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836064941113024e-05,
+      "loss": 0.8787,
+      "step": 16188
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2835247178031457e-05,
+      "loss": 0.8867,
+      "step": 16189
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2834429394331662e-05,
+      "loss": 0.7057,
+      "step": 16190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2833611590019577e-05,
+      "loss": 1.0017,
+      "step": 16191
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2832793765101159e-05,
+      "loss": 0.8048,
+      "step": 16192
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2831975919582348e-05,
+      "loss": 1.282,
+      "step": 16193
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2831158053469092e-05,
+      "loss": 0.7923,
+      "step": 16194
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283034016676734e-05,
+      "loss": 0.8997,
+      "step": 16195
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2829522259483044e-05,
+      "loss": 0.8515,
+      "step": 16196
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2828704331622146e-05,
+      "loss": 0.8415,
+      "step": 16197
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2827886383190595e-05,
+      "loss": 0.9253,
+      "step": 16198
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2827068414194341e-05,
+      "loss": 0.5798,
+      "step": 16199
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2826250424639329e-05,
+      "loss": 1.2844,
+      "step": 16200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2825432414531514e-05,
+      "loss": 1.0702,
+      "step": 16201
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2824614383876837e-05,
+      "loss": 0.9371,
+      "step": 16202
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2823796332681253e-05,
+      "loss": 1.0836,
+      "step": 16203
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2822978260950709e-05,
+      "loss": 0.8206,
+      "step": 16204
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2822160168691153e-05,
+      "loss": 1.1329,
+      "step": 16205
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2821342055908532e-05,
+      "loss": 1.0364,
+      "step": 16206
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2820523922608803e-05,
+      "loss": 0.9857,
+      "step": 16207
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.281970576879791e-05,
+      "loss": 1.2263,
+      "step": 16208
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.28188875944818e-05,
+      "loss": 0.7357,
+      "step": 16209
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2818069399666428e-05,
+      "loss": 1.1931,
+      "step": 16210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2817251184357744e-05,
+      "loss": 1.222,
+      "step": 16211
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2816432948561695e-05,
+      "loss": 0.9736,
+      "step": 16212
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2815614692284232e-05,
+      "loss": 0.9536,
+      "step": 16213
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.281479641553131e-05,
+      "loss": 0.9176,
+      "step": 16214
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2813978118308872e-05,
+      "loss": 0.683,
+      "step": 16215
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2813159800622873e-05,
+      "loss": 1.0899,
+      "step": 16216
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2812341462479267e-05,
+      "loss": 0.9066,
+      "step": 16217
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2811523103883995e-05,
+      "loss": 0.8031,
+      "step": 16218
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2810704724843022e-05,
+      "loss": 1.0181,
+      "step": 16219
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2809886325362288e-05,
+      "loss": 1.0447,
+      "step": 16220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280906790544775e-05,
+      "loss": 0.8966,
+      "step": 16221
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2808249465105357e-05,
+      "loss": 0.5647,
+      "step": 16222
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2807431004341064e-05,
+      "loss": 0.8471,
+      "step": 16223
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280661252316082e-05,
+      "loss": 0.986,
+      "step": 16224
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2805794021570577e-05,
+      "loss": 1.2029,
+      "step": 16225
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2804975499576289e-05,
+      "loss": 0.8499,
+      "step": 16226
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2804156957183908e-05,
+      "loss": 0.9663,
+      "step": 16227
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2803338394399388e-05,
+      "loss": 1.2093,
+      "step": 16228
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2802519811228677e-05,
+      "loss": 0.642,
+      "step": 16229
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2801701207677731e-05,
+      "loss": 0.834,
+      "step": 16230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2800882583752504e-05,
+      "loss": 0.8769,
+      "step": 16231
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2800063939458948e-05,
+      "loss": 1.0464,
+      "step": 16232
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2799245274803013e-05,
+      "loss": 1.1198,
+      "step": 16233
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279842658979066e-05,
+      "loss": 0.8246,
+      "step": 16234
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2797607884427836e-05,
+      "loss": 1.0507,
+      "step": 16235
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2796789158720497e-05,
+      "loss": 0.9506,
+      "step": 16236
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2795970412674596e-05,
+      "loss": 1.0112,
+      "step": 16237
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279515164629609e-05,
+      "loss": 0.8009,
+      "step": 16238
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2794332859590925e-05,
+      "loss": 0.8581,
+      "step": 16239
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2793514052565066e-05,
+      "loss": 0.8719,
+      "step": 16240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2792695225224463e-05,
+      "loss": 1.1022,
+      "step": 16241
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2791876377575068e-05,
+      "loss": 1.0791,
+      "step": 16242
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2791057509622839e-05,
+      "loss": 0.8383,
+      "step": 16243
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2790238621373731e-05,
+      "loss": 0.9499,
+      "step": 16244
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2789419712833698e-05,
+      "loss": 0.9692,
+      "step": 16245
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2788600784008688e-05,
+      "loss": 0.907,
+      "step": 16246
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2787781834904673e-05,
+      "loss": 1.0188,
+      "step": 16247
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2786962865527594e-05,
+      "loss": 0.5999,
+      "step": 16248
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2786143875883412e-05,
+      "loss": 0.7489,
+      "step": 16249
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2785324865978084e-05,
+      "loss": 0.7207,
+      "step": 16250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2784505835817562e-05,
+      "loss": 1.0081,
+      "step": 16251
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2783686785407806e-05,
+      "loss": 1.0453,
+      "step": 16252
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2782867714754772e-05,
+      "loss": 0.7203,
+      "step": 16253
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2782048623864413e-05,
+      "loss": 0.806,
+      "step": 16254
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2781229512742685e-05,
+      "loss": 1.3278,
+      "step": 16255
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278041038139555e-05,
+      "loss": 0.8478,
+      "step": 16256
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2779591229828962e-05,
+      "loss": 0.6652,
+      "step": 16257
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2778772058048878e-05,
+      "loss": 1.0945,
+      "step": 16258
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2777952866061253e-05,
+      "loss": 0.883,
+      "step": 16259
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277713365387205e-05,
+      "loss": 1.0312,
+      "step": 16260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277631442148722e-05,
+      "loss": 1.0424,
+      "step": 16261
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2775495168912726e-05,
+      "loss": 0.9609,
+      "step": 16262
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277467589615452e-05,
+      "loss": 0.702,
+      "step": 16263
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2773856603218562e-05,
+      "loss": 0.8268,
+      "step": 16264
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2773037290110813e-05,
+      "loss": 0.8514,
+      "step": 16265
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2772217956837227e-05,
+      "loss": 0.7411,
+      "step": 16266
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2771398603403767e-05,
+      "loss": 0.6372,
+      "step": 16267
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2770579229816385e-05,
+      "loss": 0.766,
+      "step": 16268
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2769759836081045e-05,
+      "loss": 1.1288,
+      "step": 16269
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2768940422203703e-05,
+      "loss": 1.0615,
+      "step": 16270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.276812098819032e-05,
+      "loss": 0.5528,
+      "step": 16271
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2767301534046857e-05,
+      "loss": 0.955,
+      "step": 16272
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2766482059779263e-05,
+      "loss": 0.9276,
+      "step": 16273
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2765662565393508e-05,
+      "loss": 1.2894,
+      "step": 16274
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2764843050895548e-05,
+      "loss": 0.7735,
+      "step": 16275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.276402351629134e-05,
+      "loss": 0.936,
+      "step": 16276
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.276320396158685e-05,
+      "loss": 1.0231,
+      "step": 16277
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2762384386788032e-05,
+      "loss": 0.9883,
+      "step": 16278
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2761564791900845e-05,
+      "loss": 1.0569,
+      "step": 16279
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2760745176931255e-05,
+      "loss": 1.1912,
+      "step": 16280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2759925541885221e-05,
+      "loss": 0.9156,
+      "step": 16281
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27591058867687e-05,
+      "loss": 1.0905,
+      "step": 16282
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2758286211587656e-05,
+      "loss": 1.0348,
+      "step": 16283
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2757466516348047e-05,
+      "loss": 0.9256,
+      "step": 16284
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2756646801055835e-05,
+      "loss": 0.9448,
+      "step": 16285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2755827065716981e-05,
+      "loss": 0.7765,
+      "step": 16286
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2755007310337449e-05,
+      "loss": 0.6548,
+      "step": 16287
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2754187534923195e-05,
+      "loss": 1.0141,
+      "step": 16288
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2753367739480185e-05,
+      "loss": 1.2158,
+      "step": 16289
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2752547924014378e-05,
+      "loss": 1.1054,
+      "step": 16290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2751728088531736e-05,
+      "loss": 0.7317,
+      "step": 16291
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750908233038226e-05,
+      "loss": 0.6354,
+      "step": 16292
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750088357539802e-05,
+      "loss": 0.7469,
+      "step": 16293
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.274926846204243e-05,
+      "loss": 1.1763,
+      "step": 16294
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2748448546552074e-05,
+      "loss": 0.9188,
+      "step": 16295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2747628611074696e-05,
+      "loss": 0.6789,
+      "step": 16296
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2746808655616256e-05,
+      "loss": 0.6711,
+      "step": 16297
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2745988680182719e-05,
+      "loss": 0.66,
+      "step": 16298
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2745168684780048e-05,
+      "loss": 0.7146,
+      "step": 16299
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2744348669414205e-05,
+      "loss": 0.9812,
+      "step": 16300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.274352863409115e-05,
+      "loss": 1.1127,
+      "step": 16301
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2742708578816859e-05,
+      "loss": 1.0592,
+      "step": 16302
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2741888503597277e-05,
+      "loss": 0.8523,
+      "step": 16303
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2741068408438382e-05,
+      "loss": 1.2118,
+      "step": 16304
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2740248293346134e-05,
+      "loss": 1.3294,
+      "step": 16305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2739428158326494e-05,
+      "loss": 0.8554,
+      "step": 16306
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2738608003385429e-05,
+      "loss": 1.3068,
+      "step": 16307
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2737787828528902e-05,
+      "loss": 1.229,
+      "step": 16308
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2736967633762879e-05,
+      "loss": 0.8585,
+      "step": 16309
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273614741909332e-05,
+      "loss": 0.8726,
+      "step": 16310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2735327184526196e-05,
+      "loss": 1.024,
+      "step": 16311
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273450693006747e-05,
+      "loss": 0.6676,
+      "step": 16312
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2733686655723105e-05,
+      "loss": 0.7357,
+      "step": 16313
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2732866361499068e-05,
+      "loss": 1.0053,
+      "step": 16314
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2732046047401321e-05,
+      "loss": 0.9061,
+      "step": 16315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273122571343583e-05,
+      "loss": 1.2426,
+      "step": 16316
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273040535960857e-05,
+      "loss": 0.9851,
+      "step": 16317
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2729584985925492e-05,
+      "loss": 1.0483,
+      "step": 16318
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272876459239257e-05,
+      "loss": 0.9951,
+      "step": 16319
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2727944179015773e-05,
+      "loss": 0.6603,
+      "step": 16320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2727123745801058e-05,
+      "loss": 0.6649,
+      "step": 16321
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2726303292754399e-05,
+      "loss": 1.0382,
+      "step": 16322
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272548281988176e-05,
+      "loss": 0.8794,
+      "step": 16323
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2724662327189105e-05,
+      "loss": 1.1358,
+      "step": 16324
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2723841814682403e-05,
+      "loss": 1.5351,
+      "step": 16325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2723021282367622e-05,
+      "loss": 1.0512,
+      "step": 16326
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272220073025073e-05,
+      "loss": 0.7148,
+      "step": 16327
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2721380158337688e-05,
+      "loss": 0.6707,
+      "step": 16328
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272055956663447e-05,
+      "loss": 0.843,
+      "step": 16329
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271973895514704e-05,
+      "loss": 0.9637,
+      "step": 16330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2718918323881364e-05,
+      "loss": 0.9091,
+      "step": 16331
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2718097672843416e-05,
+      "loss": 0.8472,
+      "step": 16332
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2717277002039159e-05,
+      "loss": 0.9085,
+      "step": 16333
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2716456311474561e-05,
+      "loss": 0.8803,
+      "step": 16334
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271563560115559e-05,
+      "loss": 0.8022,
+      "step": 16335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2714814871088218e-05,
+      "loss": 0.9951,
+      "step": 16336
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2713994121278409e-05,
+      "loss": 0.7933,
+      "step": 16337
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2713173351732134e-05,
+      "loss": 0.9085,
+      "step": 16338
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2712352562455358e-05,
+      "loss": 0.8636,
+      "step": 16339
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2711531753454057e-05,
+      "loss": 0.9532,
+      "step": 16340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2710710924734196e-05,
+      "loss": 0.8478,
+      "step": 16341
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2709890076301744e-05,
+      "loss": 0.9064,
+      "step": 16342
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2709069208162665e-05,
+      "loss": 0.881,
+      "step": 16343
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270824832032294e-05,
+      "loss": 1.1356,
+      "step": 16344
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270742741278853e-05,
+      "loss": 0.9107,
+      "step": 16345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2706606485565409e-05,
+      "loss": 0.7895,
+      "step": 16346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2705785538659542e-05,
+      "loss": 1.0409,
+      "step": 16347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2704964572076906e-05,
+      "loss": 1.1202,
+      "step": 16348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2704143585823463e-05,
+      "loss": 1.0811,
+      "step": 16349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2703322579905191e-05,
+      "loss": 1.2547,
+      "step": 16350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2702501554328058e-05,
+      "loss": 1.1646,
+      "step": 16351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2701680509098029e-05,
+      "loss": 0.5712,
+      "step": 16352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2700859444221081e-05,
+      "loss": 1.0459,
+      "step": 16353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2700038359703184e-05,
+      "loss": 0.7495,
+      "step": 16354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2699217255550306e-05,
+      "loss": 1.012,
+      "step": 16355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2698396131768424e-05,
+      "loss": 0.6925,
+      "step": 16356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2697574988363506e-05,
+      "loss": 1.2204,
+      "step": 16357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269675382534152e-05,
+      "loss": 0.9707,
+      "step": 16358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2695932642708442e-05,
+      "loss": 1.2003,
+      "step": 16359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2695111440470243e-05,
+      "loss": 1.0823,
+      "step": 16360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2694290218632892e-05,
+      "loss": 0.925,
+      "step": 16361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2693468977202366e-05,
+      "loss": 0.9885,
+      "step": 16362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2692647716184634e-05,
+      "loss": 1.0992,
+      "step": 16363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2691826435585669e-05,
+      "loss": 0.9888,
+      "step": 16364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269100513541144e-05,
+      "loss": 1.0211,
+      "step": 16365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269018381566793e-05,
+      "loss": 0.7643,
+      "step": 16366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2689362476361096e-05,
+      "loss": 0.6765,
+      "step": 16367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2688541117496923e-05,
+      "loss": 1.2175,
+      "step": 16368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2687719739081384e-05,
+      "loss": 0.7477,
+      "step": 16369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2686898341120442e-05,
+      "loss": 0.7439,
+      "step": 16370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2686076923620084e-05,
+      "loss": 1.2745,
+      "step": 16371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2685255486586271e-05,
+      "loss": 0.9121,
+      "step": 16372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2684434030024982e-05,
+      "loss": 0.6875,
+      "step": 16373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2683612553942193e-05,
+      "loss": 0.9758,
+      "step": 16374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2682791058343875e-05,
+      "loss": 1.0386,
+      "step": 16375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2681969543236004e-05,
+      "loss": 1.1202,
+      "step": 16376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2681148008624548e-05,
+      "loss": 0.8745,
+      "step": 16377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268032645451549e-05,
+      "loss": 1.1836,
+      "step": 16378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2679504880914798e-05,
+      "loss": 1.1055,
+      "step": 16379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2678683287828451e-05,
+      "loss": 0.6515,
+      "step": 16380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2677861675262422e-05,
+      "loss": 0.9724,
+      "step": 16381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2677040043222682e-05,
+      "loss": 1.206,
+      "step": 16382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2676218391715214e-05,
+      "loss": 0.9021,
+      "step": 16383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2675396720745984e-05,
+      "loss": 0.9712,
+      "step": 16384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2674575030320976e-05,
+      "loss": 1.1112,
+      "step": 16385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267375332044616e-05,
+      "loss": 0.9483,
+      "step": 16386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2672931591127513e-05,
+      "loss": 0.8604,
+      "step": 16387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267210984237101e-05,
+      "loss": 1.1563,
+      "step": 16388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2671288074182626e-05,
+      "loss": 1.0798,
+      "step": 16389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2670466286568341e-05,
+      "loss": 0.8956,
+      "step": 16390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2669644479534129e-05,
+      "loss": 0.8794,
+      "step": 16391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2668822653085967e-05,
+      "loss": 1.2915,
+      "step": 16392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2668000807229827e-05,
+      "loss": 0.8727,
+      "step": 16393
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.266717894197169e-05,
+      "loss": 0.8309,
+      "step": 16394
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.266635705731753e-05,
+      "loss": 0.8935,
+      "step": 16395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2665535153273329e-05,
+      "loss": 0.6944,
+      "step": 16396
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2664713229845056e-05,
+      "loss": 0.8369,
+      "step": 16397
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2663891287038696e-05,
+      "loss": 0.9586,
+      "step": 16398
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2663069324860221e-05,
+      "loss": 0.7347,
+      "step": 16399
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2662247343315612e-05,
+      "loss": 0.7153,
+      "step": 16400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2661425342410842e-05,
+      "loss": 0.9689,
+      "step": 16401
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2660603322151894e-05,
+      "loss": 0.9896,
+      "step": 16402
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2659781282544743e-05,
+      "loss": 0.9421,
+      "step": 16403
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2658959223595364e-05,
+      "loss": 1.0437,
+      "step": 16404
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2658137145309742e-05,
+      "loss": 0.8368,
+      "step": 16405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2657315047693848e-05,
+      "loss": 1.2027,
+      "step": 16406
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2656492930753668e-05,
+      "loss": 1.0385,
+      "step": 16407
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2655670794495174e-05,
+      "loss": 0.8427,
+      "step": 16408
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2654848638924346e-05,
+      "loss": 1.0482,
+      "step": 16409
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2654026464047165e-05,
+      "loss": 0.6363,
+      "step": 16410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.265320426986961e-05,
+      "loss": 1.1005,
+      "step": 16411
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2652382056397658e-05,
+      "loss": 0.9273,
+      "step": 16412
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2651559823637288e-05,
+      "loss": 1.1123,
+      "step": 16413
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2650737571594482e-05,
+      "loss": 0.8446,
+      "step": 16414
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2649915300275218e-05,
+      "loss": 0.742,
+      "step": 16415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2649093009685476e-05,
+      "loss": 1.1081,
+      "step": 16416
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2648270699831233e-05,
+      "loss": 1.0526,
+      "step": 16417
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2647448370718474e-05,
+      "loss": 1.1954,
+      "step": 16418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2646626022353174e-05,
+      "loss": 1.0655,
+      "step": 16419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2645803654741318e-05,
+      "loss": 1.1828,
+      "step": 16420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2644981267888886e-05,
+      "loss": 0.8059,
+      "step": 16421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2644158861801851e-05,
+      "loss": 0.8335,
+      "step": 16422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2643336436486201e-05,
+      "loss": 0.9613,
+      "step": 16423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2642513991947918e-05,
+      "loss": 1.0552,
+      "step": 16424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2641691528192976e-05,
+      "loss": 0.9997,
+      "step": 16425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2640869045227361e-05,
+      "loss": 0.8017,
+      "step": 16426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2640046543057056e-05,
+      "loss": 0.7243,
+      "step": 16427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2639224021688035e-05,
+      "loss": 1.0914,
+      "step": 16428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2638401481126286e-05,
+      "loss": 0.6655,
+      "step": 16429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637578921377789e-05,
+      "loss": 1.1495,
+      "step": 16430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2636756342448524e-05,
+      "loss": 0.8668,
+      "step": 16431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2635933744344472e-05,
+      "loss": 1.1051,
+      "step": 16432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2635111127071622e-05,
+      "loss": 0.992,
+      "step": 16433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2634288490635946e-05,
+      "loss": 0.9076,
+      "step": 16434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2633465835043434e-05,
+      "loss": 0.7646,
+      "step": 16435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2632643160300071e-05,
+      "loss": 0.7966,
+      "step": 16436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2631820466411826e-05,
+      "loss": 0.9031,
+      "step": 16437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2630997753384695e-05,
+      "loss": 0.7905,
+      "step": 16438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2630175021224655e-05,
+      "loss": 1.0843,
+      "step": 16439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262935226993769e-05,
+      "loss": 0.8418,
+      "step": 16440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2628529499529787e-05,
+      "loss": 0.9312,
+      "step": 16441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2627706710006922e-05,
+      "loss": 1.0173,
+      "step": 16442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262688390137508e-05,
+      "loss": 0.9546,
+      "step": 16443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2626061073640251e-05,
+      "loss": 0.7634,
+      "step": 16444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2625238226808411e-05,
+      "loss": 0.9609,
+      "step": 16445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262441536088555e-05,
+      "loss": 1.0438,
+      "step": 16446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2623592475877647e-05,
+      "loss": 0.9467,
+      "step": 16447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262276957179069e-05,
+      "loss": 0.835,
+      "step": 16448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2621946648630656e-05,
+      "loss": 1.2913,
+      "step": 16449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262112370640354e-05,
+      "loss": 0.8286,
+      "step": 16450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.262030074511532e-05,
+      "loss": 1.1573,
+      "step": 16451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2619477764771983e-05,
+      "loss": 1.1258,
+      "step": 16452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2618654765379512e-05,
+      "loss": 1.0952,
+      "step": 16453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2617831746943891e-05,
+      "loss": 1.025,
+      "step": 16454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261700870947111e-05,
+      "loss": 1.0077,
+      "step": 16455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261618565296715e-05,
+      "loss": 1.1971,
+      "step": 16456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2615362577437997e-05,
+      "loss": 0.8016,
+      "step": 16457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2614539482889637e-05,
+      "loss": 0.8621,
+      "step": 16458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2613716369328055e-05,
+      "loss": 1.189,
+      "step": 16459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2612893236759238e-05,
+      "loss": 1.2074,
+      "step": 16460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261207008518917e-05,
+      "loss": 1.0271,
+      "step": 16461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261124691462384e-05,
+      "loss": 1.1168,
+      "step": 16462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2610423725069232e-05,
+      "loss": 1.1133,
+      "step": 16463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2609600516531334e-05,
+      "loss": 1.0131,
+      "step": 16464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2608777289016128e-05,
+      "loss": 0.957,
+      "step": 16465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2607954042529607e-05,
+      "loss": 1.1217,
+      "step": 16466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2607130777077752e-05,
+      "loss": 0.7901,
+      "step": 16467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2606307492666553e-05,
+      "loss": 0.9266,
+      "step": 16468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2605484189301997e-05,
+      "loss": 0.6727,
+      "step": 16469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2604660866990072e-05,
+      "loss": 0.9196,
+      "step": 16470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2603837525736764e-05,
+      "loss": 0.8173,
+      "step": 16471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2603014165548059e-05,
+      "loss": 0.8983,
+      "step": 16472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2602190786429946e-05,
+      "loss": 1.0743,
+      "step": 16473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.260136738838841e-05,
+      "loss": 0.9267,
+      "step": 16474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2600543971429446e-05,
+      "loss": 0.9688,
+      "step": 16475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2599720535559034e-05,
+      "loss": 0.7868,
+      "step": 16476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2598897080783168e-05,
+      "loss": 1.0565,
+      "step": 16477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2598073607107834e-05,
+      "loss": 1.1805,
+      "step": 16478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259725011453902e-05,
+      "loss": 1.1714,
+      "step": 16479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2596426603082711e-05,
+      "loss": 1.0078,
+      "step": 16480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2595603072744906e-05,
+      "loss": 0.8433,
+      "step": 16481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2594779523531584e-05,
+      "loss": 1.0318,
+      "step": 16482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593955955448735e-05,
+      "loss": 0.5867,
+      "step": 16483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593132368502352e-05,
+      "loss": 0.8574,
+      "step": 16484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2592308762698422e-05,
+      "loss": 0.9608,
+      "step": 16485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2591485138042937e-05,
+      "loss": 0.5974,
+      "step": 16486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2590661494541883e-05,
+      "loss": 0.9793,
+      "step": 16487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258983783220125e-05,
+      "loss": 1.0278,
+      "step": 16488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2589014151027027e-05,
+      "loss": 0.7915,
+      "step": 16489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2588190451025209e-05,
+      "loss": 0.8615,
+      "step": 16490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2587366732201783e-05,
+      "loss": 0.6717,
+      "step": 16491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2586542994562736e-05,
+      "loss": 0.8125,
+      "step": 16492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2585719238114062e-05,
+      "loss": 0.8463,
+      "step": 16493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258489546286175e-05,
+      "loss": 1.1036,
+      "step": 16494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2584071668811792e-05,
+      "loss": 1.0064,
+      "step": 16495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258324785597018e-05,
+      "loss": 0.8814,
+      "step": 16496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2582424024342901e-05,
+      "loss": 0.9169,
+      "step": 16497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2581600173935946e-05,
+      "loss": 0.9731,
+      "step": 16498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2580776304755308e-05,
+      "loss": 1.2052,
+      "step": 16499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257995241680698e-05,
+      "loss": 0.6469,
+      "step": 16500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2579128510096952e-05,
+      "loss": 0.8892,
+      "step": 16501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2578304584631214e-05,
+      "loss": 0.9732,
+      "step": 16502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257748064041576e-05,
+      "loss": 1.0017,
+      "step": 16503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2576656677456578e-05,
+      "loss": 1.3124,
+      "step": 16504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2575832695759663e-05,
+      "loss": 0.7519,
+      "step": 16505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2575008695331011e-05,
+      "loss": 1.1049,
+      "step": 16506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2574184676176602e-05,
+      "loss": 1.1468,
+      "step": 16507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2573360638302441e-05,
+      "loss": 1.024,
+      "step": 16508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2572536581714515e-05,
+      "loss": 0.8765,
+      "step": 16509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2571712506418819e-05,
+      "loss": 1.1972,
+      "step": 16510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2570888412421342e-05,
+      "loss": 0.7994,
+      "step": 16511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257006429972808e-05,
+      "loss": 0.8286,
+      "step": 16512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2569240168345023e-05,
+      "loss": 1.3142,
+      "step": 16513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2568416018278167e-05,
+      "loss": 0.8569,
+      "step": 16514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2567591849533507e-05,
+      "loss": 1.0241,
+      "step": 16515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2566767662117031e-05,
+      "loss": 1.1628,
+      "step": 16516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2565943456034736e-05,
+      "loss": 0.7008,
+      "step": 16517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2565119231292617e-05,
+      "loss": 0.8485,
+      "step": 16518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2564294987896666e-05,
+      "loss": 0.9027,
+      "step": 16519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2563470725852874e-05,
+      "loss": 0.8809,
+      "step": 16520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2562646445167242e-05,
+      "loss": 0.933,
+      "step": 16521
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2561822145845757e-05,
+      "loss": 0.7813,
+      "step": 16522
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.256099782789442e-05,
+      "loss": 0.888,
+      "step": 16523
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560173491319221e-05,
+      "loss": 1.1273,
+      "step": 16524
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2559349136126156e-05,
+      "loss": 1.1121,
+      "step": 16525
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255852476232122e-05,
+      "loss": 1.0454,
+      "step": 16526
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2557700369910408e-05,
+      "loss": 0.8535,
+      "step": 16527
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2556875958899714e-05,
+      "loss": 1.0397,
+      "step": 16528
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2556051529295133e-05,
+      "loss": 0.8732,
+      "step": 16529
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2555227081102663e-05,
+      "loss": 0.9709,
+      "step": 16530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2554402614328299e-05,
+      "loss": 0.8099,
+      "step": 16531
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2553578128978035e-05,
+      "loss": 0.664,
+      "step": 16532
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2552753625057868e-05,
+      "loss": 0.82,
+      "step": 16533
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2551929102573793e-05,
+      "loss": 0.9236,
+      "step": 16534
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2551104561531802e-05,
+      "loss": 1.2988,
+      "step": 16535
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2550280001937902e-05,
+      "loss": 0.6493,
+      "step": 16536
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2549455423798077e-05,
+      "loss": 1.2006,
+      "step": 16537
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2548630827118331e-05,
+      "loss": 0.9544,
+      "step": 16538
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2547806211904658e-05,
+      "loss": 0.9964,
+      "step": 16539
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2546981578163058e-05,
+      "loss": 0.8251,
+      "step": 16540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2546156925899524e-05,
+      "loss": 0.9386,
+      "step": 16541
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2545332255120053e-05,
+      "loss": 1.0656,
+      "step": 16542
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2544507565830644e-05,
+      "loss": 1.144,
+      "step": 16543
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2543682858037292e-05,
+      "loss": 0.8385,
+      "step": 16544
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2542858131745997e-05,
+      "loss": 0.8975,
+      "step": 16545
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2542033386962755e-05,
+      "loss": 0.7839,
+      "step": 16546
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2541208623693564e-05,
+      "loss": 0.8142,
+      "step": 16547
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2540383841944422e-05,
+      "loss": 0.8165,
+      "step": 16548
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2539559041721325e-05,
+      "loss": 1.0899,
+      "step": 16549
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2538734223030274e-05,
+      "loss": 0.8556,
+      "step": 16550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2537909385877267e-05,
+      "loss": 0.8655,
+      "step": 16551
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25370845302683e-05,
+      "loss": 0.7215,
+      "step": 16552
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253625965620937e-05,
+      "loss": 0.8867,
+      "step": 16553
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2535434763706479e-05,
+      "loss": 0.7765,
+      "step": 16554
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2534609852765629e-05,
+      "loss": 1.0061,
+      "step": 16555
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2533784923392811e-05,
+      "loss": 1.0956,
+      "step": 16556
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253295997559403e-05,
+      "loss": 0.9906,
+      "step": 16557
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2532135009375281e-05,
+      "loss": 1.1679,
+      "step": 16558
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2531310024742565e-05,
+      "loss": 0.8952,
+      "step": 16559
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253048502170188e-05,
+      "loss": 0.8434,
+      "step": 16560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2529660000259231e-05,
+      "loss": 0.8477,
+      "step": 16561
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252883496042061e-05,
+      "loss": 0.9495,
+      "step": 16562
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2528009902192025e-05,
+      "loss": 1.0732,
+      "step": 16563
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2527184825579467e-05,
+      "loss": 0.7553,
+      "step": 16564
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252635973058894e-05,
+      "loss": 0.888,
+      "step": 16565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2525534617226446e-05,
+      "loss": 0.7529,
+      "step": 16566
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2524709485497986e-05,
+      "loss": 0.732,
+      "step": 16567
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2523884335409553e-05,
+      "loss": 0.9128,
+      "step": 16568
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2523059166967158e-05,
+      "loss": 1.0163,
+      "step": 16569
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2522233980176793e-05,
+      "loss": 0.8149,
+      "step": 16570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2521408775044463e-05,
+      "loss": 1.0611,
+      "step": 16571
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2520583551576171e-05,
+      "loss": 0.5797,
+      "step": 16572
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2519758309777914e-05,
+      "loss": 1.2559,
+      "step": 16573
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2518933049655692e-05,
+      "loss": 0.9372,
+      "step": 16574
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2518107771215511e-05,
+      "loss": 0.8679,
+      "step": 16575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2517282474463375e-05,
+      "loss": 0.8233,
+      "step": 16576
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2516457159405275e-05,
+      "loss": 0.9484,
+      "step": 16577
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.251563182604722e-05,
+      "loss": 0.9723,
+      "step": 16578
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2514806474395214e-05,
+      "loss": 0.7511,
+      "step": 16579
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2513981104455255e-05,
+      "loss": 1.2839,
+      "step": 16580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2513155716233345e-05,
+      "loss": 1.12,
+      "step": 16581
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2512330309735489e-05,
+      "loss": 1.0695,
+      "step": 16582
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511504884967687e-05,
+      "loss": 1.0451,
+      "step": 16583
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2510679441935942e-05,
+      "loss": 0.7022,
+      "step": 16584
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2509853980646259e-05,
+      "loss": 1.1878,
+      "step": 16585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2509028501104639e-05,
+      "loss": 0.9223,
+      "step": 16586
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2508203003317086e-05,
+      "loss": 1.0936,
+      "step": 16587
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25073774872896e-05,
+      "loss": 0.9248,
+      "step": 16588
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2506551953028188e-05,
+      "loss": 0.6737,
+      "step": 16589
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2505726400538849e-05,
+      "loss": 0.8847,
+      "step": 16590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2504900829827593e-05,
+      "loss": 0.6307,
+      "step": 16591
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2504075240900417e-05,
+      "loss": 0.6501,
+      "step": 16592
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250324963376333e-05,
+      "loss": 0.8192,
+      "step": 16593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2502424008422333e-05,
+      "loss": 1.2966,
+      "step": 16594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2501598364883432e-05,
+      "loss": 0.8464,
+      "step": 16595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2500772703152628e-05,
+      "loss": 0.9749,
+      "step": 16596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2499947023235927e-05,
+      "loss": 0.9432,
+      "step": 16597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2499121325139336e-05,
+      "loss": 1.068,
+      "step": 16598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2498295608868853e-05,
+      "loss": 0.9597,
+      "step": 16599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2497469874430491e-05,
+      "loss": 0.5743,
+      "step": 16600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2496644121830248e-05,
+      "loss": 0.925,
+      "step": 16601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2495818351074134e-05,
+      "loss": 0.7094,
+      "step": 16602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2494992562168152e-05,
+      "loss": 0.7432,
+      "step": 16603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2494166755118304e-05,
+      "loss": 0.8868,
+      "step": 16604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24933409299306e-05,
+      "loss": 0.9937,
+      "step": 16605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2492515086611045e-05,
+      "loss": 0.9385,
+      "step": 16606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2491689225165641e-05,
+      "loss": 1.061,
+      "step": 16607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490863345600398e-05,
+      "loss": 1.0321,
+      "step": 16608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249003744792132e-05,
+      "loss": 0.9591,
+      "step": 16609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2489211532134413e-05,
+      "loss": 0.9661,
+      "step": 16610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2488385598245686e-05,
+      "loss": 0.8559,
+      "step": 16611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2487559646261138e-05,
+      "loss": 0.902,
+      "step": 16612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2486733676186782e-05,
+      "loss": 0.7822,
+      "step": 16613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2485907688028623e-05,
+      "loss": 0.8356,
+      "step": 16614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2485081681792665e-05,
+      "loss": 0.7536,
+      "step": 16615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2484255657484923e-05,
+      "loss": 0.8396,
+      "step": 16616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2483429615111392e-05,
+      "loss": 0.7936,
+      "step": 16617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2482603554678085e-05,
+      "loss": 0.9873,
+      "step": 16618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2481777476191013e-05,
+      "loss": 1.0494,
+      "step": 16619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2480951379656175e-05,
+      "loss": 1.1361,
+      "step": 16620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2480125265079589e-05,
+      "loss": 0.8465,
+      "step": 16621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2479299132467254e-05,
+      "loss": 0.8905,
+      "step": 16622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2478472981825179e-05,
+      "loss": 0.787,
+      "step": 16623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2477646813159373e-05,
+      "loss": 0.9237,
+      "step": 16624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2476820626475846e-05,
+      "loss": 0.8176,
+      "step": 16625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2475994421780604e-05,
+      "loss": 0.7497,
+      "step": 16626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2475168199079654e-05,
+      "loss": 0.8168,
+      "step": 16627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2474341958379008e-05,
+      "loss": 0.7459,
+      "step": 16628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247351569968467e-05,
+      "loss": 0.5667,
+      "step": 16629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2472689423002653e-05,
+      "loss": 1.2172,
+      "step": 16630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2471863128338968e-05,
+      "loss": 0.8955,
+      "step": 16631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2471036815699613e-05,
+      "loss": 1.038,
+      "step": 16632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2470210485090607e-05,
+      "loss": 0.7894,
+      "step": 16633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2469384136517956e-05,
+      "loss": 1.0582,
+      "step": 16634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2468557769987667e-05,
+      "loss": 0.8697,
+      "step": 16635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2467731385505757e-05,
+      "loss": 0.7421,
+      "step": 16636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2466904983078228e-05,
+      "loss": 0.8795,
+      "step": 16637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2466078562711087e-05,
+      "loss": 0.887,
+      "step": 16638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2465252124410354e-05,
+      "loss": 1.0319,
+      "step": 16639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2464425668182034e-05,
+      "loss": 0.9783,
+      "step": 16640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2463599194032137e-05,
+      "loss": 0.7796,
+      "step": 16641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246277270196667e-05,
+      "loss": 1.1731,
+      "step": 16642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246194619199165e-05,
+      "loss": 0.9621,
+      "step": 16643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2461119664113079e-05,
+      "loss": 0.745,
+      "step": 16644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2460293118336977e-05,
+      "loss": 1.0709,
+      "step": 16645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2459466554669351e-05,
+      "loss": 1.0438,
+      "step": 16646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2458639973116207e-05,
+      "loss": 0.9732,
+      "step": 16647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245781337368356e-05,
+      "loss": 0.8899,
+      "step": 16648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2456986756377424e-05,
+      "loss": 0.7015,
+      "step": 16649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2456160121203808e-05,
+      "loss": 0.9556,
+      "step": 16650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2455333468168723e-05,
+      "loss": 0.787,
+      "step": 16651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2454506797278178e-05,
+      "loss": 0.6876,
+      "step": 16652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2453680108538186e-05,
+      "loss": 0.8453,
+      "step": 16653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.245285340195476e-05,
+      "loss": 0.9609,
+      "step": 16654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2452026677533914e-05,
+      "loss": 0.9364,
+      "step": 16655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2451199935281659e-05,
+      "loss": 1.1389,
+      "step": 16656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2450373175204005e-05,
+      "loss": 0.7631,
+      "step": 16657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2449546397306964e-05,
+      "loss": 1.085,
+      "step": 16658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2448719601596549e-05,
+      "loss": 1.2378,
+      "step": 16659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2447892788078773e-05,
+      "loss": 1.2806,
+      "step": 16660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2447065956759653e-05,
+      "loss": 1.0883,
+      "step": 16661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244623910764519e-05,
+      "loss": 0.9227,
+      "step": 16662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244541224074141e-05,
+      "loss": 1.0452,
+      "step": 16663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244458535605432e-05,
+      "loss": 0.9232,
+      "step": 16664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2443758453589934e-05,
+      "loss": 0.9,
+      "step": 16665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2442931533354264e-05,
+      "loss": 0.8631,
+      "step": 16666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2442104595353324e-05,
+      "loss": 0.923,
+      "step": 16667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244127763959313e-05,
+      "loss": 0.7969,
+      "step": 16668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2440450666079692e-05,
+      "loss": 1.3185,
+      "step": 16669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2439623674819026e-05,
+      "loss": 0.9081,
+      "step": 16670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2438796665817147e-05,
+      "loss": 0.9193,
+      "step": 16671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2437969639080067e-05,
+      "loss": 0.8598,
+      "step": 16672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2437142594613803e-05,
+      "loss": 1.0055,
+      "step": 16673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2436315532424366e-05,
+      "loss": 1.0042,
+      "step": 16674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.243548845251777e-05,
+      "loss": 1.324,
+      "step": 16675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2434661354900035e-05,
+      "loss": 0.92,
+      "step": 16676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2433834239577171e-05,
+      "loss": 1.2028,
+      "step": 16677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2433007106555193e-05,
+      "loss": 0.8052,
+      "step": 16678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2432179955840118e-05,
+      "loss": 0.8879,
+      "step": 16679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.243135278743796e-05,
+      "loss": 0.9057,
+      "step": 16680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2430525601354735e-05,
+      "loss": 0.8459,
+      "step": 16681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2429698397596458e-05,
+      "loss": 0.842,
+      "step": 16682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2428871176169144e-05,
+      "loss": 1.018,
+      "step": 16683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2428043937078807e-05,
+      "loss": 0.8495,
+      "step": 16684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2427216680331467e-05,
+      "loss": 0.7475,
+      "step": 16685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2426389405933142e-05,
+      "loss": 0.9049,
+      "step": 16686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2425562113889835e-05,
+      "loss": 0.644,
+      "step": 16687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2424734804207576e-05,
+      "loss": 0.7547,
+      "step": 16688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2423907476892376e-05,
+      "loss": 1.3434,
+      "step": 16689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2423080131950252e-05,
+      "loss": 0.8167,
+      "step": 16690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2422252769387219e-05,
+      "loss": 0.7922,
+      "step": 16691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2421425389209296e-05,
+      "loss": 1.0812,
+      "step": 16692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2420597991422496e-05,
+      "loss": 1.1026,
+      "step": 16693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2419770576032844e-05,
+      "loss": 0.9413,
+      "step": 16694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2418943143046346e-05,
+      "loss": 1.4265,
+      "step": 16695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2418115692469028e-05,
+      "loss": 0.8505,
+      "step": 16696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2417288224306906e-05,
+      "loss": 1.2878,
+      "step": 16697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2416460738565992e-05,
+      "loss": 1.0605,
+      "step": 16698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2415633235252304e-05,
+      "loss": 0.6554,
+      "step": 16699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2414805714371869e-05,
+      "loss": 0.7856,
+      "step": 16700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2413978175930698e-05,
+      "loss": 0.8512,
+      "step": 16701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2413150619934808e-05,
+      "loss": 0.7892,
+      "step": 16702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241232304639022e-05,
+      "loss": 0.6415,
+      "step": 16703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2411495455302948e-05,
+      "loss": 0.651,
+      "step": 16704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2410667846679014e-05,
+      "loss": 0.7231,
+      "step": 16705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240984022052444e-05,
+      "loss": 0.7005,
+      "step": 16706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2409012576845238e-05,
+      "loss": 1.0055,
+      "step": 16707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2408184915647426e-05,
+      "loss": 0.9546,
+      "step": 16708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2407357236937028e-05,
+      "loss": 0.9228,
+      "step": 16709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2406529540720063e-05,
+      "loss": 0.9267,
+      "step": 16710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2405701827002546e-05,
+      "loss": 0.7187,
+      "step": 16711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2404874095790498e-05,
+      "loss": 1.1152,
+      "step": 16712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2404046347089937e-05,
+      "loss": 1.0516,
+      "step": 16713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2403218580906885e-05,
+      "loss": 1.0893,
+      "step": 16714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2402390797247362e-05,
+      "loss": 1.003,
+      "step": 16715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.240156299611739e-05,
+      "loss": 0.7928,
+      "step": 16716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2400735177522978e-05,
+      "loss": 1.126,
+      "step": 16717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2399907341470154e-05,
+      "loss": 0.7202,
+      "step": 16718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239907948796494e-05,
+      "loss": 0.6794,
+      "step": 16719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2398251617013355e-05,
+      "loss": 1.3021,
+      "step": 16720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2397423728621414e-05,
+      "loss": 1.0866,
+      "step": 16721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2396595822795146e-05,
+      "loss": 0.6882,
+      "step": 16722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2395767899540561e-05,
+      "loss": 1.1877,
+      "step": 16723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2394939958863689e-05,
+      "loss": 1.0377,
+      "step": 16724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.239411200077055e-05,
+      "loss": 0.7766,
+      "step": 16725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2393284025267159e-05,
+      "loss": 0.8156,
+      "step": 16726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2392456032359544e-05,
+      "loss": 0.8809,
+      "step": 16727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2391628022053719e-05,
+      "loss": 0.8922,
+      "step": 16728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2390799994355713e-05,
+      "loss": 0.6906,
+      "step": 16729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389971949271539e-05,
+      "loss": 0.8313,
+      "step": 16730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389143886807232e-05,
+      "loss": 1.0808,
+      "step": 16731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2388315806968798e-05,
+      "loss": 1.2865,
+      "step": 16732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387487709762268e-05,
+      "loss": 0.9936,
+      "step": 16733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2386659595193664e-05,
+      "loss": 0.6675,
+      "step": 16734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2385831463269006e-05,
+      "loss": 0.9366,
+      "step": 16735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2385003313994316e-05,
+      "loss": 0.8485,
+      "step": 16736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2384175147375617e-05,
+      "loss": 0.9668,
+      "step": 16737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2383346963418931e-05,
+      "loss": 0.7464,
+      "step": 16738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2382518762130283e-05,
+      "loss": 0.8095,
+      "step": 16739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2381690543515692e-05,
+      "loss": 1.2306,
+      "step": 16740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2380862307581185e-05,
+      "loss": 0.8377,
+      "step": 16741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.238003405433278e-05,
+      "loss": 1.1006,
+      "step": 16742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2379205783776507e-05,
+      "loss": 0.9273,
+      "step": 16743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2378377495918384e-05,
+      "loss": 0.9406,
+      "step": 16744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2377549190764432e-05,
+      "loss": 0.6637,
+      "step": 16745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2376720868320684e-05,
+      "loss": 0.5939,
+      "step": 16746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2375892528593155e-05,
+      "loss": 1.0742,
+      "step": 16747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237506417158787e-05,
+      "loss": 0.9674,
+      "step": 16748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2374235797310856e-05,
+      "loss": 1.0681,
+      "step": 16749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2373407405768137e-05,
+      "loss": 0.9906,
+      "step": 16750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2372578996965733e-05,
+      "loss": 0.6701,
+      "step": 16751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2371750570909675e-05,
+      "loss": 0.8725,
+      "step": 16752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2370922127605982e-05,
+      "loss": 1.0152,
+      "step": 16753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2370093667060677e-05,
+      "loss": 0.8447,
+      "step": 16754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236926518927979e-05,
+      "loss": 1.2796,
+      "step": 16755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2368436694269346e-05,
+      "loss": 0.5464,
+      "step": 16756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2367608182035363e-05,
+      "loss": 1.1881,
+      "step": 16757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2366779652583872e-05,
+      "loss": 1.0917,
+      "step": 16758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2365951105920897e-05,
+      "loss": 0.8887,
+      "step": 16759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2365122542052461e-05,
+      "loss": 1.0314,
+      "step": 16760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2364293960984595e-05,
+      "loss": 0.9179,
+      "step": 16761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2363465362723317e-05,
+      "loss": 0.8116,
+      "step": 16762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2362636747274656e-05,
+      "loss": 1.0133,
+      "step": 16763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2361808114644642e-05,
+      "loss": 1.0388,
+      "step": 16764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2360979464839296e-05,
+      "loss": 0.7805,
+      "step": 16765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2360150797864644e-05,
+      "loss": 0.944,
+      "step": 16766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2359322113726715e-05,
+      "loss": 0.9206,
+      "step": 16767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235849341243153e-05,
+      "loss": 0.7003,
+      "step": 16768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235766469398512e-05,
+      "loss": 0.8338,
+      "step": 16769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356835958393513e-05,
+      "loss": 1.2057,
+      "step": 16770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356007205662734e-05,
+      "loss": 0.7095,
+      "step": 16771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2355178435798805e-05,
+      "loss": 0.9556,
+      "step": 16772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235434964880776e-05,
+      "loss": 0.9388,
+      "step": 16773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2353520844695622e-05,
+      "loss": 0.8226,
+      "step": 16774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2352692023468418e-05,
+      "loss": 0.9712,
+      "step": 16775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351863185132178e-05,
+      "loss": 0.8791,
+      "step": 16776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351034329692928e-05,
+      "loss": 1.289,
+      "step": 16777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2350205457156691e-05,
+      "loss": 0.969,
+      "step": 16778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2349376567529503e-05,
+      "loss": 1.0672,
+      "step": 16779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2348547660817388e-05,
+      "loss": 1.2096,
+      "step": 16780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2347718737026372e-05,
+      "loss": 0.905,
+      "step": 16781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2346889796162485e-05,
+      "loss": 0.9006,
+      "step": 16782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2346060838231754e-05,
+      "loss": 0.8055,
+      "step": 16783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345231863240211e-05,
+      "loss": 0.875,
+      "step": 16784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2344402871193876e-05,
+      "loss": 1.0763,
+      "step": 16785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234357386209879e-05,
+      "loss": 0.7553,
+      "step": 16786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234274483596097e-05,
+      "loss": 1.0556,
+      "step": 16787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2341915792786449e-05,
+      "loss": 0.7702,
+      "step": 16788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234108673258126e-05,
+      "loss": 0.5091,
+      "step": 16789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2340257655351426e-05,
+      "loss": 0.7786,
+      "step": 16790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2339428561102977e-05,
+      "loss": 0.6892,
+      "step": 16791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2338599449841946e-05,
+      "loss": 0.6738,
+      "step": 16792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2337770321574361e-05,
+      "loss": 0.8234,
+      "step": 16793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2336941176306248e-05,
+      "loss": 0.9903,
+      "step": 16794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.233611201404364e-05,
+      "loss": 0.9469,
+      "step": 16795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2335282834792565e-05,
+      "loss": 1.1892,
+      "step": 16796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2334453638559057e-05,
+      "loss": 1.0259,
+      "step": 16797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.233362442534914e-05,
+      "loss": 1.0762,
+      "step": 16798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332795195168847e-05,
+      "loss": 1.1973,
+      "step": 16799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2331965948024209e-05,
+      "loss": 0.949,
+      "step": 16800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2331136683921259e-05,
+      "loss": 1.1202,
+      "step": 16801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2330307402866019e-05,
+      "loss": 0.8439,
+      "step": 16802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2329478104864528e-05,
+      "loss": 0.7872,
+      "step": 16803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232864878992281e-05,
+      "loss": 1.0385,
+      "step": 16804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2327819458046903e-05,
+      "loss": 0.6962,
+      "step": 16805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2326990109242834e-05,
+      "loss": 1.0644,
+      "step": 16806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2326160743516634e-05,
+      "loss": 0.922,
+      "step": 16807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2325331360874335e-05,
+      "loss": 0.9672,
+      "step": 16808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2324501961321966e-05,
+      "loss": 0.8719,
+      "step": 16809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2323672544865562e-05,
+      "loss": 0.8809,
+      "step": 16810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322843111511155e-05,
+      "loss": 0.9501,
+      "step": 16811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2322013661264773e-05,
+      "loss": 0.8111,
+      "step": 16812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232118419413245e-05,
+      "loss": 1.2234,
+      "step": 16813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232035471012022e-05,
+      "loss": 0.8557,
+      "step": 16814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2319525209234109e-05,
+      "loss": 0.9692,
+      "step": 16815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2318695691480155e-05,
+      "loss": 0.8404,
+      "step": 16816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231786615686439e-05,
+      "loss": 0.9747,
+      "step": 16817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2317036605392843e-05,
+      "loss": 1.0458,
+      "step": 16818
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2316207037071549e-05,
+      "loss": 0.9594,
+      "step": 16819
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2315377451906539e-05,
+      "loss": 1.2598,
+      "step": 16820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231454784990385e-05,
+      "loss": 0.6352,
+      "step": 16821
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2313718231069508e-05,
+      "loss": 1.1581,
+      "step": 16822
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2312888595409553e-05,
+      "loss": 1.0425,
+      "step": 16823
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2312058942930011e-05,
+      "loss": 1.2017,
+      "step": 16824
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2311229273636927e-05,
+      "loss": 0.8397,
+      "step": 16825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2310399587536324e-05,
+      "loss": 0.9075,
+      "step": 16826
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2309569884634236e-05,
+      "loss": 0.9114,
+      "step": 16827
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.23087401649367e-05,
+      "loss": 1.1261,
+      "step": 16828
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2307910428449751e-05,
+      "loss": 0.9673,
+      "step": 16829
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.230708067517942e-05,
+      "loss": 0.8418,
+      "step": 16830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2306250905131744e-05,
+      "loss": 0.7567,
+      "step": 16831
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2305421118312756e-05,
+      "loss": 1.1947,
+      "step": 16832
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2304591314728484e-05,
+      "loss": 0.9524,
+      "step": 16833
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2303761494384972e-05,
+      "loss": 1.2486,
+      "step": 16834
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.230293165728825e-05,
+      "loss": 1.0613,
+      "step": 16835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2302101803444355e-05,
+      "loss": 1.2445,
+      "step": 16836
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2301271932859317e-05,
+      "loss": 0.8185,
+      "step": 16837
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2300442045539176e-05,
+      "loss": 0.5314,
+      "step": 16838
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2299612141489963e-05,
+      "loss": 1.0045,
+      "step": 16839
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2298782220717718e-05,
+      "loss": 0.7353,
+      "step": 16840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2297952283228473e-05,
+      "loss": 0.9022,
+      "step": 16841
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.229712232902826e-05,
+      "loss": 1.0508,
+      "step": 16842
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2296292358123124e-05,
+      "loss": 0.9123,
+      "step": 16843
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2295462370519091e-05,
+      "loss": 0.6503,
+      "step": 16844
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2294632366222201e-05,
+      "loss": 0.8148,
+      "step": 16845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2293802345238493e-05,
+      "loss": 1.2754,
+      "step": 16846
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2292972307573996e-05,
+      "loss": 0.7102,
+      "step": 16847
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2292142253234749e-05,
+      "loss": 0.5976,
+      "step": 16848
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.229131218222679e-05,
+      "loss": 1.0174,
+      "step": 16849
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2290482094556156e-05,
+      "loss": 0.9463,
+      "step": 16850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2289651990228881e-05,
+      "loss": 0.8137,
+      "step": 16851
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2288821869251003e-05,
+      "loss": 0.6854,
+      "step": 16852
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2287991731628557e-05,
+      "loss": 0.693,
+      "step": 16853
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2287161577367584e-05,
+      "loss": 0.8784,
+      "step": 16854
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2286331406474113e-05,
+      "loss": 0.7379,
+      "step": 16855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.228550121895419e-05,
+      "loss": 0.6231,
+      "step": 16856
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2284671014813847e-05,
+      "loss": 0.8199,
+      "step": 16857
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2283840794059122e-05,
+      "loss": 0.9443,
+      "step": 16858
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2283010556696055e-05,
+      "loss": 0.5893,
+      "step": 16859
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2282180302730683e-05,
+      "loss": 0.8802,
+      "step": 16860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2281350032169036e-05,
+      "loss": 1.1484,
+      "step": 16861
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2280519745017165e-05,
+      "loss": 0.9851,
+      "step": 16862
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2279689441281098e-05,
+      "loss": 0.8769,
+      "step": 16863
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278859120966876e-05,
+      "loss": 0.8293,
+      "step": 16864
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2278028784080539e-05,
+      "loss": 0.7647,
+      "step": 16865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2277198430628121e-05,
+      "loss": 0.9713,
+      "step": 16866
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2276368060615667e-05,
+      "loss": 1.084,
+      "step": 16867
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227553767404921e-05,
+      "loss": 0.6825,
+      "step": 16868
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227470727093479e-05,
+      "loss": 1.0542,
+      "step": 16869
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2273876851278443e-05,
+      "loss": 0.7623,
+      "step": 16870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2273046415086219e-05,
+      "loss": 0.9992,
+      "step": 16871
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2272215962364142e-05,
+      "loss": 1.1588,
+      "step": 16872
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227138549311826e-05,
+      "loss": 1.3152,
+      "step": 16873
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227055500735461e-05,
+      "loss": 0.9154,
+      "step": 16874
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2269724505079234e-05,
+      "loss": 1.0882,
+      "step": 16875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2268893986298167e-05,
+      "loss": 0.819,
+      "step": 16876
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2268063451017454e-05,
+      "loss": 0.899,
+      "step": 16877
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2267232899243129e-05,
+      "loss": 0.8559,
+      "step": 16878
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2266402330981233e-05,
+      "loss": 1.1653,
+      "step": 16879
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226557174623781e-05,
+      "loss": 0.7818,
+      "step": 16880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2264741145018902e-05,
+      "loss": 0.5437,
+      "step": 16881
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263910527330538e-05,
+      "loss": 0.9195,
+      "step": 16882
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263079893178769e-05,
+      "loss": 0.831,
+      "step": 16883
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2262249242569631e-05,
+      "loss": 0.9387,
+      "step": 16884
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226141857550916e-05,
+      "loss": 1.0119,
+      "step": 16885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226058789200341e-05,
+      "loss": 1.2317,
+      "step": 16886
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2259757192058412e-05,
+      "loss": 0.6556,
+      "step": 16887
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2258926475680205e-05,
+      "loss": 0.8977,
+      "step": 16888
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2258095742874836e-05,
+      "loss": 0.7801,
+      "step": 16889
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2257264993648345e-05,
+      "loss": 1.2364,
+      "step": 16890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2256434228006771e-05,
+      "loss": 0.5733,
+      "step": 16891
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2255603445956158e-05,
+      "loss": 0.9596,
+      "step": 16892
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2254772647502546e-05,
+      "loss": 1.3593,
+      "step": 16893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2253941832651975e-05,
+      "loss": 0.5909,
+      "step": 16894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2253111001410489e-05,
+      "loss": 0.8604,
+      "step": 16895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2252280153784132e-05,
+      "loss": 0.8525,
+      "step": 16896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2251449289778943e-05,
+      "loss": 0.9309,
+      "step": 16897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2250618409400963e-05,
+      "loss": 0.9116,
+      "step": 16898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2249787512656238e-05,
+      "loss": 0.8202,
+      "step": 16899
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2248956599550805e-05,
+      "loss": 0.737,
+      "step": 16900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2248125670090713e-05,
+      "loss": 0.9865,
+      "step": 16901
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2247294724282002e-05,
+      "loss": 1.0051,
+      "step": 16902
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224646376213071e-05,
+      "loss": 0.8104,
+      "step": 16903
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224563278364289e-05,
+      "loss": 0.9122,
+      "step": 16904
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2244801788824577e-05,
+      "loss": 0.9487,
+      "step": 16905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2243970777681813e-05,
+      "loss": 0.7506,
+      "step": 16906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2243139750220647e-05,
+      "loss": 0.902,
+      "step": 16907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224230870644712e-05,
+      "loss": 0.5214,
+      "step": 16908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2241477646367273e-05,
+      "loss": 1.2499,
+      "step": 16909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2240646569987154e-05,
+      "loss": 0.7664,
+      "step": 16910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2239815477312804e-05,
+      "loss": 0.7478,
+      "step": 16911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2238984368350263e-05,
+      "loss": 0.8643,
+      "step": 16912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2238153243105583e-05,
+      "loss": 0.9268,
+      "step": 16913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2237322101584806e-05,
+      "loss": 1.095,
+      "step": 16914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2236490943793971e-05,
+      "loss": 1.0193,
+      "step": 16915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2235659769739126e-05,
+      "loss": 1.0312,
+      "step": 16916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2234828579426315e-05,
+      "loss": 1.2901,
+      "step": 16917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2233997372861581e-05,
+      "loss": 0.8333,
+      "step": 16918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2233166150050972e-05,
+      "loss": 0.7922,
+      "step": 16919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223233491100053e-05,
+      "loss": 0.7044,
+      "step": 16920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2231503655716302e-05,
+      "loss": 0.7484,
+      "step": 16921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223067238420433e-05,
+      "loss": 1.1231,
+      "step": 16922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222984109647066e-05,
+      "loss": 1.1919,
+      "step": 16923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222900979252134e-05,
+      "loss": 0.7944,
+      "step": 16924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2228178472362411e-05,
+      "loss": 0.7329,
+      "step": 16925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2227347135999926e-05,
+      "loss": 0.9713,
+      "step": 16926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.222651578343992e-05,
+      "loss": 0.9293,
+      "step": 16927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2225684414688444e-05,
+      "loss": 0.9046,
+      "step": 16928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2224853029751545e-05,
+      "loss": 0.6451,
+      "step": 16929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2224021628635267e-05,
+      "loss": 1.0478,
+      "step": 16930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2223190211345658e-05,
+      "loss": 1.0464,
+      "step": 16931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2222358777888764e-05,
+      "loss": 1.0657,
+      "step": 16932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2221527328270628e-05,
+      "loss": 0.901,
+      "step": 16933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2220695862497298e-05,
+      "loss": 0.8605,
+      "step": 16934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2219864380574822e-05,
+      "loss": 0.7562,
+      "step": 16935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2219032882509247e-05,
+      "loss": 0.6691,
+      "step": 16936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2218201368306618e-05,
+      "loss": 0.7798,
+      "step": 16937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221736983797298e-05,
+      "loss": 1.3737,
+      "step": 16938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2216538291514386e-05,
+      "loss": 0.968,
+      "step": 16939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2215706728936877e-05,
+      "loss": 1.02,
+      "step": 16940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2214875150246503e-05,
+      "loss": 1.3444,
+      "step": 16941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2214043555449314e-05,
+      "loss": 1.0582,
+      "step": 16942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221321194455135e-05,
+      "loss": 0.7582,
+      "step": 16943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2212380317558663e-05,
+      "loss": 0.8151,
+      "step": 16944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2211548674477302e-05,
+      "loss": 1.0094,
+      "step": 16945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2210717015313313e-05,
+      "loss": 0.7094,
+      "step": 16946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2209885340072746e-05,
+      "loss": 0.8686,
+      "step": 16947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2209053648761646e-05,
+      "loss": 1.0666,
+      "step": 16948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2208221941386063e-05,
+      "loss": 1.1225,
+      "step": 16949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2207390217952044e-05,
+      "loss": 0.8567,
+      "step": 16950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220655847846564e-05,
+      "loss": 0.7098,
+      "step": 16951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2205726722932895e-05,
+      "loss": 0.958,
+      "step": 16952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2204894951359862e-05,
+      "loss": 1.3083,
+      "step": 16953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2204063163752588e-05,
+      "loss": 0.9666,
+      "step": 16954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2203231360117123e-05,
+      "loss": 0.5673,
+      "step": 16955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2202399540459515e-05,
+      "loss": 0.7708,
+      "step": 16956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2201567704785814e-05,
+      "loss": 1.0586,
+      "step": 16957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2200735853102061e-05,
+      "loss": 1.3164,
+      "step": 16958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2199903985414319e-05,
+      "loss": 0.877,
+      "step": 16959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2199072101728628e-05,
+      "loss": 0.6594,
+      "step": 16960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2198240202051043e-05,
+      "loss": 1.1379,
+      "step": 16961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219740828638761e-05,
+      "loss": 0.9606,
+      "step": 16962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219657635474438e-05,
+      "loss": 1.1417,
+      "step": 16963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.21957444071274e-05,
+      "loss": 1.1623,
+      "step": 16964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2194912443542728e-05,
+      "loss": 0.986,
+      "step": 16965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219408046399641e-05,
+      "loss": 1.0992,
+      "step": 16966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2193248468494487e-05,
+      "loss": 0.9759,
+      "step": 16967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2192416457043024e-05,
+      "loss": 1.0345,
+      "step": 16968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2191584429648065e-05,
+      "loss": 0.8886,
+      "step": 16969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2190752386315656e-05,
+      "loss": 1.3233,
+      "step": 16970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2189920327051858e-05,
+      "loss": 1.0574,
+      "step": 16971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2189088251862713e-05,
+      "loss": 0.7564,
+      "step": 16972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2188256160754274e-05,
+      "loss": 0.8982,
+      "step": 16973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2187424053732597e-05,
+      "loss": 0.7399,
+      "step": 16974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2186591930803726e-05,
+      "loss": 0.857,
+      "step": 16975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2185759791973718e-05,
+      "loss": 1.3329,
+      "step": 16976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218492763724862e-05,
+      "loss": 0.8972,
+      "step": 16977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218409546663449e-05,
+      "loss": 1.1244,
+      "step": 16978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218326328013737e-05,
+      "loss": 0.7652,
+      "step": 16979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2182431077763317e-05,
+      "loss": 0.8986,
+      "step": 16980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2181598859518389e-05,
+      "loss": 0.9836,
+      "step": 16981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2180766625408624e-05,
+      "loss": 0.8551,
+      "step": 16982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2179934375440086e-05,
+      "loss": 0.975,
+      "step": 16983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2179102109618824e-05,
+      "loss": 0.7512,
+      "step": 16984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2178269827950888e-05,
+      "loss": 1.3459,
+      "step": 16985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2177437530442333e-05,
+      "loss": 0.6878,
+      "step": 16986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2176605217099208e-05,
+      "loss": 1.0803,
+      "step": 16987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2175772887927569e-05,
+      "loss": 1.08,
+      "step": 16988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.217494054293347e-05,
+      "loss": 0.9774,
+      "step": 16989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2174108182122959e-05,
+      "loss": 0.7804,
+      "step": 16990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2173275805502094e-05,
+      "loss": 0.9399,
+      "step": 16991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172443413076926e-05,
+      "loss": 0.7121,
+      "step": 16992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2171611004853508e-05,
+      "loss": 1.0559,
+      "step": 16993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2170778580837892e-05,
+      "loss": 0.8993,
+      "step": 16994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2169946141036133e-05,
+      "loss": 0.8027,
+      "step": 16995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216911368545429e-05,
+      "loss": 1.0864,
+      "step": 16996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2168281214098406e-05,
+      "loss": 1.0234,
+      "step": 16997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216744872697454e-05,
+      "loss": 0.9692,
+      "step": 16998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216661622408875e-05,
+      "loss": 0.6218,
+      "step": 16999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2165783705447084e-05,
+      "loss": 0.6152,
+      "step": 17000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2164951171055597e-05,
+      "loss": 1.0267,
+      "step": 17001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2164118620920347e-05,
+      "loss": 0.8219,
+      "step": 17002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2163286055047385e-05,
+      "loss": 0.8998,
+      "step": 17003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2162453473442766e-05,
+      "loss": 0.989,
+      "step": 17004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2161620876112546e-05,
+      "loss": 1.004,
+      "step": 17005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2160788263062781e-05,
+      "loss": 1.0064,
+      "step": 17006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.215995563429952e-05,
+      "loss": 0.8905,
+      "step": 17007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2159122989828825e-05,
+      "loss": 0.7429,
+      "step": 17008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2158290329656745e-05,
+      "loss": 0.9531,
+      "step": 17009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2157457653789337e-05,
+      "loss": 1.1107,
+      "step": 17010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2156624962232664e-05,
+      "loss": 0.9236,
+      "step": 17011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2155792254992772e-05,
+      "loss": 1.0179,
+      "step": 17012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2154959532075714e-05,
+      "loss": 1.1833,
+      "step": 17013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2154126793487555e-05,
+      "loss": 0.9363,
+      "step": 17014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2153294039234347e-05,
+      "loss": 0.7915,
+      "step": 17015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2152461269322145e-05,
+      "loss": 0.7611,
+      "step": 17016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2151628483757007e-05,
+      "loss": 0.9552,
+      "step": 17017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2150795682544987e-05,
+      "loss": 1.0118,
+      "step": 17018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214996286569214e-05,
+      "loss": 0.8082,
+      "step": 17019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2149130033204526e-05,
+      "loss": 0.8974,
+      "step": 17020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.21482971850882e-05,
+      "loss": 0.9118,
+      "step": 17021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2147464321349216e-05,
+      "loss": 1.1238,
+      "step": 17022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2146631441993636e-05,
+      "loss": 0.6914,
+      "step": 17023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2145798547027514e-05,
+      "loss": 0.7686,
+      "step": 17024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2144965636456903e-05,
+      "loss": 0.8879,
+      "step": 17025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2144132710287867e-05,
+      "loss": 0.8481,
+      "step": 17026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214329976852646e-05,
+      "loss": 0.9984,
+      "step": 17027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2142466811178737e-05,
+      "loss": 0.8761,
+      "step": 17028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2141633838250757e-05,
+      "loss": 0.8959,
+      "step": 17029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214080084974858e-05,
+      "loss": 0.7886,
+      "step": 17030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2139967845678262e-05,
+      "loss": 0.8478,
+      "step": 17031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.213913482604586e-05,
+      "loss": 1.2959,
+      "step": 17032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138301790857429e-05,
+      "loss": 1.0077,
+      "step": 17033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2137468740119032e-05,
+      "loss": 0.7797,
+      "step": 17034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2136635673836725e-05,
+      "loss": 0.9069,
+      "step": 17035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2135802592016567e-05,
+      "loss": 0.7351,
+      "step": 17036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2134969494664612e-05,
+      "loss": 0.9645,
+      "step": 17037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2134136381786924e-05,
+      "loss": 0.6021,
+      "step": 17038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2133303253389562e-05,
+      "loss": 0.9306,
+      "step": 17039
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2132470109478577e-05,
+      "loss": 0.8587,
+      "step": 17040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2131636950060037e-05,
+      "loss": 0.9352,
+      "step": 17041
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2130803775139995e-05,
+      "loss": 0.7165,
+      "step": 17042
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2129970584724506e-05,
+      "loss": 1.0259,
+      "step": 17043
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212913737881964e-05,
+      "loss": 0.8635,
+      "step": 17044
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2128304157431449e-05,
+      "loss": 0.9439,
+      "step": 17045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2127470920565995e-05,
+      "loss": 1.3131,
+      "step": 17046
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2126637668229335e-05,
+      "loss": 1.1807,
+      "step": 17047
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2125804400427532e-05,
+      "loss": 0.8547,
+      "step": 17048
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2124971117166641e-05,
+      "loss": 0.9441,
+      "step": 17049
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2124137818452723e-05,
+      "loss": 0.7358,
+      "step": 17050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2123304504291842e-05,
+      "loss": 1.0267,
+      "step": 17051
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2122471174690051e-05,
+      "loss": 0.7296,
+      "step": 17052
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2121637829653419e-05,
+      "loss": 0.8091,
+      "step": 17053
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2120804469187997e-05,
+      "loss": 0.8612,
+      "step": 17054
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2119971093299852e-05,
+      "loss": 1.2753,
+      "step": 17055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.211913770199504e-05,
+      "loss": 1.0293,
+      "step": 17056
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2118304295279624e-05,
+      "loss": 0.95,
+      "step": 17057
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117470873159665e-05,
+      "loss": 0.7944,
+      "step": 17058
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2116637435641218e-05,
+      "loss": 0.9538,
+      "step": 17059
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2115803982730351e-05,
+      "loss": 0.9897,
+      "step": 17060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2114970514433124e-05,
+      "loss": 0.9529,
+      "step": 17061
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2114137030755598e-05,
+      "loss": 0.8822,
+      "step": 17062
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2113303531703831e-05,
+      "loss": 0.9589,
+      "step": 17063
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2112470017283886e-05,
+      "loss": 0.8767,
+      "step": 17064
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2111636487501821e-05,
+      "loss": 1.0481,
+      "step": 17065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2110802942363708e-05,
+      "loss": 1.0406,
+      "step": 17066
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109969381875596e-05,
+      "loss": 0.8345,
+      "step": 17067
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109135806043556e-05,
+      "loss": 0.779,
+      "step": 17068
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2108302214873642e-05,
+      "loss": 1.0889,
+      "step": 17069
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2107468608371924e-05,
+      "loss": 0.9385,
+      "step": 17070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2106634986544459e-05,
+      "loss": 0.8533,
+      "step": 17071
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2105801349397311e-05,
+      "loss": 1.2126,
+      "step": 17072
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210496769693654e-05,
+      "loss": 1.0372,
+      "step": 17073
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210413402916821e-05,
+      "loss": 1.0814,
+      "step": 17074
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2103300346098384e-05,
+      "loss": 0.8366,
+      "step": 17075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2102466647733127e-05,
+      "loss": 1.0128,
+      "step": 17076
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2101632934078495e-05,
+      "loss": 0.8377,
+      "step": 17077
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2100799205140558e-05,
+      "loss": 1.084,
+      "step": 17078
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2099965460925376e-05,
+      "loss": 1.0734,
+      "step": 17079
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209913170143901e-05,
+      "loss": 1.2737,
+      "step": 17080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2098297926687528e-05,
+      "loss": 1.0266,
+      "step": 17081
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2097464136676988e-05,
+      "loss": 1.0559,
+      "step": 17082
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2096630331413454e-05,
+      "loss": 1.0672,
+      "step": 17083
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2095796510902995e-05,
+      "loss": 0.8071,
+      "step": 17084
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209496267515167e-05,
+      "loss": 0.8247,
+      "step": 17085
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2094128824165543e-05,
+      "loss": 0.8434,
+      "step": 17086
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209329495795068e-05,
+      "loss": 0.9133,
+      "step": 17087
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2092461076513145e-05,
+      "loss": 0.7002,
+      "step": 17088
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2091627179858996e-05,
+      "loss": 0.7869,
+      "step": 17089
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2090793267994303e-05,
+      "loss": 0.718,
+      "step": 17090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2089959340925134e-05,
+      "loss": 0.887,
+      "step": 17091
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2089125398657545e-05,
+      "loss": 0.868,
+      "step": 17092
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2088291441197603e-05,
+      "loss": 1.3975,
+      "step": 17093
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2087457468551375e-05,
+      "loss": 0.9312,
+      "step": 17094
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2086623480724923e-05,
+      "loss": 1.0611,
+      "step": 17095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2085789477724318e-05,
+      "loss": 0.946,
+      "step": 17096
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2084955459555617e-05,
+      "loss": 1.0372,
+      "step": 17097
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2084121426224886e-05,
+      "loss": 1.4462,
+      "step": 17098
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2083287377738195e-05,
+      "loss": 1.1587,
+      "step": 17099
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2082453314101607e-05,
+      "loss": 0.8369,
+      "step": 17100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2081619235321189e-05,
+      "loss": 1.1742,
+      "step": 17101
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2080785141403001e-05,
+      "loss": 0.9177,
+      "step": 17102
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2079951032353117e-05,
+      "loss": 0.6418,
+      "step": 17103
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2079116908177592e-05,
+      "loss": 0.8174,
+      "step": 17104
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2078282768882503e-05,
+      "loss": 0.91,
+      "step": 17105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2077448614473913e-05,
+      "loss": 0.7447,
+      "step": 17106
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.207661444495788e-05,
+      "loss": 0.707,
+      "step": 17107
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2075780260340478e-05,
+      "loss": 1.0411,
+      "step": 17108
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074946060627772e-05,
+      "loss": 0.7998,
+      "step": 17109
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074111845825829e-05,
+      "loss": 0.9568,
+      "step": 17110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2073277615940713e-05,
+      "loss": 0.7445,
+      "step": 17111
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2072443370978494e-05,
+      "loss": 0.8097,
+      "step": 17112
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.207160911094523e-05,
+      "loss": 1.0081,
+      "step": 17113
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070774835847002e-05,
+      "loss": 0.9363,
+      "step": 17114
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2069940545689867e-05,
+      "loss": 0.8615,
+      "step": 17115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2069106240479893e-05,
+      "loss": 0.913,
+      "step": 17116
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2068271920223152e-05,
+      "loss": 1.1556,
+      "step": 17117
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2067437584925705e-05,
+      "loss": 1.0112,
+      "step": 17118
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2066603234593624e-05,
+      "loss": 0.8746,
+      "step": 17119
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206576886923297e-05,
+      "loss": 1.1087,
+      "step": 17120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2064934488849822e-05,
+      "loss": 1.2615,
+      "step": 17121
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2064100093450237e-05,
+      "loss": 0.9688,
+      "step": 17122
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2063265683040289e-05,
+      "loss": 0.9654,
+      "step": 17123
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062431257626041e-05,
+      "loss": 0.5494,
+      "step": 17124
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2061596817213568e-05,
+      "loss": 0.9257,
+      "step": 17125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2060762361808932e-05,
+      "loss": 0.9091,
+      "step": 17126
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2059927891418204e-05,
+      "loss": 0.9418,
+      "step": 17127
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2059093406047451e-05,
+      "loss": 0.8336,
+      "step": 17128
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205825890570274e-05,
+      "loss": 0.8171,
+      "step": 17129
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2057424390390141e-05,
+      "loss": 1.0184,
+      "step": 17130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2056589860115728e-05,
+      "loss": 0.7757,
+      "step": 17131
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2055755314885565e-05,
+      "loss": 0.8938,
+      "step": 17132
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2054920754705717e-05,
+      "loss": 0.8061,
+      "step": 17133
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2054086179582261e-05,
+      "loss": 0.8097,
+      "step": 17134
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2053251589521258e-05,
+      "loss": 1.2042,
+      "step": 17135
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2052416984528787e-05,
+      "loss": 0.7735,
+      "step": 17136
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2051582364610908e-05,
+      "loss": 0.8417,
+      "step": 17137
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2050747729773694e-05,
+      "loss": 0.6777,
+      "step": 17138
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2049913080023214e-05,
+      "loss": 0.8237,
+      "step": 17139
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2049078415365543e-05,
+      "loss": 0.8376,
+      "step": 17140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2048243735806742e-05,
+      "loss": 0.8051,
+      "step": 17141
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2047409041352886e-05,
+      "loss": 0.8856,
+      "step": 17142
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2046574332010045e-05,
+      "loss": 0.5368,
+      "step": 17143
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2045739607784286e-05,
+      "loss": 0.757,
+      "step": 17144
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2044904868681684e-05,
+      "loss": 1.0939,
+      "step": 17145
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2044070114708309e-05,
+      "loss": 0.9176,
+      "step": 17146
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2043235345870223e-05,
+      "loss": 1.0977,
+      "step": 17147
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2042400562173506e-05,
+      "loss": 0.8344,
+      "step": 17148
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2041565763624225e-05,
+      "loss": 1.1281,
+      "step": 17149
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.204073095022845e-05,
+      "loss": 0.9067,
+      "step": 17150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2039896121992255e-05,
+      "loss": 0.9811,
+      "step": 17151
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2039061278921708e-05,
+      "loss": 1.0713,
+      "step": 17152
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2038226421022879e-05,
+      "loss": 0.8462,
+      "step": 17153
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2037391548301845e-05,
+      "loss": 0.9501,
+      "step": 17154
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036556660764668e-05,
+      "loss": 1.1188,
+      "step": 17155
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2035721758417429e-05,
+      "loss": 1.1721,
+      "step": 17156
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2034886841266195e-05,
+      "loss": 0.8345,
+      "step": 17157
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203405190931704e-05,
+      "loss": 0.7977,
+      "step": 17158
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2033216962576026e-05,
+      "loss": 0.7906,
+      "step": 17159
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203238200104924e-05,
+      "loss": 1.2217,
+      "step": 17160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2031547024742744e-05,
+      "loss": 0.7621,
+      "step": 17161
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2030712033662613e-05,
+      "loss": 0.7683,
+      "step": 17162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2029877027814916e-05,
+      "loss": 0.8801,
+      "step": 17163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202904200720573e-05,
+      "loss": 0.9622,
+      "step": 17164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2028206971841123e-05,
+      "loss": 1.0906,
+      "step": 17165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2027371921727174e-05,
+      "loss": 0.9862,
+      "step": 17166
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202653685686995e-05,
+      "loss": 1.1594,
+      "step": 17167
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202570177727552e-05,
+      "loss": 1.0793,
+      "step": 17168
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2024866682949968e-05,
+      "loss": 0.6889,
+      "step": 17169
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2024031573899357e-05,
+      "loss": 1.0471,
+      "step": 17170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2023196450129765e-05,
+      "loss": 0.8122,
+      "step": 17171
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2022361311647264e-05,
+      "loss": 0.6925,
+      "step": 17172
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2021526158457926e-05,
+      "loss": 0.654,
+      "step": 17173
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2020690990567824e-05,
+      "loss": 0.7488,
+      "step": 17174
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019855807983036e-05,
+      "loss": 0.9753,
+      "step": 17175
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019020610709633e-05,
+      "loss": 1.3197,
+      "step": 17176
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2018185398753684e-05,
+      "loss": 0.7197,
+      "step": 17177
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2017350172121267e-05,
+      "loss": 0.9247,
+      "step": 17178
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2016514930818457e-05,
+      "loss": 0.9031,
+      "step": 17179
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2015679674851329e-05,
+      "loss": 0.8501,
+      "step": 17180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2014844404225951e-05,
+      "loss": 0.7856,
+      "step": 17181
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2014009118948402e-05,
+      "loss": 1.0345,
+      "step": 17182
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2013173819024752e-05,
+      "loss": 0.9305,
+      "step": 17183
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201233850446108e-05,
+      "loss": 0.8534,
+      "step": 17184
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2011503175263461e-05,
+      "loss": 1.1051,
+      "step": 17185
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2010667831437968e-05,
+      "loss": 0.9668,
+      "step": 17186
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2009832472990674e-05,
+      "loss": 1.0993,
+      "step": 17187
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008997099927655e-05,
+      "loss": 1.0321,
+      "step": 17188
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008161712254987e-05,
+      "loss": 1.0383,
+      "step": 17189
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200732630997874e-05,
+      "loss": 0.9175,
+      "step": 17190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2006490893104999e-05,
+      "loss": 1.0516,
+      "step": 17191
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005655461639827e-05,
+      "loss": 1.0107,
+      "step": 17192
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2004820015589311e-05,
+      "loss": 0.7715,
+      "step": 17193
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003984554959521e-05,
+      "loss": 0.7055,
+      "step": 17194
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2003149079756528e-05,
+      "loss": 1.016,
+      "step": 17195
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002313589986416e-05,
+      "loss": 0.7066,
+      "step": 17196
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001478085655258e-05,
+      "loss": 1.0195,
+      "step": 17197
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200064256676913e-05,
+      "loss": 0.9897,
+      "step": 17198
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999807033334103e-05,
+      "loss": 0.8506,
+      "step": 17199
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998971485356259e-05,
+      "loss": 1.035,
+      "step": 17200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998135922841671e-05,
+      "loss": 0.8772,
+      "step": 17201
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997300345796421e-05,
+      "loss": 0.5417,
+      "step": 17202
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1996464754226577e-05,
+      "loss": 0.8816,
+      "step": 17203
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199562914813822e-05,
+      "loss": 0.5716,
+      "step": 17204
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1994793527537427e-05,
+      "loss": 0.7694,
+      "step": 17205
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1993957892430277e-05,
+      "loss": 0.8515,
+      "step": 17206
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199312224282284e-05,
+      "loss": 1.1293,
+      "step": 17207
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1992286578721197e-05,
+      "loss": 1.0266,
+      "step": 17208
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991450900131425e-05,
+      "loss": 1.1379,
+      "step": 17209
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990615207059602e-05,
+      "loss": 0.8867,
+      "step": 17210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989779499511805e-05,
+      "loss": 0.9422,
+      "step": 17211
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198894377749411e-05,
+      "loss": 1.206,
+      "step": 17212
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1988108041012593e-05,
+      "loss": 1.1236,
+      "step": 17213
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987272290073332e-05,
+      "loss": 0.9847,
+      "step": 17214
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198643652468241e-05,
+      "loss": 1.097,
+      "step": 17215
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1985600744845902e-05,
+      "loss": 0.8328,
+      "step": 17216
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198476495056988e-05,
+      "loss": 0.8975,
+      "step": 17217
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.198392914186043e-05,
+      "loss": 0.7133,
+      "step": 17218
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983093318723627e-05,
+      "loss": 1.0374,
+      "step": 17219
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1982257481165547e-05,
+      "loss": 0.7992,
+      "step": 17220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1981421629192273e-05,
+      "loss": 0.9215,
+      "step": 17221
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980585762809882e-05,
+      "loss": 0.7318,
+      "step": 17222
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979749882024448e-05,
+      "loss": 1.0678,
+      "step": 17223
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978913986842053e-05,
+      "loss": 1.0192,
+      "step": 17224
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197807807726878e-05,
+      "loss": 1.0046,
+      "step": 17225
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.19772421533107e-05,
+      "loss": 0.6,
+      "step": 17226
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976406214973897e-05,
+      "loss": 0.9794,
+      "step": 17227
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975570262264447e-05,
+      "loss": 0.803,
+      "step": 17228
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1974734295188431e-05,
+      "loss": 0.803,
+      "step": 17229
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973898313751929e-05,
+      "loss": 1.1296,
+      "step": 17230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197306231796102e-05,
+      "loss": 0.9595,
+      "step": 17231
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.197222630782178e-05,
+      "loss": 0.8612,
+      "step": 17232
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971390283340291e-05,
+      "loss": 0.8963,
+      "step": 17233
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970554244522636e-05,
+      "loss": 1.0668,
+      "step": 17234
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1969718191374888e-05,
+      "loss": 1.0169,
+      "step": 17235
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968882123903137e-05,
+      "loss": 0.8166,
+      "step": 17236
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968046042113452e-05,
+      "loss": 0.9887,
+      "step": 17237
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1967209946011916e-05,
+      "loss": 0.9803,
+      "step": 17238
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966373835604612e-05,
+      "loss": 0.8753,
+      "step": 17239
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1965537710897621e-05,
+      "loss": 0.866,
+      "step": 17240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1964701571897019e-05,
+      "loss": 0.5741,
+      "step": 17241
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963865418608893e-05,
+      "loss": 0.7398,
+      "step": 17242
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1963029251039318e-05,
+      "loss": 0.5114,
+      "step": 17243
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962193069194372e-05,
+      "loss": 1.2785,
+      "step": 17244
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1961356873080144e-05,
+      "loss": 0.8941,
+      "step": 17245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1960520662702713e-05,
+      "loss": 0.8432,
+      "step": 17246
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1959684438068154e-05,
+      "loss": 0.6825,
+      "step": 17247
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958848199182553e-05,
+      "loss": 0.9344,
+      "step": 17248
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958011946051991e-05,
+      "loss": 1.1448,
+      "step": 17249
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1957175678682548e-05,
+      "loss": 1.0993,
+      "step": 17250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.195633939708031e-05,
+      "loss": 1.0499,
+      "step": 17251
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1955503101251351e-05,
+      "loss": 0.6708,
+      "step": 17252
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1954666791201754e-05,
+      "loss": 1.3161,
+      "step": 17253
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1953830466937606e-05,
+      "loss": 0.7726,
+      "step": 17254
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1952994128464987e-05,
+      "loss": 0.9105,
+      "step": 17255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1952157775789976e-05,
+      "loss": 0.9042,
+      "step": 17256
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.195132140891866e-05,
+      "loss": 0.9897,
+      "step": 17257
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1950485027857112e-05,
+      "loss": 0.8204,
+      "step": 17258
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1949648632611427e-05,
+      "loss": 0.8292,
+      "step": 17259
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1948812223187675e-05,
+      "loss": 0.7727,
+      "step": 17260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194797579959195e-05,
+      "loss": 1.1215,
+      "step": 17261
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1947139361830325e-05,
+      "loss": 1.2018,
+      "step": 17262
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1946302909908884e-05,
+      "loss": 0.978,
+      "step": 17263
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1945466443833716e-05,
+      "loss": 0.9804,
+      "step": 17264
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1944629963610897e-05,
+      "loss": 0.9715,
+      "step": 17265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1943793469246514e-05,
+      "loss": 0.8404,
+      "step": 17266
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194295696074665e-05,
+      "loss": 0.9352,
+      "step": 17267
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1942120438117389e-05,
+      "loss": 0.8622,
+      "step": 17268
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1941283901364806e-05,
+      "loss": 0.8198,
+      "step": 17269
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1940447350494999e-05,
+      "loss": 0.9892,
+      "step": 17270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.193961078551404e-05,
+      "loss": 1.3544,
+      "step": 17271
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1938774206428011e-05,
+      "loss": 1.0718,
+      "step": 17272
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937937613243007e-05,
+      "loss": 0.687,
+      "step": 17273
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937101005965103e-05,
+      "loss": 0.7929,
+      "step": 17274
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1936264384600386e-05,
+      "loss": 1.1249,
+      "step": 17275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1935427749154943e-05,
+      "loss": 1.0584,
+      "step": 17276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1934591099634849e-05,
+      "loss": 1.5349,
+      "step": 17277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1933754436046195e-05,
+      "loss": 0.6204,
+      "step": 17278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932917758395065e-05,
+      "loss": 0.7443,
+      "step": 17279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1932081066687544e-05,
+      "loss": 0.6372,
+      "step": 17280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.193124436092971e-05,
+      "loss": 0.8873,
+      "step": 17281
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1930407641127657e-05,
+      "loss": 0.6767,
+      "step": 17282
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929570907287464e-05,
+      "loss": 1.2774,
+      "step": 17283
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1928734159415214e-05,
+      "loss": 0.7331,
+      "step": 17284
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927897397517e-05,
+      "loss": 0.9562,
+      "step": 17285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.19270606215989e-05,
+      "loss": 0.8182,
+      "step": 17286
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1926223831667001e-05,
+      "loss": 1.0498,
+      "step": 17287
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925387027727389e-05,
+      "loss": 0.9268,
+      "step": 17288
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924550209786147e-05,
+      "loss": 0.7287,
+      "step": 17289
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923713377849364e-05,
+      "loss": 0.9058,
+      "step": 17290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922876531923125e-05,
+      "loss": 0.8124,
+      "step": 17291
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1922039672013513e-05,
+      "loss": 1.1041,
+      "step": 17292
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1921202798126612e-05,
+      "loss": 1.2816,
+      "step": 17293
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1920365910268512e-05,
+      "loss": 1.1235,
+      "step": 17294
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919529008445302e-05,
+      "loss": 0.9854,
+      "step": 17295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191869209266306e-05,
+      "loss": 0.7526,
+      "step": 17296
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917855162927877e-05,
+      "loss": 0.9507,
+      "step": 17297
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191701821924584e-05,
+      "loss": 1.0987,
+      "step": 17298
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191618126162303e-05,
+      "loss": 0.7059,
+      "step": 17299
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191534429006554e-05,
+      "loss": 1.0509,
+      "step": 17300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1914507304579453e-05,
+      "loss": 1.0316,
+      "step": 17301
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1913670305170855e-05,
+      "loss": 0.5031,
+      "step": 17302
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1912833291845836e-05,
+      "loss": 0.9524,
+      "step": 17303
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191199626461048e-05,
+      "loss": 0.8468,
+      "step": 17304
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1911159223470874e-05,
+      "loss": 0.6704,
+      "step": 17305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1910322168433108e-05,
+      "loss": 0.5166,
+      "step": 17306
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1909485099503267e-05,
+      "loss": 0.8673,
+      "step": 17307
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1908648016687434e-05,
+      "loss": 0.7387,
+      "step": 17308
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1907810919991702e-05,
+      "loss": 0.9669,
+      "step": 17309
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1906973809422163e-05,
+      "loss": 0.7661,
+      "step": 17310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1906136684984892e-05,
+      "loss": 0.7418,
+      "step": 17311
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1905299546685987e-05,
+      "loss": 0.7385,
+      "step": 17312
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.190446239453153e-05,
+      "loss": 1.1289,
+      "step": 17313
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1903625228527613e-05,
+      "loss": 0.7922,
+      "step": 17314
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1902788048680321e-05,
+      "loss": 0.8007,
+      "step": 17315
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1901950854995744e-05,
+      "loss": 1.2401,
+      "step": 17316
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1901113647479965e-05,
+      "loss": 1.1532,
+      "step": 17317
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1900276426139082e-05,
+      "loss": 0.7666,
+      "step": 17318
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1899439190979173e-05,
+      "loss": 0.6908,
+      "step": 17319
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1898601942006336e-05,
+      "loss": 0.7992,
+      "step": 17320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1897764679226653e-05,
+      "loss": 0.9438,
+      "step": 17321
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1896927402646212e-05,
+      "loss": 0.8227,
+      "step": 17322
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1896090112271107e-05,
+      "loss": 0.8661,
+      "step": 17323
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1895252808107423e-05,
+      "loss": 0.8108,
+      "step": 17324
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1894415490161253e-05,
+      "loss": 0.7643,
+      "step": 17325
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1893578158438679e-05,
+      "loss": 0.9038,
+      "step": 17326
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1892740812945795e-05,
+      "loss": 1.0263,
+      "step": 17327
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.189190345368869e-05,
+      "loss": 1.2506,
+      "step": 17328
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1891066080673457e-05,
+      "loss": 0.6734,
+      "step": 17329
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1890228693906179e-05,
+      "loss": 1.2372,
+      "step": 17330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1889391293392945e-05,
+      "loss": 0.9626,
+      "step": 17331
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188855387913985e-05,
+      "loss": 0.9657,
+      "step": 17332
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188771645115298e-05,
+      "loss": 1.0384,
+      "step": 17333
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1886879009438428e-05,
+      "loss": 0.871,
+      "step": 17334
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1886041554002284e-05,
+      "loss": 0.6689,
+      "step": 17335
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1885204084850633e-05,
+      "loss": 1.0163,
+      "step": 17336
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.188436660198957e-05,
+      "loss": 1.1119,
+      "step": 17337
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1883529105425185e-05,
+      "loss": 0.7394,
+      "step": 17338
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1882691595163564e-05,
+      "loss": 0.9543,
+      "step": 17339
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1881854071210805e-05,
+      "loss": 0.7163,
+      "step": 17340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1881016533572992e-05,
+      "loss": 1.2291,
+      "step": 17341
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1880178982256216e-05,
+      "loss": 0.8357,
+      "step": 17342
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1879341417266572e-05,
+      "loss": 0.7014,
+      "step": 17343
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1878503838610146e-05,
+      "loss": 1.3635,
+      "step": 17344
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1877666246293036e-05,
+      "loss": 0.7226,
+      "step": 17345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1876828640321325e-05,
+      "loss": 0.7817,
+      "step": 17346
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1875991020701109e-05,
+      "loss": 0.9673,
+      "step": 17347
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1875153387438477e-05,
+      "loss": 1.3753,
+      "step": 17348
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1874315740539522e-05,
+      "loss": 1.228,
+      "step": 17349
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1873478080010336e-05,
+      "loss": 0.9336,
+      "step": 17350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1872640405857007e-05,
+      "loss": 1.0481,
+      "step": 17351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187180271808563e-05,
+      "loss": 0.9437,
+      "step": 17352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1870965016702297e-05,
+      "loss": 1.0655,
+      "step": 17353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1870127301713094e-05,
+      "loss": 0.8732,
+      "step": 17354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1869289573124125e-05,
+      "loss": 0.7368,
+      "step": 17355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186845183094147e-05,
+      "loss": 1.0363,
+      "step": 17356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1867614075171226e-05,
+      "loss": 1.1698,
+      "step": 17357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1866776305819484e-05,
+      "loss": 0.8378,
+      "step": 17358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1865938522892342e-05,
+      "loss": 1.144,
+      "step": 17359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1865100726395883e-05,
+      "loss": 0.6443,
+      "step": 17360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1864262916336206e-05,
+      "loss": 1.0385,
+      "step": 17361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1863425092719405e-05,
+      "loss": 1.1384,
+      "step": 17362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1862587255551565e-05,
+      "loss": 0.932,
+      "step": 17363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1861749404838785e-05,
+      "loss": 1.0397,
+      "step": 17364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186091154058716e-05,
+      "loss": 0.9651,
+      "step": 17365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1860073662802776e-05,
+      "loss": 1.0301,
+      "step": 17366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.185923577149173e-05,
+      "loss": 1.4784,
+      "step": 17367
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1858397866660117e-05,
+      "loss": 0.865,
+      "step": 17368
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1857559948314023e-05,
+      "loss": 1.1289,
+      "step": 17369
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1856722016459554e-05,
+      "loss": 0.9499,
+      "step": 17370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1855884071102794e-05,
+      "loss": 1.1682,
+      "step": 17371
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1855046112249838e-05,
+      "loss": 0.9631,
+      "step": 17372
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.185420813990678e-05,
+      "loss": 0.7417,
+      "step": 17373
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1853370154079715e-05,
+      "loss": 1.1651,
+      "step": 17374
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1852532154774737e-05,
+      "loss": 0.932,
+      "step": 17375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.185169414199794e-05,
+      "loss": 1.1097,
+      "step": 17376
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1850856115755416e-05,
+      "loss": 0.6678,
+      "step": 17377
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.185001807605326e-05,
+      "loss": 0.8213,
+      "step": 17378
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184918002289757e-05,
+      "loss": 0.9125,
+      "step": 17379
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1848341956294438e-05,
+      "loss": 0.6212,
+      "step": 17380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1847503876249956e-05,
+      "loss": 0.6306,
+      "step": 17381
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184666578277022e-05,
+      "loss": 1.2669,
+      "step": 17382
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1845827675861326e-05,
+      "loss": 1.1692,
+      "step": 17383
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184498955552937e-05,
+      "loss": 0.9918,
+      "step": 17384
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1844151421780442e-05,
+      "loss": 1.0131,
+      "step": 17385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1843313274620642e-05,
+      "loss": 0.6368,
+      "step": 17386
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184247511405606e-05,
+      "loss": 1.0962,
+      "step": 17387
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1841636940092798e-05,
+      "loss": 0.9679,
+      "step": 17388
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1840798752736948e-05,
+      "loss": 0.9743,
+      "step": 17389
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1839960551994603e-05,
+      "loss": 0.9291,
+      "step": 17390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1839122337871863e-05,
+      "loss": 0.8836,
+      "step": 17391
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1838284110374818e-05,
+      "loss": 0.8479,
+      "step": 17392
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1837445869509569e-05,
+      "loss": 0.8914,
+      "step": 17393
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1836607615282207e-05,
+      "loss": 0.9661,
+      "step": 17394
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1835769347698835e-05,
+      "loss": 1.091,
+      "step": 17395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183493106676554e-05,
+      "loss": 0.8035,
+      "step": 17396
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1834092772488424e-05,
+      "loss": 0.9654,
+      "step": 17397
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1833254464873582e-05,
+      "loss": 0.9328,
+      "step": 17398
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1832416143927113e-05,
+      "loss": 1.2823,
+      "step": 17399
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1831577809655105e-05,
+      "loss": 1.2313,
+      "step": 17400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1830739462063662e-05,
+      "loss": 0.9124,
+      "step": 17401
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1829901101158881e-05,
+      "loss": 1.0621,
+      "step": 17402
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.182906272694685e-05,
+      "loss": 1.3329,
+      "step": 17403
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1828224339433677e-05,
+      "loss": 0.9924,
+      "step": 17404
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1827385938625451e-05,
+      "loss": 1.1933,
+      "step": 17405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1826547524528271e-05,
+      "loss": 1.235,
+      "step": 17406
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1825709097148237e-05,
+      "loss": 1.2025,
+      "step": 17407
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1824870656491443e-05,
+      "loss": 0.7183,
+      "step": 17408
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1824032202563984e-05,
+      "loss": 0.823,
+      "step": 17409
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1823193735371965e-05,
+      "loss": 0.9713,
+      "step": 17410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1822355254921478e-05,
+      "loss": 0.7769,
+      "step": 17411
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1821516761218617e-05,
+      "loss": 0.7217,
+      "step": 17412
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1820678254269487e-05,
+      "loss": 1.2779,
+      "step": 17413
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1819839734080184e-05,
+      "loss": 0.4824,
+      "step": 17414
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.18190012006568e-05,
+      "loss": 1.1905,
+      "step": 17415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181816265400544e-05,
+      "loss": 1.0739,
+      "step": 17416
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1817324094132201e-05,
+      "loss": 1.0049,
+      "step": 17417
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1816485521043175e-05,
+      "loss": 0.9471,
+      "step": 17418
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1815646934744468e-05,
+      "loss": 0.8099,
+      "step": 17419
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1814808335242174e-05,
+      "loss": 1.0442,
+      "step": 17420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1813969722542392e-05,
+      "loss": 1.1532,
+      "step": 17421
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181313109665122e-05,
+      "loss": 0.8981,
+      "step": 17422
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1812292457574759e-05,
+      "loss": 0.9203,
+      "step": 17423
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1811453805319102e-05,
+      "loss": 1.1206,
+      "step": 17424
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1810615139890359e-05,
+      "loss": 0.821,
+      "step": 17425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1809776461294617e-05,
+      "loss": 1.1013,
+      "step": 17426
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1808937769537978e-05,
+      "loss": 1.1914,
+      "step": 17427
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1808099064626544e-05,
+      "loss": 1.0798,
+      "step": 17428
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1807260346566414e-05,
+      "loss": 1.0075,
+      "step": 17429
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1806421615363685e-05,
+      "loss": 1.0359,
+      "step": 17430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180558287102446e-05,
+      "loss": 0.9784,
+      "step": 17431
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804744113554834e-05,
+      "loss": 1.3406,
+      "step": 17432
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1803905342960907e-05,
+      "loss": 0.8891,
+      "step": 17433
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1803066559248781e-05,
+      "loss": 0.9728,
+      "step": 17434
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1802227762424558e-05,
+      "loss": 0.9643,
+      "step": 17435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1801388952494328e-05,
+      "loss": 0.8832,
+      "step": 17436
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1800550129464204e-05,
+      "loss": 1.1309,
+      "step": 17437
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1799711293340276e-05,
+      "loss": 0.9254,
+      "step": 17438
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1798872444128646e-05,
+      "loss": 1.0416,
+      "step": 17439
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179803358183542e-05,
+      "loss": 0.8352,
+      "step": 17440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1797194706466693e-05,
+      "loss": 0.8084,
+      "step": 17441
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1796355818028564e-05,
+      "loss": 1.1601,
+      "step": 17442
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1795516916527138e-05,
+      "loss": 0.7567,
+      "step": 17443
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1794678001968514e-05,
+      "loss": 1.0429,
+      "step": 17444
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179383907435879e-05,
+      "loss": 0.6501,
+      "step": 17445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793000133704073e-05,
+      "loss": 1.0441,
+      "step": 17446
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1792161180010458e-05,
+      "loss": 0.5885,
+      "step": 17447
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791322213284045e-05,
+      "loss": 0.7592,
+      "step": 17448
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1790483233530941e-05,
+      "loss": 0.8774,
+      "step": 17449
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1789644240757245e-05,
+      "loss": 1.1104,
+      "step": 17450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1788805234969054e-05,
+      "loss": 0.5823,
+      "step": 17451
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1787966216172474e-05,
+      "loss": 0.7993,
+      "step": 17452
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1787127184373607e-05,
+      "loss": 0.8164,
+      "step": 17453
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1786288139578553e-05,
+      "loss": 1.27,
+      "step": 17454
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785449081793413e-05,
+      "loss": 0.9519,
+      "step": 17455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1784610011024286e-05,
+      "loss": 0.7603,
+      "step": 17456
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1783770927277276e-05,
+      "loss": 1.0072,
+      "step": 17457
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1782931830558489e-05,
+      "loss": 0.8902,
+      "step": 17458
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1782092720874024e-05,
+      "loss": 0.9384,
+      "step": 17459
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1781253598229982e-05,
+      "loss": 0.7777,
+      "step": 17460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1780414462632465e-05,
+      "loss": 1.0374,
+      "step": 17461
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1779575314087577e-05,
+      "loss": 0.6934,
+      "step": 17462
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177873615260142e-05,
+      "loss": 0.8754,
+      "step": 17463
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1777896978180094e-05,
+      "loss": 0.7904,
+      "step": 17464
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177705779082971e-05,
+      "loss": 1.1473,
+      "step": 17465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1776218590556358e-05,
+      "loss": 0.7514,
+      "step": 17466
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177537937736615e-05,
+      "loss": 1.0225,
+      "step": 17467
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1774540151265185e-05,
+      "loss": 0.9512,
+      "step": 17468
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1773700912259567e-05,
+      "loss": 1.0064,
+      "step": 17469
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1772861660355399e-05,
+      "loss": 1.3548,
+      "step": 17470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1772022395558786e-05,
+      "loss": 0.7357,
+      "step": 17471
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177118311787583e-05,
+      "loss": 0.7924,
+      "step": 17472
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1770343827312627e-05,
+      "loss": 0.9077,
+      "step": 17473
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1769504523875293e-05,
+      "loss": 0.6876,
+      "step": 17474
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1768665207569922e-05,
+      "loss": 1.0295,
+      "step": 17475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1767825878402626e-05,
+      "loss": 0.7789,
+      "step": 17476
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1766986536379502e-05,
+      "loss": 0.7377,
+      "step": 17477
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1766147181506656e-05,
+      "loss": 1.0828,
+      "step": 17478
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1765307813790187e-05,
+      "loss": 0.9638,
+      "step": 17479
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1764468433236209e-05,
+      "loss": 1.0047,
+      "step": 17480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176362903985082e-05,
+      "loss": 1.2817,
+      "step": 17481
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1762789633640122e-05,
+      "loss": 0.768,
+      "step": 17482
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1761950214610222e-05,
+      "loss": 0.9715,
+      "step": 17483
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1761110782767227e-05,
+      "loss": 1.1538,
+      "step": 17484
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1760271338117237e-05,
+      "loss": 1.0214,
+      "step": 17485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759431880666362e-05,
+      "loss": 0.7934,
+      "step": 17486
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17585924104207e-05,
+      "loss": 0.9561,
+      "step": 17487
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1757752927386355e-05,
+      "loss": 0.8061,
+      "step": 17488
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175691343156944e-05,
+      "loss": 0.6802,
+      "step": 17489
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1756073922976056e-05,
+      "loss": 1.0327,
+      "step": 17490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1755234401612305e-05,
+      "loss": 0.8925,
+      "step": 17491
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1754394867484296e-05,
+      "loss": 1.0559,
+      "step": 17492
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1753555320598134e-05,
+      "loss": 1.0691,
+      "step": 17493
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1752715760959918e-05,
+      "loss": 1.1846,
+      "step": 17494
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1751876188575764e-05,
+      "loss": 1.1414,
+      "step": 17495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175103660345177e-05,
+      "loss": 0.7532,
+      "step": 17496
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1750197005594041e-05,
+      "loss": 1.0196,
+      "step": 17497
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1749357395008687e-05,
+      "loss": 1.2258,
+      "step": 17498
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1748517771701812e-05,
+      "loss": 1.0954,
+      "step": 17499
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174767813567952e-05,
+      "loss": 0.8974,
+      "step": 17500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1746838486947923e-05,
+      "loss": 1.2597,
+      "step": 17501
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1745998825513118e-05,
+      "loss": 0.6371,
+      "step": 17502
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1745159151381215e-05,
+      "loss": 0.9862,
+      "step": 17503
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1744319464558322e-05,
+      "loss": 0.9743,
+      "step": 17504
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1743479765050549e-05,
+      "loss": 1.1076,
+      "step": 17505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1742640052863991e-05,
+      "loss": 1.2306,
+      "step": 17506
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1741800328004764e-05,
+      "loss": 1.0632,
+      "step": 17507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1740960590478972e-05,
+      "loss": 0.9418,
+      "step": 17508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174012084029272e-05,
+      "loss": 0.6372,
+      "step": 17509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1739281077452116e-05,
+      "loss": 0.9752,
+      "step": 17510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1738441301963268e-05,
+      "loss": 0.9713,
+      "step": 17511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173760151383228e-05,
+      "loss": 0.9783,
+      "step": 17512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1736761713065262e-05,
+      "loss": 0.9842,
+      "step": 17513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173592189966832e-05,
+      "loss": 0.8794,
+      "step": 17514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1735082073647561e-05,
+      "loss": 1.0076,
+      "step": 17515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1734242235009093e-05,
+      "loss": 0.9497,
+      "step": 17516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1733402383759023e-05,
+      "loss": 0.9182,
+      "step": 17517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1732562519903459e-05,
+      "loss": 1.0867,
+      "step": 17518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1731722643448504e-05,
+      "loss": 0.9359,
+      "step": 17519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1730882754400274e-05,
+      "loss": 1.0362,
+      "step": 17520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173004285276487e-05,
+      "loss": 0.6568,
+      "step": 17521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1729202938548403e-05,
+      "loss": 0.7396,
+      "step": 17522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172836301175698e-05,
+      "loss": 1.0225,
+      "step": 17523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172752307239671e-05,
+      "loss": 1.2891,
+      "step": 17524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17266831204737e-05,
+      "loss": 0.932,
+      "step": 17525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172584315599406e-05,
+      "loss": 1.0264,
+      "step": 17526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1725003178963895e-05,
+      "loss": 1.1007,
+      "step": 17527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1724163189389312e-05,
+      "loss": 0.8491,
+      "step": 17528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1723323187276428e-05,
+      "loss": 1.0362,
+      "step": 17529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1722483172631344e-05,
+      "loss": 0.9353,
+      "step": 17530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1721643145460173e-05,
+      "loss": 0.7272,
+      "step": 17531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172080310576902e-05,
+      "loss": 0.8942,
+      "step": 17532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1719963053563999e-05,
+      "loss": 1.0549,
+      "step": 17533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171912298885121e-05,
+      "loss": 1.0071,
+      "step": 17534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1718282911636774e-05,
+      "loss": 0.8981,
+      "step": 17535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1717442821926793e-05,
+      "loss": 0.8796,
+      "step": 17536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1716602719727373e-05,
+      "loss": 0.6715,
+      "step": 17537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1715762605044628e-05,
+      "loss": 0.8314,
+      "step": 17538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171492247788467e-05,
+      "loss": 0.9883,
+      "step": 17539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1714082338253603e-05,
+      "loss": 0.9597,
+      "step": 17540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1713242186157538e-05,
+      "loss": 1.077,
+      "step": 17541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171240202160259e-05,
+      "loss": 0.5843,
+      "step": 17542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1711561844594856e-05,
+      "loss": 0.967,
+      "step": 17543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1710721655140462e-05,
+      "loss": 0.8713,
+      "step": 17544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1709881453245508e-05,
+      "loss": 0.9847,
+      "step": 17545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1709041238916105e-05,
+      "loss": 0.8596,
+      "step": 17546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1708201012158367e-05,
+      "loss": 0.8574,
+      "step": 17547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1707360772978398e-05,
+      "loss": 0.8992,
+      "step": 17548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1706520521382312e-05,
+      "loss": 1.2568,
+      "step": 17549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1705680257376224e-05,
+      "loss": 1.0193,
+      "step": 17550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1704839980966236e-05,
+      "loss": 1.0125,
+      "step": 17551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170399969215846e-05,
+      "loss": 0.4198,
+      "step": 17552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1703159390959014e-05,
+      "loss": 1.3306,
+      "step": 17553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1702319077374002e-05,
+      "loss": 0.5613,
+      "step": 17554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1701478751409536e-05,
+      "loss": 0.6423,
+      "step": 17555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1700638413071727e-05,
+      "loss": 1.0219,
+      "step": 17556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169979806236669e-05,
+      "loss": 0.9516,
+      "step": 17557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1698957699300527e-05,
+      "loss": 0.5897,
+      "step": 17558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1698117323879358e-05,
+      "loss": 0.9546,
+      "step": 17559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1697276936109293e-05,
+      "loss": 0.9282,
+      "step": 17560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.169643653599644e-05,
+      "loss": 0.7947,
+      "step": 17561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1695596123546912e-05,
+      "loss": 1.1518,
+      "step": 17562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1694755698766822e-05,
+      "loss": 0.5811,
+      "step": 17563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1693915261662277e-05,
+      "loss": 0.7684,
+      "step": 17564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1693074812239397e-05,
+      "loss": 0.8553,
+      "step": 17565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1692234350504286e-05,
+      "loss": 0.9596,
+      "step": 17566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1691393876463057e-05,
+      "loss": 1.197,
+      "step": 17567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1690553390121826e-05,
+      "loss": 1.031,
+      "step": 17568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1689712891486704e-05,
+      "loss": 1.5238,
+      "step": 17569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1688872380563803e-05,
+      "loss": 0.9546,
+      "step": 17570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1688031857359233e-05,
+      "loss": 0.5155,
+      "step": 17571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1687191321879107e-05,
+      "loss": 0.9159,
+      "step": 17572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1686350774129538e-05,
+      "loss": 0.9024,
+      "step": 17573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168551021411664e-05,
+      "loss": 0.8357,
+      "step": 17574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1684669641846526e-05,
+      "loss": 1.2408,
+      "step": 17575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1683829057325304e-05,
+      "loss": 0.9273,
+      "step": 17576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1682988460559093e-05,
+      "loss": 0.8217,
+      "step": 17577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1682147851554002e-05,
+      "loss": 1.0709,
+      "step": 17578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1681307230316146e-05,
+      "loss": 0.8712,
+      "step": 17579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1680466596851635e-05,
+      "loss": 0.8711,
+      "step": 17580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1679625951166587e-05,
+      "loss": 0.9906,
+      "step": 17581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1678785293267107e-05,
+      "loss": 1.2376,
+      "step": 17582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1677944623159321e-05,
+      "loss": 0.7389,
+      "step": 17583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167710394084933e-05,
+      "loss": 0.7072,
+      "step": 17584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1676263246343255e-05,
+      "loss": 0.9721,
+      "step": 17585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167542253964721e-05,
+      "loss": 0.9306,
+      "step": 17586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1674581820767304e-05,
+      "loss": 1.2182,
+      "step": 17587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1673741089709654e-05,
+      "loss": 0.9559,
+      "step": 17588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167290034648037e-05,
+      "loss": 0.7241,
+      "step": 17589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1672059591085572e-05,
+      "loss": 0.9651,
+      "step": 17590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1671218823531368e-05,
+      "loss": 0.8784,
+      "step": 17591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1670378043823878e-05,
+      "loss": 0.9601,
+      "step": 17592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1669537251969212e-05,
+      "loss": 1.0018,
+      "step": 17593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1668696447973486e-05,
+      "loss": 0.9026,
+      "step": 17594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1667855631842815e-05,
+      "loss": 0.9102,
+      "step": 17595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1667014803583312e-05,
+      "loss": 1.1772,
+      "step": 17596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1666173963201093e-05,
+      "loss": 0.9236,
+      "step": 17597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1665333110702268e-05,
+      "loss": 0.9039,
+      "step": 17598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1664492246092958e-05,
+      "loss": 0.829,
+      "step": 17599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1663651369379277e-05,
+      "loss": 1.2016,
+      "step": 17600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166281048056734e-05,
+      "loss": 1.0908,
+      "step": 17601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1661969579663257e-05,
+      "loss": 1.2715,
+      "step": 17602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1661128666673147e-05,
+      "loss": 0.9917,
+      "step": 17603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1660287741603124e-05,
+      "loss": 0.8161,
+      "step": 17604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.165944680445931e-05,
+      "loss": 0.8577,
+      "step": 17605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1658605855247809e-05,
+      "loss": 0.7567,
+      "step": 17606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1657764893974741e-05,
+      "loss": 0.9516,
+      "step": 17607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1656923920646226e-05,
+      "loss": 1.0356,
+      "step": 17608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1656082935268375e-05,
+      "loss": 0.9505,
+      "step": 17609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1655241937847305e-05,
+      "loss": 1.6203,
+      "step": 17610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1654400928389133e-05,
+      "loss": 1.3038,
+      "step": 17611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1653559906899973e-05,
+      "loss": 0.8988,
+      "step": 17612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.165271887338594e-05,
+      "loss": 0.9899,
+      "step": 17613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1651877827853155e-05,
+      "loss": 1.1717,
+      "step": 17614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1651036770307731e-05,
+      "loss": 0.9983,
+      "step": 17615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1650195700755782e-05,
+      "loss": 0.9246,
+      "step": 17616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1649354619203427e-05,
+      "loss": 0.5679,
+      "step": 17617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1648513525656781e-05,
+      "loss": 0.7639,
+      "step": 17618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1647672420121962e-05,
+      "loss": 0.6914,
+      "step": 17619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164683130260509e-05,
+      "loss": 0.9436,
+      "step": 17620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1645990173112277e-05,
+      "loss": 0.6998,
+      "step": 17621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1645149031649634e-05,
+      "loss": 0.7335,
+      "step": 17622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1644307878223291e-05,
+      "loss": 0.9467,
+      "step": 17623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1643466712839357e-05,
+      "loss": 0.7367,
+      "step": 17624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164262553550395e-05,
+      "loss": 1.0912,
+      "step": 17625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1641784346223188e-05,
+      "loss": 0.7048,
+      "step": 17626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1640943145003189e-05,
+      "loss": 0.7645,
+      "step": 17627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1640101931850065e-05,
+      "loss": 0.8222,
+      "step": 17628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1639260706769942e-05,
+      "loss": 1.0507,
+      "step": 17629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1638419469768934e-05,
+      "loss": 0.7999,
+      "step": 17630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1637578220853155e-05,
+      "loss": 1.095,
+      "step": 17631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1636736960028725e-05,
+      "loss": 0.8792,
+      "step": 17632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1635895687301764e-05,
+      "loss": 1.1478,
+      "step": 17633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1635054402678385e-05,
+      "loss": 0.769,
+      "step": 17634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1634213106164714e-05,
+      "loss": 0.7751,
+      "step": 17635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163337179776686e-05,
+      "loss": 1.2491,
+      "step": 17636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1632530477490942e-05,
+      "loss": 0.7157,
+      "step": 17637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1631689145343085e-05,
+      "loss": 1.1353,
+      "step": 17638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1630847801329403e-05,
+      "loss": 1.1489,
+      "step": 17639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1630006445456015e-05,
+      "loss": 1.1319,
+      "step": 17640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1629165077729039e-05,
+      "loss": 0.9528,
+      "step": 17641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1628323698154594e-05,
+      "loss": 1.1206,
+      "step": 17642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1627482306738795e-05,
+      "loss": 0.7392,
+      "step": 17643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1626640903487767e-05,
+      "loss": 0.8723,
+      "step": 17644
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1625799488407627e-05,
+      "loss": 0.9189,
+      "step": 17645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162495806150449e-05,
+      "loss": 0.9339,
+      "step": 17646
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1624116622784478e-05,
+      "loss": 0.7655,
+      "step": 17647
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162327517225371e-05,
+      "loss": 1.0462,
+      "step": 17648
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1622433709918306e-05,
+      "loss": 1.0127,
+      "step": 17649
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1621592235784383e-05,
+      "loss": 0.9547,
+      "step": 17650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162075074985806e-05,
+      "loss": 1.1715,
+      "step": 17651
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1619909252145455e-05,
+      "loss": 0.9678,
+      "step": 17652
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1619067742652694e-05,
+      "loss": 0.9722,
+      "step": 17653
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1618226221385894e-05,
+      "loss": 1.0677,
+      "step": 17654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161738468835117e-05,
+      "loss": 0.9428,
+      "step": 17655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1616543143554646e-05,
+      "loss": 0.96,
+      "step": 17656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161570158700244e-05,
+      "loss": 1.0815,
+      "step": 17657
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1614860018700675e-05,
+      "loss": 0.6972,
+      "step": 17658
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1614018438655465e-05,
+      "loss": 1.2149,
+      "step": 17659
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1613176846872938e-05,
+      "loss": 1.2964,
+      "step": 17660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1612335243359204e-05,
+      "loss": 0.7776,
+      "step": 17661
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1611493628120394e-05,
+      "loss": 0.7263,
+      "step": 17662
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610652001162623e-05,
+      "loss": 0.6642,
+      "step": 17663
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1609810362492011e-05,
+      "loss": 0.7567,
+      "step": 17664
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1608968712114679e-05,
+      "loss": 1.0427,
+      "step": 17665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1608127050036749e-05,
+      "loss": 0.8451,
+      "step": 17666
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160728537626434e-05,
+      "loss": 0.6996,
+      "step": 17667
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1606443690803569e-05,
+      "loss": 1.0058,
+      "step": 17668
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1605601993660565e-05,
+      "loss": 1.0474,
+      "step": 17669
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1604760284841446e-05,
+      "loss": 0.9306,
+      "step": 17670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160391856435233e-05,
+      "loss": 0.9115,
+      "step": 17671
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160307683219934e-05,
+      "loss": 0.9662,
+      "step": 17672
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.16022350883886e-05,
+      "loss": 0.9557,
+      "step": 17673
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1601393332926225e-05,
+      "loss": 0.7806,
+      "step": 17674
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1600551565818344e-05,
+      "loss": 0.6968,
+      "step": 17675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1599709787071071e-05,
+      "loss": 1.4969,
+      "step": 17676
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159886799669053e-05,
+      "loss": 0.7776,
+      "step": 17677
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1598026194682845e-05,
+      "loss": 0.7663,
+      "step": 17678
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1597184381054137e-05,
+      "loss": 1.1552,
+      "step": 17679
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1596342555810526e-05,
+      "loss": 1.1977,
+      "step": 17680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1595500718958133e-05,
+      "loss": 1.0883,
+      "step": 17681
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1594658870503084e-05,
+      "loss": 0.8722,
+      "step": 17682
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1593817010451496e-05,
+      "loss": 0.6972,
+      "step": 17683
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592975138809495e-05,
+      "loss": 0.8018,
+      "step": 17684
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592133255583204e-05,
+      "loss": 1.0303,
+      "step": 17685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159129136077874e-05,
+      "loss": 0.7189,
+      "step": 17686
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.159044945440223e-05,
+      "loss": 0.6609,
+      "step": 17687
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1589607536459795e-05,
+      "loss": 1.1763,
+      "step": 17688
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1588765606957555e-05,
+      "loss": 0.6764,
+      "step": 17689
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1587923665901642e-05,
+      "loss": 0.6552,
+      "step": 17690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1587081713298165e-05,
+      "loss": 0.7455,
+      "step": 17691
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586239749153255e-05,
+      "loss": 1.0801,
+      "step": 17692
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1585397773473034e-05,
+      "loss": 1.2768,
+      "step": 17693
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1584555786263625e-05,
+      "loss": 1.0029,
+      "step": 17694
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.158371378753115e-05,
+      "loss": 1.0266,
+      "step": 17695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1582871777281732e-05,
+      "loss": 0.9274,
+      "step": 17696
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1582029755521496e-05,
+      "loss": 1.0101,
+      "step": 17697
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1581187722256562e-05,
+      "loss": 0.8205,
+      "step": 17698
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1580345677493055e-05,
+      "loss": 0.7143,
+      "step": 17699
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1579503621237102e-05,
+      "loss": 0.7512,
+      "step": 17700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1578661553494821e-05,
+      "loss": 0.7642,
+      "step": 17701
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1577819474272338e-05,
+      "loss": 0.6775,
+      "step": 17702
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576977383575777e-05,
+      "loss": 0.5903,
+      "step": 17703
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576135281411259e-05,
+      "loss": 1.1321,
+      "step": 17704
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1575293167784913e-05,
+      "loss": 0.8719,
+      "step": 17705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1574451042702861e-05,
+      "loss": 0.9891,
+      "step": 17706
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1573608906171221e-05,
+      "loss": 1.0288,
+      "step": 17707
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1572766758196127e-05,
+      "loss": 0.7593,
+      "step": 17708
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1571924598783697e-05,
+      "loss": 0.9452,
+      "step": 17709
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1571082427940058e-05,
+      "loss": 0.9674,
+      "step": 17710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1570240245671331e-05,
+      "loss": 1.2203,
+      "step": 17711
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1569398051983644e-05,
+      "loss": 0.807,
+      "step": 17712
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1568555846883119e-05,
+      "loss": 0.9667,
+      "step": 17713
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1567713630375883e-05,
+      "loss": 0.788,
+      "step": 17714
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156687140246806e-05,
+      "loss": 0.7952,
+      "step": 17715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1566029163165773e-05,
+      "loss": 0.8594,
+      "step": 17716
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1565186912475149e-05,
+      "loss": 0.8407,
+      "step": 17717
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156434465040231e-05,
+      "loss": 0.8635,
+      "step": 17718
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1563502376953384e-05,
+      "loss": 1.3549,
+      "step": 17719
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1562660092134497e-05,
+      "loss": 0.7308,
+      "step": 17720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1561817795951771e-05,
+      "loss": 0.6096,
+      "step": 17721
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1560975488411328e-05,
+      "loss": 1.1007,
+      "step": 17722
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1560133169519302e-05,
+      "loss": 0.7706,
+      "step": 17723
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1559290839281815e-05,
+      "loss": 0.6883,
+      "step": 17724
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.155844849770499e-05,
+      "loss": 0.7075,
+      "step": 17725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1557606144794957e-05,
+      "loss": 0.9086,
+      "step": 17726
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1556763780557838e-05,
+      "loss": 0.8999,
+      "step": 17727
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1555921404999759e-05,
+      "loss": 0.7885,
+      "step": 17728
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1555079018126847e-05,
+      "loss": 0.8507,
+      "step": 17729
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1554236619945229e-05,
+      "loss": 1.1237,
+      "step": 17730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1553394210461028e-05,
+      "loss": 0.9629,
+      "step": 17731
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552551789680373e-05,
+      "loss": 1.1525,
+      "step": 17732
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1551709357609388e-05,
+      "loss": 0.7616,
+      "step": 17733
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1550866914254202e-05,
+      "loss": 1.0726,
+      "step": 17734
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1550024459620938e-05,
+      "loss": 1.1293,
+      "step": 17735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1549181993715726e-05,
+      "loss": 0.9176,
+      "step": 17736
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1548339516544689e-05,
+      "loss": 0.72,
+      "step": 17737
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1547497028113953e-05,
+      "loss": 0.9966,
+      "step": 17738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1546654528429649e-05,
+      "loss": 0.8931,
+      "step": 17739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1545812017497902e-05,
+      "loss": 0.9511,
+      "step": 17740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1544969495324836e-05,
+      "loss": 0.8154,
+      "step": 17741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1544126961916581e-05,
+      "loss": 1.2088,
+      "step": 17742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1543284417279267e-05,
+      "loss": 0.5546,
+      "step": 17743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1542441861419011e-05,
+      "loss": 0.7623,
+      "step": 17744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1541599294341952e-05,
+      "loss": 0.87,
+      "step": 17745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154075671605421e-05,
+      "loss": 1.123,
+      "step": 17746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539914126561912e-05,
+      "loss": 0.9466,
+      "step": 17747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539071525871191e-05,
+      "loss": 0.6166,
+      "step": 17748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1538228913988168e-05,
+      "loss": 0.763,
+      "step": 17749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1537386290918976e-05,
+      "loss": 1.1789,
+      "step": 17750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153654365666974e-05,
+      "loss": 1.098,
+      "step": 17751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1535701011246585e-05,
+      "loss": 0.6104,
+      "step": 17752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1534858354655643e-05,
+      "loss": 1.0312,
+      "step": 17753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153401568690304e-05,
+      "loss": 0.669,
+      "step": 17754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153317300799491e-05,
+      "loss": 0.7593,
+      "step": 17755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1532330317937368e-05,
+      "loss": 0.8504,
+      "step": 17756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1531487616736555e-05,
+      "loss": 1.0128,
+      "step": 17757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1530644904398592e-05,
+      "loss": 1.5055,
+      "step": 17758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1529802180929605e-05,
+      "loss": 1.0114,
+      "step": 17759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528959446335735e-05,
+      "loss": 1.124,
+      "step": 17760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528116700623098e-05,
+      "loss": 0.9674,
+      "step": 17761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1527273943797825e-05,
+      "loss": 1.013,
+      "step": 17762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152643117586605e-05,
+      "loss": 1.0468,
+      "step": 17763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1525588396833895e-05,
+      "loss": 0.9979,
+      "step": 17764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1524745606707492e-05,
+      "loss": 0.9657,
+      "step": 17765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152390280549297e-05,
+      "loss": 1.1506,
+      "step": 17766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1523059993196459e-05,
+      "loss": 0.7132,
+      "step": 17767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1522217169824083e-05,
+      "loss": 0.9501,
+      "step": 17768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1521374335381979e-05,
+      "loss": 0.8188,
+      "step": 17769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1520531489876273e-05,
+      "loss": 0.8102,
+      "step": 17770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1519688633313089e-05,
+      "loss": 0.7757,
+      "step": 17771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151884576569856e-05,
+      "loss": 0.7058,
+      "step": 17772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151800288703882e-05,
+      "loss": 0.8376,
+      "step": 17773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151715999733999e-05,
+      "loss": 0.7414,
+      "step": 17774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1516317096608207e-05,
+      "loss": 0.7148,
+      "step": 17775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1515474184849598e-05,
+      "loss": 0.8784,
+      "step": 17776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1514631262070291e-05,
+      "loss": 0.6812,
+      "step": 17777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1513788328276416e-05,
+      "loss": 1.3791,
+      "step": 17778
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1512945383474105e-05,
+      "loss": 0.9198,
+      "step": 17779
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151210242766949e-05,
+      "loss": 0.8682,
+      "step": 17780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1511259460868696e-05,
+      "loss": 0.7405,
+      "step": 17781
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1510416483077855e-05,
+      "loss": 1.0399,
+      "step": 17782
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1509573494303099e-05,
+      "loss": 1.0744,
+      "step": 17783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1508730494550555e-05,
+      "loss": 0.9916,
+      "step": 17784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.150788748382636e-05,
+      "loss": 0.8716,
+      "step": 17785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1507044462136631e-05,
+      "loss": 1.113,
+      "step": 17786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1506201429487513e-05,
+      "loss": 0.8326,
+      "step": 17787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1505358385885133e-05,
+      "loss": 0.7711,
+      "step": 17788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1504515331335617e-05,
+      "loss": 0.9093,
+      "step": 17789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1503672265845098e-05,
+      "loss": 1.0333,
+      "step": 17790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1502829189419708e-05,
+      "loss": 0.7229,
+      "step": 17791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1501986102065579e-05,
+      "loss": 0.8099,
+      "step": 17792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1501143003788834e-05,
+      "loss": 0.6683,
+      "step": 17793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1500299894595617e-05,
+      "loss": 1.0044,
+      "step": 17794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149945677449205e-05,
+      "loss": 1.1107,
+      "step": 17795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1498613643484266e-05,
+      "loss": 0.6989,
+      "step": 17796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14977705015784e-05,
+      "loss": 0.9186,
+      "step": 17797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149692734878058e-05,
+      "loss": 0.905,
+      "step": 17798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1496084185096934e-05,
+      "loss": 0.8824,
+      "step": 17799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1495241010533602e-05,
+      "loss": 0.9529,
+      "step": 17800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149439782509671e-05,
+      "loss": 0.728,
+      "step": 17801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149355462879239e-05,
+      "loss": 0.7148,
+      "step": 17802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1492711421626776e-05,
+      "loss": 0.5714,
+      "step": 17803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1491868203606e-05,
+      "loss": 0.9764,
+      "step": 17804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1491024974736191e-05,
+      "loss": 1.0983,
+      "step": 17805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1490181735023484e-05,
+      "loss": 0.9016,
+      "step": 17806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1489338484474009e-05,
+      "loss": 0.9353,
+      "step": 17807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1488495223093899e-05,
+      "loss": 0.9004,
+      "step": 17808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1487651950889284e-05,
+      "loss": 0.7498,
+      "step": 17809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1486808667866306e-05,
+      "loss": 0.9084,
+      "step": 17810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1485965374031083e-05,
+      "loss": 1.1393,
+      "step": 17811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1485122069389759e-05,
+      "loss": 1.0481,
+      "step": 17812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148427875394846e-05,
+      "loss": 0.9877,
+      "step": 17813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148343542771332e-05,
+      "loss": 0.9951,
+      "step": 17814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1482592090690478e-05,
+      "loss": 0.9474,
+      "step": 17815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1481748742886059e-05,
+      "loss": 0.9859,
+      "step": 17816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1480905384306196e-05,
+      "loss": 0.8685,
+      "step": 17817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1480062014957028e-05,
+      "loss": 1.244,
+      "step": 17818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1479218634844685e-05,
+      "loss": 0.9713,
+      "step": 17819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1478375243975298e-05,
+      "loss": 1.1382,
+      "step": 17820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1477531842355e-05,
+      "loss": 0.9135,
+      "step": 17821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.147668842998993e-05,
+      "loss": 0.9017,
+      "step": 17822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1475845006886214e-05,
+      "loss": 0.5776,
+      "step": 17823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1475001573049993e-05,
+      "loss": 1.0733,
+      "step": 17824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1474158128487398e-05,
+      "loss": 0.6685,
+      "step": 17825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1473314673204557e-05,
+      "loss": 1.1443,
+      "step": 17826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1472471207207611e-05,
+      "loss": 0.6447,
+      "step": 17827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.147162773050269e-05,
+      "loss": 0.916,
+      "step": 17828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1470784243095927e-05,
+      "loss": 0.7462,
+      "step": 17829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1469940744993463e-05,
+      "loss": 1.3211,
+      "step": 17830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146909723620142e-05,
+      "loss": 0.9314,
+      "step": 17831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146825371672594e-05,
+      "loss": 1.3897,
+      "step": 17832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1467410186573158e-05,
+      "loss": 0.9871,
+      "step": 17833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1466566645749206e-05,
+      "loss": 0.7345,
+      "step": 17834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1465723094260219e-05,
+      "loss": 0.923,
+      "step": 17835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146487953211233e-05,
+      "loss": 0.865,
+      "step": 17836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1464035959311674e-05,
+      "loss": 1.044,
+      "step": 17837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1463192375864388e-05,
+      "loss": 0.7675,
+      "step": 17838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462348781776602e-05,
+      "loss": 0.8542,
+      "step": 17839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1461505177054456e-05,
+      "loss": 0.9622,
+      "step": 17840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1460661561704078e-05,
+      "loss": 1.2112,
+      "step": 17841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145981793573161e-05,
+      "loss": 1.32,
+      "step": 17842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1458974299143182e-05,
+      "loss": 0.9028,
+      "step": 17843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1458130651944934e-05,
+      "loss": 1.0783,
+      "step": 17844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1457286994142995e-05,
+      "loss": 0.9213,
+      "step": 17845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1456443325743505e-05,
+      "loss": 0.8207,
+      "step": 17846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1455599646752596e-05,
+      "loss": 1.0018,
+      "step": 17847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1454755957176405e-05,
+      "loss": 1.0313,
+      "step": 17848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1453912257021067e-05,
+      "loss": 1.0651,
+      "step": 17849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1453068546292718e-05,
+      "loss": 1.1006,
+      "step": 17850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1452224824997493e-05,
+      "loss": 0.8829,
+      "step": 17851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145138109314153e-05,
+      "loss": 1.0622,
+      "step": 17852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1450537350730959e-05,
+      "loss": 0.8625,
+      "step": 17853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1449693597771917e-05,
+      "loss": 1.0577,
+      "step": 17854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144884983427055e-05,
+      "loss": 0.8828,
+      "step": 17855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1448006060232981e-05,
+      "loss": 0.8357,
+      "step": 17856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447162275665354e-05,
+      "loss": 0.995,
+      "step": 17857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1446318480573799e-05,
+      "loss": 0.8073,
+      "step": 17858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1445474674964456e-05,
+      "loss": 1.0755,
+      "step": 17859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1444630858843462e-05,
+      "loss": 0.8083,
+      "step": 17860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1443787032216954e-05,
+      "loss": 0.8513,
+      "step": 17861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1442943195091062e-05,
+      "loss": 0.893,
+      "step": 17862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1442099347471926e-05,
+      "loss": 0.9189,
+      "step": 17863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1441255489365688e-05,
+      "loss": 0.8473,
+      "step": 17864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1440411620778478e-05,
+      "loss": 1.2524,
+      "step": 17865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1439567741716436e-05,
+      "loss": 0.8434,
+      "step": 17866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1438723852185695e-05,
+      "loss": 0.775,
+      "step": 17867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1437879952192396e-05,
+      "loss": 0.7494,
+      "step": 17868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1437036041742672e-05,
+      "loss": 0.8267,
+      "step": 17869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1436192120842669e-05,
+      "loss": 1.0323,
+      "step": 17870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1435348189498513e-05,
+      "loss": 0.692,
+      "step": 17871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1434504247716342e-05,
+      "loss": 1.0041,
+      "step": 17872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14336602955023e-05,
+      "loss": 0.9806,
+      "step": 17873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1432816332862522e-05,
+      "loss": 0.9182,
+      "step": 17874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1431972359803145e-05,
+      "loss": 1.0874,
+      "step": 17875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1431128376330306e-05,
+      "loss": 0.7795,
+      "step": 17876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.143028438245014e-05,
+      "loss": 1.0959,
+      "step": 17877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1429440378168786e-05,
+      "loss": 0.8305,
+      "step": 17878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1428596363492385e-05,
+      "loss": 1.1481,
+      "step": 17879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1427752338427075e-05,
+      "loss": 0.7926,
+      "step": 17880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.142690830297899e-05,
+      "loss": 0.9462,
+      "step": 17881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1426064257154268e-05,
+      "loss": 0.8862,
+      "step": 17882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1425220200959051e-05,
+      "loss": 0.7961,
+      "step": 17883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1424376134399468e-05,
+      "loss": 0.9808,
+      "step": 17884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1423532057481673e-05,
+      "loss": 1.0287,
+      "step": 17885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.142268797021179e-05,
+      "loss": 0.8571,
+      "step": 17886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1421843872595958e-05,
+      "loss": 0.8738,
+      "step": 17887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1420999764640323e-05,
+      "loss": 0.8879,
+      "step": 17888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1420155646351021e-05,
+      "loss": 0.9291,
+      "step": 17889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141931151773419e-05,
+      "loss": 0.967,
+      "step": 17890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1418467378795967e-05,
+      "loss": 1.3715,
+      "step": 17891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1417623229542492e-05,
+      "loss": 1.079,
+      "step": 17892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1416779069979902e-05,
+      "loss": 0.7046,
+      "step": 17893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1415934900114339e-05,
+      "loss": 0.7641,
+      "step": 17894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141509071995194e-05,
+      "loss": 0.8042,
+      "step": 17895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1414246529498843e-05,
+      "loss": 0.9743,
+      "step": 17896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141340232876119e-05,
+      "loss": 0.8975,
+      "step": 17897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1412558117745116e-05,
+      "loss": 1.1886,
+      "step": 17898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1411713896456762e-05,
+      "loss": 1.0627,
+      "step": 17899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410869664902271e-05,
+      "loss": 0.9805,
+      "step": 17900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410025423087778e-05,
+      "loss": 0.9165,
+      "step": 17901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140918117101942e-05,
+      "loss": 0.8618,
+      "step": 17902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1408336908703345e-05,
+      "loss": 0.9387,
+      "step": 17903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1407492636145686e-05,
+      "loss": 0.9906,
+      "step": 17904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1406648353352582e-05,
+      "loss": 0.8649,
+      "step": 17905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1405804060330179e-05,
+      "loss": 1.1964,
+      "step": 17906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140495975708461e-05,
+      "loss": 0.6946,
+      "step": 17907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1404115443622016e-05,
+      "loss": 0.7564,
+      "step": 17908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1403271119948543e-05,
+      "loss": 0.8501,
+      "step": 17909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1402426786070326e-05,
+      "loss": 0.8061,
+      "step": 17910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1401582441993501e-05,
+      "loss": 1.1565,
+      "step": 17911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1400738087724217e-05,
+      "loss": 0.7775,
+      "step": 17912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.139989372326861e-05,
+      "loss": 0.9049,
+      "step": 17913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.139904934863282e-05,
+      "loss": 0.7915,
+      "step": 17914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1398204963822988e-05,
+      "loss": 0.9813,
+      "step": 17915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1397360568845253e-05,
+      "loss": 0.9535,
+      "step": 17916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1396516163705755e-05,
+      "loss": 0.7103,
+      "step": 17917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1395671748410639e-05,
+      "loss": 0.7721,
+      "step": 17918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1394827322966041e-05,
+      "loss": 0.8795,
+      "step": 17919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1393982887378108e-05,
+      "loss": 1.1356,
+      "step": 17920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1393138441652974e-05,
+      "loss": 0.8629,
+      "step": 17921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1392293985796783e-05,
+      "loss": 1.0178,
+      "step": 17922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1391449519815676e-05,
+      "loss": 0.9863,
+      "step": 17923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.139060504371579e-05,
+      "loss": 1.2776,
+      "step": 17924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1389760557503275e-05,
+      "loss": 0.9747,
+      "step": 17925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1388916061184264e-05,
+      "loss": 1.0126,
+      "step": 17926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13880715547649e-05,
+      "loss": 0.8903,
+      "step": 17927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1387227038251327e-05,
+      "loss": 1.1133,
+      "step": 17928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1386382511649687e-05,
+      "loss": 0.9391,
+      "step": 17929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1385537974966116e-05,
+      "loss": 1.145,
+      "step": 17930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1384693428206761e-05,
+      "loss": 1.0231,
+      "step": 17931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138384887137776e-05,
+      "loss": 0.7903,
+      "step": 17932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1383004304485256e-05,
+      "loss": 0.8014,
+      "step": 17933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1382159727535393e-05,
+      "loss": 1.3065,
+      "step": 17934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1381315140534313e-05,
+      "loss": 1.215,
+      "step": 17935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1380470543488152e-05,
+      "loss": 0.7775,
+      "step": 17936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1379625936403058e-05,
+      "loss": 0.9139,
+      "step": 17937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137878131928517e-05,
+      "loss": 1.1598,
+      "step": 17938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137793669214063e-05,
+      "loss": 0.5761,
+      "step": 17939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1377092054975586e-05,
+      "loss": 0.8871,
+      "step": 17940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137624740779617e-05,
+      "loss": 1.1465,
+      "step": 17941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1375402750608533e-05,
+      "loss": 1.1187,
+      "step": 17942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1374558083418811e-05,
+      "loss": 1.2004,
+      "step": 17943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1373713406233156e-05,
+      "loss": 1.0092,
+      "step": 17944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1372868719057702e-05,
+      "loss": 0.8434,
+      "step": 17945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1372024021898593e-05,
+      "loss": 0.8565,
+      "step": 17946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1371179314761973e-05,
+      "loss": 0.6264,
+      "step": 17947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1370334597653984e-05,
+      "loss": 0.9318,
+      "step": 17948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1369489870580772e-05,
+      "loss": 0.9077,
+      "step": 17949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1368645133548478e-05,
+      "loss": 0.8259,
+      "step": 17950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1367800386563241e-05,
+      "loss": 1.142,
+      "step": 17951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1366955629631211e-05,
+      "loss": 0.9333,
+      "step": 17952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1366110862758526e-05,
+      "loss": 1.1118,
+      "step": 17953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.136526608595133e-05,
+      "loss": 0.6412,
+      "step": 17954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1364421299215773e-05,
+      "loss": 1.2688,
+      "step": 17955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1363576502557988e-05,
+      "loss": 1.1568,
+      "step": 17956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1362731695984123e-05,
+      "loss": 0.8071,
+      "step": 17957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1361886879500324e-05,
+      "loss": 1.0736,
+      "step": 17958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1361042053112732e-05,
+      "loss": 1.1075,
+      "step": 17959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1360197216827489e-05,
+      "loss": 0.9913,
+      "step": 17960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1359352370650742e-05,
+      "loss": 0.7468,
+      "step": 17961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1358507514588635e-05,
+      "loss": 0.9659,
+      "step": 17962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1357662648647308e-05,
+      "loss": 0.6851,
+      "step": 17963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135681777283291e-05,
+      "loss": 1.3654,
+      "step": 17964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1355972887151583e-05,
+      "loss": 0.908,
+      "step": 17965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1355127991609467e-05,
+      "loss": 1.2399,
+      "step": 17966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1354283086212712e-05,
+      "loss": 0.8973,
+      "step": 17967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1353438170967459e-05,
+      "loss": 1.1553,
+      "step": 17968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1352593245879852e-05,
+      "loss": 0.7733,
+      "step": 17969
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135174831095604e-05,
+      "loss": 0.9707,
+      "step": 17970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1350903366202165e-05,
+      "loss": 0.7285,
+      "step": 17971
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1350058411624367e-05,
+      "loss": 1.0669,
+      "step": 17972
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1349213447228795e-05,
+      "loss": 0.8321,
+      "step": 17973
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1348368473021596e-05,
+      "loss": 1.0704,
+      "step": 17974
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134752348900891e-05,
+      "loss": 0.7526,
+      "step": 17975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1346678495196883e-05,
+      "loss": 1.2413,
+      "step": 17976
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134583349159166e-05,
+      "loss": 0.7008,
+      "step": 17977
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1344988478199385e-05,
+      "loss": 0.7093,
+      "step": 17978
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1344143455026206e-05,
+      "loss": 0.6457,
+      "step": 17979
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134329842207827e-05,
+      "loss": 1.1598,
+      "step": 17980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1342453379361715e-05,
+      "loss": 0.6855,
+      "step": 17981
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1341608326882691e-05,
+      "loss": 0.9727,
+      "step": 17982
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.134076326464734e-05,
+      "loss": 0.8608,
+      "step": 17983
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339918192661813e-05,
+      "loss": 0.8513,
+      "step": 17984
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.133907311093225e-05,
+      "loss": 0.5304,
+      "step": 17985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1338228019464801e-05,
+      "loss": 0.8886,
+      "step": 17986
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1337382918265604e-05,
+      "loss": 0.9357,
+      "step": 17987
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1336537807340816e-05,
+      "loss": 0.7249,
+      "step": 17988
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1335692686696573e-05,
+      "loss": 0.6025,
+      "step": 17989
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1334847556339026e-05,
+      "loss": 0.6842,
+      "step": 17990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1334002416274319e-05,
+      "loss": 0.8708,
+      "step": 17991
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13331572665086e-05,
+      "loss": 1.0091,
+      "step": 17992
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1332312107048011e-05,
+      "loss": 1.0727,
+      "step": 17993
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1331466937898698e-05,
+      "loss": 0.8697,
+      "step": 17994
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1330621759066816e-05,
+      "loss": 1.2557,
+      "step": 17995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1329776570558499e-05,
+      "loss": 1.0327,
+      "step": 17996
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1328931372379901e-05,
+      "loss": 0.8289,
+      "step": 17997
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1328086164537169e-05,
+      "loss": 0.9326,
+      "step": 17998
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1327240947036444e-05,
+      "loss": 0.9109,
+      "step": 17999
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1326395719883876e-05,
+      "loss": 1.1888,
+      "step": 18000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1325550483085613e-05,
+      "loss": 1.0224,
+      "step": 18001
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1324705236647797e-05,
+      "loss": 0.9146,
+      "step": 18002
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1323859980576576e-05,
+      "loss": 1.0158,
+      "step": 18003
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1323014714878103e-05,
+      "loss": 0.7784,
+      "step": 18004
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1322169439558517e-05,
+      "loss": 1.1356,
+      "step": 18005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1321324154623967e-05,
+      "loss": 1.0769,
+      "step": 18006
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1320478860080603e-05,
+      "loss": 0.7876,
+      "step": 18007
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131963355593457e-05,
+      "loss": 0.77,
+      "step": 18008
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1318788242192013e-05,
+      "loss": 1.2135,
+      "step": 18009
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1317942918859087e-05,
+      "loss": 1.0193,
+      "step": 18010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1317097585941932e-05,
+      "loss": 0.523,
+      "step": 18011
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1316252243446692e-05,
+      "loss": 0.9902,
+      "step": 18012
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1315406891379523e-05,
+      "loss": 0.7492,
+      "step": 18013
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131456152974657e-05,
+      "loss": 0.5951,
+      "step": 18014
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1313716158553978e-05,
+      "loss": 0.958,
+      "step": 18015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1312870777807897e-05,
+      "loss": 1.1868,
+      "step": 18016
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1312025387514474e-05,
+      "loss": 0.4979,
+      "step": 18017
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1311179987679855e-05,
+      "loss": 1.425,
+      "step": 18018
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131033457831019e-05,
+      "loss": 0.8,
+      "step": 18019
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130948915941163e-05,
+      "loss": 0.994,
+      "step": 18020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1308643730990314e-05,
+      "loss": 0.9088,
+      "step": 18021
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1307798293052398e-05,
+      "loss": 0.8876,
+      "step": 18022
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1306952845604029e-05,
+      "loss": 0.8321,
+      "step": 18023
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130610738865135e-05,
+      "loss": 0.9856,
+      "step": 18024
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130526192220052e-05,
+      "loss": 0.7736,
+      "step": 18025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1304416446257674e-05,
+      "loss": 0.7546,
+      "step": 18026
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1303570960828967e-05,
+      "loss": 0.4974,
+      "step": 18027
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130272546592055e-05,
+      "loss": 0.844,
+      "step": 18028
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1301879961538566e-05,
+      "loss": 1.3314,
+      "step": 18029
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130103444768917e-05,
+      "loss": 1.078,
+      "step": 18030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1300188924378503e-05,
+      "loss": 0.9509,
+      "step": 18031
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.129934339161272e-05,
+      "loss": 1.2103,
+      "step": 18032
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1298497849397967e-05,
+      "loss": 1.0713,
+      "step": 18033
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1297652297740393e-05,
+      "loss": 0.869,
+      "step": 18034
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1296806736646151e-05,
+      "loss": 1.0523,
+      "step": 18035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.129596116612138e-05,
+      "loss": 1.1827,
+      "step": 18036
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.129511558617224e-05,
+      "loss": 0.5445,
+      "step": 18037
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1294269996804876e-05,
+      "loss": 0.9847,
+      "step": 18038
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1293424398025434e-05,
+      "loss": 1.0719,
+      "step": 18039
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1292578789840069e-05,
+      "loss": 0.8283,
+      "step": 18040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1291733172254927e-05,
+      "loss": 1.0059,
+      "step": 18041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1290887545276155e-05,
+      "loss": 0.9801,
+      "step": 18042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1290041908909908e-05,
+      "loss": 1.0047,
+      "step": 18043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1289196263162331e-05,
+      "loss": 0.7596,
+      "step": 18044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1288350608039577e-05,
+      "loss": 1.1978,
+      "step": 18045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1287504943547795e-05,
+      "loss": 0.8417,
+      "step": 18046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1286659269693135e-05,
+      "loss": 1.1411,
+      "step": 18047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1285813586481744e-05,
+      "loss": 1.0489,
+      "step": 18048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1284967893919771e-05,
+      "loss": 0.6962,
+      "step": 18049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1284122192013376e-05,
+      "loss": 1.0036,
+      "step": 18050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1283276480768696e-05,
+      "loss": 0.8061,
+      "step": 18051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.128243076019189e-05,
+      "loss": 0.4716,
+      "step": 18052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1281585030289101e-05,
+      "loss": 0.8267,
+      "step": 18053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1280739291066488e-05,
+      "loss": 0.9833,
+      "step": 18054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1279893542530196e-05,
+      "loss": 0.9488,
+      "step": 18055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1279047784686374e-05,
+      "loss": 1.1362,
+      "step": 18056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1278202017541175e-05,
+      "loss": 0.9735,
+      "step": 18057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1277356241100747e-05,
+      "loss": 1.1186,
+      "step": 18058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1276510455371245e-05,
+      "loss": 0.8949,
+      "step": 18059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1275664660358818e-05,
+      "loss": 1.1213,
+      "step": 18060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1274818856069617e-05,
+      "loss": 0.9024,
+      "step": 18061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1273973042509789e-05,
+      "loss": 0.9474,
+      "step": 18062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1273127219685489e-05,
+      "loss": 0.9687,
+      "step": 18063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1272281387602863e-05,
+      "loss": 0.8764,
+      "step": 18064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1271435546268071e-05,
+      "loss": 0.958,
+      "step": 18065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1270589695687255e-05,
+      "loss": 0.9027,
+      "step": 18066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1269743835866569e-05,
+      "loss": 1.1491,
+      "step": 18067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1268897966812166e-05,
+      "loss": 0.9995,
+      "step": 18068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1268052088530196e-05,
+      "loss": 1.4657,
+      "step": 18069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1267206201026813e-05,
+      "loss": 0.7059,
+      "step": 18070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1266360304308164e-05,
+      "loss": 0.5211,
+      "step": 18071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1265514398380402e-05,
+      "loss": 0.9029,
+      "step": 18072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1264668483249674e-05,
+      "loss": 0.9545,
+      "step": 18073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1263822558922141e-05,
+      "loss": 1.2313,
+      "step": 18074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1262976625403954e-05,
+      "loss": 0.8459,
+      "step": 18075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1262130682701256e-05,
+      "loss": 0.8092,
+      "step": 18076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1261284730820203e-05,
+      "loss": 0.6576,
+      "step": 18077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1260438769766948e-05,
+      "loss": 0.9637,
+      "step": 18078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1259592799547641e-05,
+      "loss": 0.929,
+      "step": 18079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125874682016844e-05,
+      "loss": 0.7712,
+      "step": 18080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1257900831635489e-05,
+      "loss": 0.8903,
+      "step": 18081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1257054833954943e-05,
+      "loss": 1.0511,
+      "step": 18082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1256208827132954e-05,
+      "loss": 0.8106,
+      "step": 18083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1255362811175678e-05,
+      "loss": 1.0068,
+      "step": 18084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1254516786089263e-05,
+      "loss": 0.9135,
+      "step": 18085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1253670751879861e-05,
+      "loss": 0.9228,
+      "step": 18086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1252824708553629e-05,
+      "loss": 0.8063,
+      "step": 18087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1251978656116713e-05,
+      "loss": 0.9101,
+      "step": 18088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1251132594575271e-05,
+      "loss": 0.7985,
+      "step": 18089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1250286523935456e-05,
+      "loss": 1.0949,
+      "step": 18090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1249440444203414e-05,
+      "loss": 1.2241,
+      "step": 18091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1248594355385307e-05,
+      "loss": 0.5944,
+      "step": 18092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1247748257487281e-05,
+      "loss": 1.0215,
+      "step": 18093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124690215051549e-05,
+      "loss": 0.9823,
+      "step": 18094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1246056034476091e-05,
+      "loss": 0.8145,
+      "step": 18095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1245209909375233e-05,
+      "loss": 1.0495,
+      "step": 18096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1244363775219067e-05,
+      "loss": 1.0814,
+      "step": 18097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1243517632013752e-05,
+      "loss": 1.0,
+      "step": 18098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1242671479765439e-05,
+      "loss": 1.0988,
+      "step": 18099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1241825318480281e-05,
+      "loss": 0.9346,
+      "step": 18100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1240979148164432e-05,
+      "loss": 0.8334,
+      "step": 18101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1240132968824044e-05,
+      "loss": 0.96,
+      "step": 18102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1239286780465271e-05,
+      "loss": 0.9427,
+      "step": 18103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1238440583094266e-05,
+      "loss": 0.9991,
+      "step": 18104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1237594376717188e-05,
+      "loss": 0.9262,
+      "step": 18105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1236748161340181e-05,
+      "loss": 1.0755,
+      "step": 18106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1235901936969406e-05,
+      "loss": 0.8732,
+      "step": 18107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1235055703611015e-05,
+      "loss": 0.944,
+      "step": 18108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1234209461271161e-05,
+      "loss": 1.1098,
+      "step": 18109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1233363209956e-05,
+      "loss": 0.7888,
+      "step": 18110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1232516949671685e-05,
+      "loss": 0.9357,
+      "step": 18111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1231670680424368e-05,
+      "loss": 1.2178,
+      "step": 18112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1230824402220207e-05,
+      "loss": 0.9922,
+      "step": 18113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1229978115065352e-05,
+      "loss": 1.073,
+      "step": 18114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1229131818965962e-05,
+      "loss": 0.7888,
+      "step": 18115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1228285513928187e-05,
+      "loss": 0.8609,
+      "step": 18116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1227439199958186e-05,
+      "loss": 0.7311,
+      "step": 18117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.122659287706211e-05,
+      "loss": 0.8513,
+      "step": 18118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.122574654524611e-05,
+      "loss": 0.9812,
+      "step": 18119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.122490020451635e-05,
+      "loss": 0.8497,
+      "step": 18120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1224053854878977e-05,
+      "loss": 0.8503,
+      "step": 18121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1223207496340152e-05,
+      "loss": 1.3892,
+      "step": 18122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1222361128906024e-05,
+      "loss": 0.8418,
+      "step": 18123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1221514752582749e-05,
+      "loss": 0.897,
+      "step": 18124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1220668367376483e-05,
+      "loss": 0.7724,
+      "step": 18125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1219821973293384e-05,
+      "loss": 0.8468,
+      "step": 18126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1218975570339603e-05,
+      "loss": 0.5455,
+      "step": 18127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1218129158521294e-05,
+      "loss": 0.8944,
+      "step": 18128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1217282737844618e-05,
+      "loss": 0.9032,
+      "step": 18129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1216436308315724e-05,
+      "loss": 0.931,
+      "step": 18130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121558986994077e-05,
+      "loss": 0.9483,
+      "step": 18131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1214743422725913e-05,
+      "loss": 1.1177,
+      "step": 18132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1213896966677307e-05,
+      "loss": 0.8525,
+      "step": 18133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1213050501801104e-05,
+      "loss": 0.9131,
+      "step": 18134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1212204028103469e-05,
+      "loss": 0.9163,
+      "step": 18135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1211357545590548e-05,
+      "loss": 0.7952,
+      "step": 18136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.12105110542685e-05,
+      "loss": 0.9692,
+      "step": 18137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1209664554143483e-05,
+      "loss": 0.9418,
+      "step": 18138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.120881804522165e-05,
+      "loss": 1.0872,
+      "step": 18139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1207971527509159e-05,
+      "loss": 0.7516,
+      "step": 18140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1207125001012162e-05,
+      "loss": 0.9945,
+      "step": 18141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1206278465736822e-05,
+      "loss": 0.8175,
+      "step": 18142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1205431921689287e-05,
+      "loss": 0.8193,
+      "step": 18143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.120458536887572e-05,
+      "loss": 0.6796,
+      "step": 18144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1203738807302275e-05,
+      "loss": 1.2722,
+      "step": 18145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1202892236975104e-05,
+      "loss": 0.8318,
+      "step": 18146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1202045657900372e-05,
+      "loss": 0.9088,
+      "step": 18147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1201199070084225e-05,
+      "loss": 0.5846,
+      "step": 18148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1200352473532828e-05,
+      "loss": 0.7311,
+      "step": 18149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1199505868252336e-05,
+      "loss": 0.959,
+      "step": 18150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1198659254248902e-05,
+      "loss": 0.9506,
+      "step": 18151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1197812631528684e-05,
+      "loss": 0.9491,
+      "step": 18152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1196966000097842e-05,
+      "loss": 0.9414,
+      "step": 18153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1196119359962529e-05,
+      "loss": 0.8516,
+      "step": 18154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1195272711128904e-05,
+      "loss": 0.9631,
+      "step": 18155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1194426053603122e-05,
+      "loss": 1.0986,
+      "step": 18156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1193579387391342e-05,
+      "loss": 0.8542,
+      "step": 18157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1192732712499719e-05,
+      "loss": 1.0142,
+      "step": 18158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.119188602893441e-05,
+      "loss": 0.9438,
+      "step": 18159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1191039336701578e-05,
+      "loss": 0.9637,
+      "step": 18160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1190192635807374e-05,
+      "loss": 0.8837,
+      "step": 18161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1189345926257956e-05,
+      "loss": 0.6158,
+      "step": 18162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1188499208059483e-05,
+      "loss": 0.7473,
+      "step": 18163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1187652481218108e-05,
+      "loss": 0.7026,
+      "step": 18164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1186805745739999e-05,
+      "loss": 0.7885,
+      "step": 18165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1185959001631303e-05,
+      "loss": 0.9034,
+      "step": 18166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118511224889818e-05,
+      "loss": 0.9616,
+      "step": 18167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118426548754679e-05,
+      "loss": 1.0674,
+      "step": 18168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1183418717583292e-05,
+      "loss": 1.0144,
+      "step": 18169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1182571939013842e-05,
+      "loss": 0.6566,
+      "step": 18170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1181725151844596e-05,
+      "loss": 0.8009,
+      "step": 18171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180878356081712e-05,
+      "loss": 1.1963,
+      "step": 18172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1180031551731349e-05,
+      "loss": 0.8615,
+      "step": 18173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117918473879967e-05,
+      "loss": 0.9841,
+      "step": 18174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1178337917292828e-05,
+      "loss": 0.738,
+      "step": 18175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1177491087216976e-05,
+      "loss": 1.1121,
+      "step": 18176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1176644248578284e-05,
+      "loss": 1.0122,
+      "step": 18177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1175797401382902e-05,
+      "loss": 0.7185,
+      "step": 18178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1174950545636992e-05,
+      "loss": 0.5807,
+      "step": 18179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1174103681346711e-05,
+      "loss": 0.9045,
+      "step": 18180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1173256808518215e-05,
+      "loss": 0.7697,
+      "step": 18181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1172409927157666e-05,
+      "loss": 0.8678,
+      "step": 18182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1171563037271222e-05,
+      "loss": 0.7684,
+      "step": 18183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1170716138865045e-05,
+      "loss": 0.899,
+      "step": 18184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1169869231945287e-05,
+      "loss": 0.9479,
+      "step": 18185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.116902231651811e-05,
+      "loss": 1.0466,
+      "step": 18186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1168175392589675e-05,
+      "loss": 0.9574,
+      "step": 18187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1167328460166136e-05,
+      "loss": 0.7296,
+      "step": 18188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1166481519253659e-05,
+      "loss": 1.1899,
+      "step": 18189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1165634569858395e-05,
+      "loss": 0.7882,
+      "step": 18190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1164787611986505e-05,
+      "loss": 0.8397,
+      "step": 18191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1163940645644153e-05,
+      "loss": 1.1072,
+      "step": 18192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1163093670837496e-05,
+      "loss": 1.1032,
+      "step": 18193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1162246687572693e-05,
+      "loss": 1.0586,
+      "step": 18194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1161399695855903e-05,
+      "loss": 1.1624,
+      "step": 18195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1160552695693286e-05,
+      "loss": 0.7351,
+      "step": 18196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1159705687090998e-05,
+      "loss": 0.783,
+      "step": 18197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1158858670055202e-05,
+      "loss": 0.6944,
+      "step": 18198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115801164459206e-05,
+      "loss": 0.8607,
+      "step": 18199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1157164610707725e-05,
+      "loss": 0.8059,
+      "step": 18200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1156317568408361e-05,
+      "loss": 0.97,
+      "step": 18201
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115547051770013e-05,
+      "loss": 1.1529,
+      "step": 18202
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1154623458589189e-05,
+      "loss": 1.0417,
+      "step": 18203
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1153776391081696e-05,
+      "loss": 1.1366,
+      "step": 18204
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1152929315183813e-05,
+      "loss": 0.8677,
+      "step": 18205
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1152082230901702e-05,
+      "loss": 0.8332,
+      "step": 18206
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1151235138241516e-05,
+      "loss": 0.778,
+      "step": 18207
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1150388037209425e-05,
+      "loss": 0.8653,
+      "step": 18208
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1149540927811584e-05,
+      "loss": 0.7534,
+      "step": 18209
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1148693810054152e-05,
+      "loss": 0.8541,
+      "step": 18210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1147846683943293e-05,
+      "loss": 0.8738,
+      "step": 18211
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146999549485165e-05,
+      "loss": 0.9483,
+      "step": 18212
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146152406685926e-05,
+      "loss": 1.1524,
+      "step": 18213
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1145305255551744e-05,
+      "loss": 0.9658,
+      "step": 18214
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1144458096088773e-05,
+      "loss": 1.3172,
+      "step": 18215
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1143610928303175e-05,
+      "loss": 1.1405,
+      "step": 18216
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1142763752201112e-05,
+      "loss": 0.989,
+      "step": 18217
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1141916567788744e-05,
+      "loss": 0.9655,
+      "step": 18218
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1141069375072231e-05,
+      "loss": 0.7989,
+      "step": 18219
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1140222174057736e-05,
+      "loss": 0.8643,
+      "step": 18220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1139374964751419e-05,
+      "loss": 0.8465,
+      "step": 18221
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1138527747159437e-05,
+      "loss": 0.8436,
+      "step": 18222
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1137680521287959e-05,
+      "loss": 0.744,
+      "step": 18223
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1136833287143143e-05,
+      "loss": 0.938,
+      "step": 18224
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1135986044731144e-05,
+      "loss": 0.9189,
+      "step": 18225
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1135138794058133e-05,
+      "loss": 0.7405,
+      "step": 18226
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1134291535130264e-05,
+      "loss": 1.0914,
+      "step": 18227
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1133444267953702e-05,
+      "loss": 0.8667,
+      "step": 18228
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1132596992534608e-05,
+      "loss": 1.1108,
+      "step": 18229
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1131749708879142e-05,
+      "loss": 0.7926,
+      "step": 18230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130902416993465e-05,
+      "loss": 1.0509,
+      "step": 18231
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130055116883742e-05,
+      "loss": 0.7907,
+      "step": 18232
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1129207808556133e-05,
+      "loss": 1.079,
+      "step": 18233
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11283604920168e-05,
+      "loss": 0.9072,
+      "step": 18234
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1127513167271902e-05,
+      "loss": 0.8557,
+      "step": 18235
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1126665834327603e-05,
+      "loss": 0.9691,
+      "step": 18236
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1125818493190065e-05,
+      "loss": 0.9178,
+      "step": 18237
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1124971143865451e-05,
+      "loss": 0.7969,
+      "step": 18238
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1124123786359923e-05,
+      "loss": 0.6884,
+      "step": 18239
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112327642067964e-05,
+      "loss": 0.7166,
+      "step": 18240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1122429046830768e-05,
+      "loss": 0.949,
+      "step": 18241
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1121581664819465e-05,
+      "loss": 0.8523,
+      "step": 18242
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1120734274651897e-05,
+      "loss": 0.7914,
+      "step": 18243
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1119886876334224e-05,
+      "loss": 1.1222,
+      "step": 18244
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1119039469872613e-05,
+      "loss": 0.9476,
+      "step": 18245
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1118192055273216e-05,
+      "loss": 0.8304,
+      "step": 18246
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1117344632542206e-05,
+      "loss": 1.1754,
+      "step": 18247
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1116497201685742e-05,
+      "loss": 1.0692,
+      "step": 18248
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1115649762709989e-05,
+      "loss": 1.1262,
+      "step": 18249
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1114802315621103e-05,
+      "loss": 1.0237,
+      "step": 18250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1113954860425251e-05,
+      "loss": 0.7596,
+      "step": 18251
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1113107397128595e-05,
+      "loss": 0.8256,
+      "step": 18252
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1112259925737299e-05,
+      "loss": 0.7676,
+      "step": 18253
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111141244625753e-05,
+      "loss": 0.9952,
+      "step": 18254
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111056495869544e-05,
+      "loss": 0.8699,
+      "step": 18255
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109717463057201e-05,
+      "loss": 0.9704,
+      "step": 18256
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1108869959348972e-05,
+      "loss": 1.1801,
+      "step": 18257
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1108022447576918e-05,
+      "loss": 0.8297,
+      "step": 18258
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1107174927747205e-05,
+      "loss": 0.9567,
+      "step": 18259
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1106327399865989e-05,
+      "loss": 0.8944,
+      "step": 18260
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1105479863939437e-05,
+      "loss": 0.9499,
+      "step": 18261
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1104632319973714e-05,
+      "loss": 0.9158,
+      "step": 18262
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1103784767974983e-05,
+      "loss": 1.2969,
+      "step": 18263
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1102937207949405e-05,
+      "loss": 0.9796,
+      "step": 18264
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1102089639903146e-05,
+      "loss": 0.904,
+      "step": 18265
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.110124206384237e-05,
+      "loss": 0.8936,
+      "step": 18266
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1100394479773237e-05,
+      "loss": 0.7765,
+      "step": 18267
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1099546887701913e-05,
+      "loss": 0.9188,
+      "step": 18268
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1098699287634567e-05,
+      "loss": 0.9113,
+      "step": 18269
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1097851679577351e-05,
+      "loss": 0.6671,
+      "step": 18270
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1097004063536438e-05,
+      "loss": 0.9328,
+      "step": 18271
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1096156439517992e-05,
+      "loss": 0.9706,
+      "step": 18272
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1095308807528173e-05,
+      "loss": 0.668,
+      "step": 18273
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1094461167573148e-05,
+      "loss": 0.9037,
+      "step": 18274
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109361351965908e-05,
+      "loss": 0.8816,
+      "step": 18275
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109276586379213e-05,
+      "loss": 1.133,
+      "step": 18276
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1091918199978464e-05,
+      "loss": 0.5422,
+      "step": 18277
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1091070528224253e-05,
+      "loss": 0.8423,
+      "step": 18278
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1090222848535653e-05,
+      "loss": 1.262,
+      "step": 18279
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1089375160918832e-05,
+      "loss": 0.9081,
+      "step": 18280
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1088527465379952e-05,
+      "loss": 0.7183,
+      "step": 18281
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1087679761925182e-05,
+      "loss": 0.7792,
+      "step": 18282
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108683205056068e-05,
+      "loss": 0.8076,
+      "step": 18283
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108598433129262e-05,
+      "loss": 0.8994,
+      "step": 18284
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085136604127161e-05,
+      "loss": 0.7063,
+      "step": 18285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1084288869070465e-05,
+      "loss": 0.8302,
+      "step": 18286
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10834411261287e-05,
+      "loss": 1.0044,
+      "step": 18287
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1082593375308033e-05,
+      "loss": 0.7932,
+      "step": 18288
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1081745616614626e-05,
+      "loss": 0.8882,
+      "step": 18289
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1080897850054643e-05,
+      "loss": 0.8936,
+      "step": 18290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1080050075634254e-05,
+      "loss": 0.9713,
+      "step": 18291
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1079202293359617e-05,
+      "loss": 1.115,
+      "step": 18292
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1078354503236904e-05,
+      "loss": 1.0719,
+      "step": 18293
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1077506705272278e-05,
+      "loss": 0.9424,
+      "step": 18294
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1076658899471902e-05,
+      "loss": 0.3487,
+      "step": 18295
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107581108584194e-05,
+      "loss": 0.8694,
+      "step": 18296
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1074963264388565e-05,
+      "loss": 0.9501,
+      "step": 18297
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1074115435117934e-05,
+      "loss": 1.1051,
+      "step": 18298
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107326759803622e-05,
+      "loss": 0.9799,
+      "step": 18299
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1072419753149585e-05,
+      "loss": 0.8525,
+      "step": 18300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107157190046419e-05,
+      "loss": 0.6473,
+      "step": 18301
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1070724039986209e-05,
+      "loss": 0.8394,
+      "step": 18302
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1069876171721803e-05,
+      "loss": 1.0568,
+      "step": 18303
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1069028295677139e-05,
+      "loss": 0.9872,
+      "step": 18304
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1068180411858382e-05,
+      "loss": 0.6219,
+      "step": 18305
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1067332520271698e-05,
+      "loss": 0.677,
+      "step": 18306
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1066484620923253e-05,
+      "loss": 0.9995,
+      "step": 18307
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1065636713819213e-05,
+      "loss": 0.8282,
+      "step": 18308
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1064788798965749e-05,
+      "loss": 1.0426,
+      "step": 18309
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1063940876369017e-05,
+      "loss": 1.1522,
+      "step": 18310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106309294603519e-05,
+      "loss": 1.3372,
+      "step": 18311
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1062245007970436e-05,
+      "loss": 0.7686,
+      "step": 18312
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1061397062180916e-05,
+      "loss": 0.89,
+      "step": 18313
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10605491086728e-05,
+      "loss": 0.6795,
+      "step": 18314
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.105970114745225e-05,
+      "loss": 0.7753,
+      "step": 18315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1058853178525438e-05,
+      "loss": 1.0698,
+      "step": 18316
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1058005201898528e-05,
+      "loss": 0.9526,
+      "step": 18317
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1057157217577687e-05,
+      "loss": 0.9001,
+      "step": 18318
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.105630922556908e-05,
+      "loss": 1.1714,
+      "step": 18319
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1055461225878877e-05,
+      "loss": 0.8543,
+      "step": 18320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1054613218513241e-05,
+      "loss": 0.8906,
+      "step": 18321
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1053765203478342e-05,
+      "loss": 1.1913,
+      "step": 18322
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1052917180780343e-05,
+      "loss": 1.0497,
+      "step": 18323
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1052069150425417e-05,
+      "loss": 1.2317,
+      "step": 18324
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1051221112419723e-05,
+      "loss": 1.0188,
+      "step": 18325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1050373066769435e-05,
+      "loss": 0.9668,
+      "step": 18326
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1049525013480717e-05,
+      "loss": 1.6645,
+      "step": 18327
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1048676952559738e-05,
+      "loss": 0.9972,
+      "step": 18328
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1047828884012664e-05,
+      "loss": 1.1114,
+      "step": 18329
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104698080784566e-05,
+      "loss": 0.8985,
+      "step": 18330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1046132724064895e-05,
+      "loss": 0.9501,
+      "step": 18331
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1045284632676535e-05,
+      "loss": 1.0405,
+      "step": 18332
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1044436533686751e-05,
+      "loss": 1.0379,
+      "step": 18333
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1043588427101711e-05,
+      "loss": 0.7167,
+      "step": 18334
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1042740312927577e-05,
+      "loss": 1.1133,
+      "step": 18335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1041892191170522e-05,
+      "loss": 0.7479,
+      "step": 18336
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104104406183671e-05,
+      "loss": 0.8429,
+      "step": 18337
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1040195924932307e-05,
+      "loss": 0.7488,
+      "step": 18338
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103934778046349e-05,
+      "loss": 1.0688,
+      "step": 18339
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1038499628436417e-05,
+      "loss": 1.0324,
+      "step": 18340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1037651468857258e-05,
+      "loss": 0.9159,
+      "step": 18341
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1036803301732185e-05,
+      "loss": 0.7338,
+      "step": 18342
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1035955127067362e-05,
+      "loss": 1.4309,
+      "step": 18343
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1035106944868958e-05,
+      "loss": 0.8295,
+      "step": 18344
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1034258755143141e-05,
+      "loss": 1.2173,
+      "step": 18345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103341055789608e-05,
+      "loss": 0.9737,
+      "step": 18346
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1032562353133942e-05,
+      "loss": 0.8044,
+      "step": 18347
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1031714140862897e-05,
+      "loss": 1.2204,
+      "step": 18348
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1030865921089116e-05,
+      "loss": 0.6876,
+      "step": 18349
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1030017693818756e-05,
+      "loss": 0.902,
+      "step": 18350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1029169459057998e-05,
+      "loss": 0.8558,
+      "step": 18351
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1028321216813002e-05,
+      "loss": 0.9466,
+      "step": 18352
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1027472967089941e-05,
+      "loss": 0.7194,
+      "step": 18353
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1026624709894987e-05,
+      "loss": 1.0038,
+      "step": 18354
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10257764452343e-05,
+      "loss": 1.0425,
+      "step": 18355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1024928173114051e-05,
+      "loss": 1.0983,
+      "step": 18356
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1024079893540412e-05,
+      "loss": 0.8295,
+      "step": 18357
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102323160651955e-05,
+      "loss": 1.0616,
+      "step": 18358
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1022383312057636e-05,
+      "loss": 0.7954,
+      "step": 18359
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1021535010160838e-05,
+      "loss": 1.036,
+      "step": 18360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102068670083532e-05,
+      "loss": 1.2351,
+      "step": 18361
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1019838384087256e-05,
+      "loss": 0.9033,
+      "step": 18362
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1018990059922816e-05,
+      "loss": 1.0434,
+      "step": 18363
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1018141728348169e-05,
+      "loss": 1.1663,
+      "step": 18364
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1017293389369477e-05,
+      "loss": 1.1551,
+      "step": 18365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1016445042992919e-05,
+      "loss": 0.5848,
+      "step": 18366
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1015596689224657e-05,
+      "loss": 1.0545,
+      "step": 18367
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101474832807086e-05,
+      "loss": 0.9441,
+      "step": 18368
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1013899959537708e-05,
+      "loss": 0.8999,
+      "step": 18369
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101305158363136e-05,
+      "loss": 0.6145,
+      "step": 18370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1012203200357984e-05,
+      "loss": 0.7862,
+      "step": 18371
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101135480972376e-05,
+      "loss": 0.7518,
+      "step": 18372
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1010506411734849e-05,
+      "loss": 0.8394,
+      "step": 18373
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1009658006397424e-05,
+      "loss": 0.6547,
+      "step": 18374
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1008809593717653e-05,
+      "loss": 0.9137,
+      "step": 18375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1007961173701707e-05,
+      "loss": 0.956,
+      "step": 18376
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1007112746355752e-05,
+      "loss": 1.0263,
+      "step": 18377
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1006264311685966e-05,
+      "loss": 0.7591,
+      "step": 18378
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1005415869698515e-05,
+      "loss": 1.0143,
+      "step": 18379
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1004567420399563e-05,
+      "loss": 1.348,
+      "step": 18380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1003718963795289e-05,
+      "loss": 1.0744,
+      "step": 18381
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1002870499891858e-05,
+      "loss": 0.7383,
+      "step": 18382
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1002022028695441e-05,
+      "loss": 0.8311,
+      "step": 18383
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.100117355021221e-05,
+      "loss": 0.8272,
+      "step": 18384
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1000325064448332e-05,
+      "loss": 1.2487,
+      "step": 18385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099947657140998e-05,
+      "loss": 0.679,
+      "step": 18386
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0998628071103322e-05,
+      "loss": 0.9502,
+      "step": 18387
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0997779563534532e-05,
+      "loss": 0.8659,
+      "step": 18388
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0996931048709776e-05,
+      "loss": 0.9796,
+      "step": 18389
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0996082526635227e-05,
+      "loss": 0.9573,
+      "step": 18390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0995233997317058e-05,
+      "loss": 1.2536,
+      "step": 18391
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0994385460761437e-05,
+      "loss": 1.0055,
+      "step": 18392
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099353691697453e-05,
+      "loss": 0.8346,
+      "step": 18393
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0992688365962516e-05,
+      "loss": 1.0987,
+      "step": 18394
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0991839807731559e-05,
+      "loss": 0.956,
+      "step": 18395
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0990991242287834e-05,
+      "loss": 0.8902,
+      "step": 18396
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099014266963751e-05,
+      "loss": 0.7867,
+      "step": 18397
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0989294089786761e-05,
+      "loss": 0.96,
+      "step": 18398
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0988445502741755e-05,
+      "loss": 1.0817,
+      "step": 18399
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0987596908508663e-05,
+      "loss": 0.7917,
+      "step": 18400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0986748307093657e-05,
+      "loss": 0.7433,
+      "step": 18401
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0985899698502904e-05,
+      "loss": 0.9894,
+      "step": 18402
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0985051082742583e-05,
+      "loss": 1.1511,
+      "step": 18403
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.098420245981886e-05,
+      "loss": 0.8682,
+      "step": 18404
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0983353829737909e-05,
+      "loss": 1.0729,
+      "step": 18405
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0982505192505897e-05,
+      "loss": 1.1561,
+      "step": 18406
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0981656548129001e-05,
+      "loss": 1.0089,
+      "step": 18407
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0980807896613386e-05,
+      "loss": 0.8124,
+      "step": 18408
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979959237965229e-05,
+      "loss": 1.2135,
+      "step": 18409
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.09791105721907e-05,
+      "loss": 0.881,
+      "step": 18410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.097826189929597e-05,
+      "loss": 0.9784,
+      "step": 18411
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.097741321928721e-05,
+      "loss": 0.9882,
+      "step": 18412
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0976564532170595e-05,
+      "loss": 0.9336,
+      "step": 18413
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0975715837952292e-05,
+      "loss": 1.0281,
+      "step": 18414
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0974867136638476e-05,
+      "loss": 0.9874,
+      "step": 18415
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0974018428235317e-05,
+      "loss": 0.8704,
+      "step": 18416
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0973169712748986e-05,
+      "loss": 0.9133,
+      "step": 18417
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.097232099018566e-05,
+      "loss": 0.9011,
+      "step": 18418
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.097147226055151e-05,
+      "loss": 0.8895,
+      "step": 18419
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0970623523852699e-05,
+      "loss": 0.8195,
+      "step": 18420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.096977478009541e-05,
+      "loss": 0.8599,
+      "step": 18421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.096892602928581e-05,
+      "loss": 1.0641,
+      "step": 18422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0968077271430073e-05,
+      "loss": 0.8825,
+      "step": 18423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0967228506534373e-05,
+      "loss": 1.1246,
+      "step": 18424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0966379734604877e-05,
+      "loss": 0.9435,
+      "step": 18425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0965530955647761e-05,
+      "loss": 0.8511,
+      "step": 18426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0964682169669196e-05,
+      "loss": 0.8948,
+      "step": 18427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0963833376675357e-05,
+      "loss": 1.1195,
+      "step": 18428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962984576672414e-05,
+      "loss": 0.9163,
+      "step": 18429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962135769666537e-05,
+      "loss": 0.739,
+      "step": 18430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0961286955663906e-05,
+      "loss": 0.789,
+      "step": 18431
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0960438134670686e-05,
+      "loss": 0.7599,
+      "step": 18432
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0959589306693055e-05,
+      "loss": 0.8875,
+      "step": 18433
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0958740471737185e-05,
+      "loss": 0.5926,
+      "step": 18434
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0957891629809248e-05,
+      "loss": 0.7191,
+      "step": 18435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0957042780915414e-05,
+      "loss": 1.1163,
+      "step": 18436
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.095619392506186e-05,
+      "loss": 1.0621,
+      "step": 18437
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0955345062254756e-05,
+      "loss": 0.9516,
+      "step": 18438
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0954496192500281e-05,
+      "loss": 0.8483,
+      "step": 18439
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0953647315804601e-05,
+      "loss": 1.0212,
+      "step": 18440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.095279843217389e-05,
+      "loss": 0.6671,
+      "step": 18441
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0951949541614327e-05,
+      "loss": 0.8694,
+      "step": 18442
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0951100644132077e-05,
+      "loss": 0.7948,
+      "step": 18443
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0950251739733323e-05,
+      "loss": 0.8462,
+      "step": 18444
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0949402828424227e-05,
+      "loss": 0.8477,
+      "step": 18445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0948553910210973e-05,
+      "loss": 0.8742,
+      "step": 18446
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0947704985099725e-05,
+      "loss": 0.9356,
+      "step": 18447
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0946856053096663e-05,
+      "loss": 1.1509,
+      "step": 18448
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0946007114207962e-05,
+      "loss": 0.8844,
+      "step": 18449
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0945158168439786e-05,
+      "loss": 0.5606,
+      "step": 18450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0944309215798319e-05,
+      "loss": 0.9387,
+      "step": 18451
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.094346025628973e-05,
+      "loss": 0.8346,
+      "step": 18452
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0942611289920194e-05,
+      "loss": 0.461,
+      "step": 18453
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0941762316695883e-05,
+      "loss": 0.8416,
+      "step": 18454
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0940913336622971e-05,
+      "loss": 0.6642,
+      "step": 18455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0940064349707632e-05,
+      "loss": 1.1073,
+      "step": 18456
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0939215355956043e-05,
+      "loss": 0.7421,
+      "step": 18457
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0938366355374374e-05,
+      "loss": 1.0037,
+      "step": 18458
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0937517347968802e-05,
+      "loss": 0.9138,
+      "step": 18459
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09366683337455e-05,
+      "loss": 0.9883,
+      "step": 18460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093581931271064e-05,
+      "loss": 0.9212,
+      "step": 18461
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934970284870401e-05,
+      "loss": 1.0355,
+      "step": 18462
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093412125023095e-05,
+      "loss": 0.6706,
+      "step": 18463
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093327220879847e-05,
+      "loss": 0.9871,
+      "step": 18464
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0932423160579126e-05,
+      "loss": 0.8942,
+      "step": 18465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0931574105579102e-05,
+      "loss": 1.0389,
+      "step": 18466
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0930725043804565e-05,
+      "loss": 0.879,
+      "step": 18467
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0929875975261692e-05,
+      "loss": 0.9057,
+      "step": 18468
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0929026899956661e-05,
+      "loss": 0.8458,
+      "step": 18469
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092817781789564e-05,
+      "loss": 1.0208,
+      "step": 18470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0927328729084807e-05,
+      "loss": 0.9713,
+      "step": 18471
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0926479633530336e-05,
+      "loss": 0.7437,
+      "step": 18472
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0925630531238403e-05,
+      "loss": 1.2469,
+      "step": 18473
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0924781422215183e-05,
+      "loss": 0.922,
+      "step": 18474
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923932306466849e-05,
+      "loss": 0.9132,
+      "step": 18475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923083183999577e-05,
+      "loss": 1.0149,
+      "step": 18476
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092223405481954e-05,
+      "loss": 0.8902,
+      "step": 18477
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0921384918932915e-05,
+      "loss": 0.9776,
+      "step": 18478
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092053577634588e-05,
+      "loss": 1.3109,
+      "step": 18479
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0919686627064603e-05,
+      "loss": 0.7568,
+      "step": 18480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0918837471095262e-05,
+      "loss": 0.6337,
+      "step": 18481
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0917988308444036e-05,
+      "loss": 0.7473,
+      "step": 18482
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0917139139117094e-05,
+      "loss": 1.0595,
+      "step": 18483
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0916289963120616e-05,
+      "loss": 0.7123,
+      "step": 18484
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0915440780460777e-05,
+      "loss": 0.9516,
+      "step": 18485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0914591591143751e-05,
+      "loss": 0.9415,
+      "step": 18486
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0913742395175708e-05,
+      "loss": 0.9847,
+      "step": 18487
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0912893192562835e-05,
+      "loss": 1.0104,
+      "step": 18488
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09120439833113e-05,
+      "loss": 0.8801,
+      "step": 18489
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0911194767427278e-05,
+      "loss": 1.143,
+      "step": 18490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910345544916947e-05,
+      "loss": 0.9094,
+      "step": 18491
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0909496315786483e-05,
+      "loss": 1.0432,
+      "step": 18492
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0908647080042058e-05,
+      "loss": 0.857,
+      "step": 18493
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0907797837689855e-05,
+      "loss": 0.835,
+      "step": 18494
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0906948588736044e-05,
+      "loss": 0.8628,
+      "step": 18495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09060993331868e-05,
+      "loss": 0.7732,
+      "step": 18496
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0905250071048304e-05,
+      "loss": 1.0182,
+      "step": 18497
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0904400802326726e-05,
+      "loss": 0.8558,
+      "step": 18498
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0903551527028246e-05,
+      "loss": 1.007,
+      "step": 18499
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.090270224515904e-05,
+      "loss": 0.9286,
+      "step": 18500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0901852956725283e-05,
+      "loss": 0.7265,
+      "step": 18501
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0901003661733148e-05,
+      "loss": 0.9285,
+      "step": 18502
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0900154360188818e-05,
+      "loss": 0.6929,
+      "step": 18503
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0899305052098465e-05,
+      "loss": 0.7562,
+      "step": 18504
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0898455737468262e-05,
+      "loss": 0.9406,
+      "step": 18505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0897606416304393e-05,
+      "loss": 0.9324,
+      "step": 18506
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.089675708861303e-05,
+      "loss": 0.9358,
+      "step": 18507
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0895907754400348e-05,
+      "loss": 0.8441,
+      "step": 18508
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.089505841367253e-05,
+      "loss": 0.95,
+      "step": 18509
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0894209066435746e-05,
+      "loss": 0.552,
+      "step": 18510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.089335971269617e-05,
+      "loss": 1.0273,
+      "step": 18511
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0892510352459986e-05,
+      "loss": 0.9159,
+      "step": 18512
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0891660985733368e-05,
+      "loss": 1.0587,
+      "step": 18513
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0890811612522492e-05,
+      "loss": 1.0443,
+      "step": 18514
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889962232833535e-05,
+      "loss": 1.0452,
+      "step": 18515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889112846672674e-05,
+      "loss": 1.0773,
+      "step": 18516
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0888263454046085e-05,
+      "loss": 0.8931,
+      "step": 18517
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0887414054959945e-05,
+      "loss": 0.9038,
+      "step": 18518
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0886564649420435e-05,
+      "loss": 0.8726,
+      "step": 18519
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0885715237433722e-05,
+      "loss": 0.9058,
+      "step": 18520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0884865819005995e-05,
+      "loss": 1.1685,
+      "step": 18521
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0884016394143424e-05,
+      "loss": 1.3125,
+      "step": 18522
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0883166962852187e-05,
+      "loss": 0.752,
+      "step": 18523
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0882317525138464e-05,
+      "loss": 0.7965,
+      "step": 18524
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0881468081008428e-05,
+      "loss": 0.8427,
+      "step": 18525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0880618630468256e-05,
+      "loss": 0.6675,
+      "step": 18526
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087976917352413e-05,
+      "loss": 0.7089,
+      "step": 18527
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0878919710182225e-05,
+      "loss": 0.9849,
+      "step": 18528
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0878070240448716e-05,
+      "loss": 0.7032,
+      "step": 18529
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0877220764329785e-05,
+      "loss": 0.7742,
+      "step": 18530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0876371281831608e-05,
+      "loss": 0.7413,
+      "step": 18531
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0875521792960361e-05,
+      "loss": 0.9575,
+      "step": 18532
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087467229772222e-05,
+      "loss": 1.0343,
+      "step": 18533
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087382279612337e-05,
+      "loss": 0.9175,
+      "step": 18534
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0872973288169979e-05,
+      "loss": 1.1914,
+      "step": 18535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087212377386823e-05,
+      "loss": 0.5494,
+      "step": 18536
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0871274253224302e-05,
+      "loss": 0.9655,
+      "step": 18537
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087042472624437e-05,
+      "loss": 1.2015,
+      "step": 18538
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0869575192934612e-05,
+      "loss": 0.8083,
+      "step": 18539
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0868725653301206e-05,
+      "loss": 0.7,
+      "step": 18540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0867876107350332e-05,
+      "loss": 0.7551,
+      "step": 18541
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0867026555088165e-05,
+      "loss": 1.0926,
+      "step": 18542
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0866176996520885e-05,
+      "loss": 0.801,
+      "step": 18543
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0865327431654673e-05,
+      "loss": 0.6858,
+      "step": 18544
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0864477860495702e-05,
+      "loss": 0.8898,
+      "step": 18545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086362828305015e-05,
+      "loss": 0.883,
+      "step": 18546
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08627786993242e-05,
+      "loss": 0.6665,
+      "step": 18547
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861929109324023e-05,
+      "loss": 0.9834,
+      "step": 18548
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861079513055807e-05,
+      "loss": 1.0189,
+      "step": 18549
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0860229910525723e-05,
+      "loss": 0.8902,
+      "step": 18550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085938030173995e-05,
+      "loss": 0.9547,
+      "step": 18551
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0858530686704669e-05,
+      "loss": 0.7856,
+      "step": 18552
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0857681065426057e-05,
+      "loss": 0.887,
+      "step": 18553
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0856831437910297e-05,
+      "loss": 0.7064,
+      "step": 18554
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085598180416356e-05,
+      "loss": 0.7967,
+      "step": 18555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0855132164192029e-05,
+      "loss": 0.9195,
+      "step": 18556
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085428251800188e-05,
+      "loss": 0.8112,
+      "step": 18557
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0853432865599296e-05,
+      "loss": 0.9947,
+      "step": 18558
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0852583206990455e-05,
+      "loss": 1.0528,
+      "step": 18559
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.085173354218153e-05,
+      "loss": 0.7456,
+      "step": 18560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0850883871178707e-05,
+      "loss": 0.9801,
+      "step": 18561
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0850034193988162e-05,
+      "loss": 0.7985,
+      "step": 18562
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0849184510616071e-05,
+      "loss": 0.7979,
+      "step": 18563
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0848334821068623e-05,
+      "loss": 0.7973,
+      "step": 18564
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0847485125351984e-05,
+      "loss": 0.8367,
+      "step": 18565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0846635423472338e-05,
+      "loss": 0.903,
+      "step": 18566
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0845785715435868e-05,
+      "loss": 1.1631,
+      "step": 18567
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0844936001248749e-05,
+      "loss": 1.0084,
+      "step": 18568
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0844086280917162e-05,
+      "loss": 0.8213,
+      "step": 18569
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0843236554447288e-05,
+      "loss": 0.9836,
+      "step": 18570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0842386821845303e-05,
+      "loss": 0.8418,
+      "step": 18571
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0841537083117384e-05,
+      "loss": 0.7042,
+      "step": 18572
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0840687338269718e-05,
+      "loss": 0.9073,
+      "step": 18573
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0839837587308482e-05,
+      "loss": 1.1668,
+      "step": 18574
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0838987830239848e-05,
+      "loss": 1.1321,
+      "step": 18575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0838138067070006e-05,
+      "loss": 0.6561,
+      "step": 18576
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0837288297805128e-05,
+      "loss": 0.885,
+      "step": 18577
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0836438522451398e-05,
+      "loss": 1.0646,
+      "step": 18578
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0835588741014992e-05,
+      "loss": 0.8726,
+      "step": 18579
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0834738953502096e-05,
+      "loss": 0.8056,
+      "step": 18580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.083388915991888e-05,
+      "loss": 1.0339,
+      "step": 18581
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833039360271533e-05,
+      "loss": 1.1817,
+      "step": 18582
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0832189554566232e-05,
+      "loss": 0.8993,
+      "step": 18583
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0831339742809156e-05,
+      "loss": 0.8872,
+      "step": 18584
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0830489925006485e-05,
+      "loss": 0.9678,
+      "step": 18585
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0829640101164397e-05,
+      "loss": 0.8607,
+      "step": 18586
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828790271289076e-05,
+      "loss": 1.0396,
+      "step": 18587
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0827940435386696e-05,
+      "loss": 1.5033,
+      "step": 18588
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0827090593463449e-05,
+      "loss": 0.6433,
+      "step": 18589
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08262407455255e-05,
+      "loss": 1.0747,
+      "step": 18590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.082539089157904e-05,
+      "loss": 0.9764,
+      "step": 18591
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0824541031630246e-05,
+      "loss": 0.9294,
+      "step": 18592
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0823691165685297e-05,
+      "loss": 0.9219,
+      "step": 18593
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0822841293750377e-05,
+      "loss": 0.8061,
+      "step": 18594
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0821991415831662e-05,
+      "loss": 0.9363,
+      "step": 18595
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0821141531935333e-05,
+      "loss": 1.1296,
+      "step": 18596
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0820291642067572e-05,
+      "loss": 0.7647,
+      "step": 18597
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0819441746234558e-05,
+      "loss": 1.1651,
+      "step": 18598
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0818591844442474e-05,
+      "loss": 1.0892,
+      "step": 18599
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0817741936697499e-05,
+      "loss": 0.7178,
+      "step": 18600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0816892023005814e-05,
+      "loss": 0.9033,
+      "step": 18601
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0816042103373601e-05,
+      "loss": 0.9453,
+      "step": 18602
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0815192177807034e-05,
+      "loss": 1.1624,
+      "step": 18603
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0814342246312306e-05,
+      "loss": 0.6326,
+      "step": 18604
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0813492308895588e-05,
+      "loss": 1.033,
+      "step": 18605
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0812642365563059e-05,
+      "loss": 0.8293,
+      "step": 18606
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081179241632091e-05,
+      "loss": 0.7659,
+      "step": 18607
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0810942461175316e-05,
+      "loss": 0.7736,
+      "step": 18608
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0810092500132457e-05,
+      "loss": 0.8725,
+      "step": 18609
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0809242533198516e-05,
+      "loss": 0.995,
+      "step": 18610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0808392560379672e-05,
+      "loss": 1.0369,
+      "step": 18611
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0807542581682107e-05,
+      "loss": 1.0869,
+      "step": 18612
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0806692597112004e-05,
+      "loss": 0.9897,
+      "step": 18613
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0805842606675547e-05,
+      "loss": 0.724,
+      "step": 18614
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0804992610378907e-05,
+      "loss": 0.8469,
+      "step": 18615
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0804142608228275e-05,
+      "loss": 1.1621,
+      "step": 18616
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0803292600229827e-05,
+      "loss": 0.6583,
+      "step": 18617
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0802442586389744e-05,
+      "loss": 1.2081,
+      "step": 18618
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0801592566714214e-05,
+      "loss": 1.1863,
+      "step": 18619
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0800742541209411e-05,
+      "loss": 0.704,
+      "step": 18620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0799892509881518e-05,
+      "loss": 0.7623,
+      "step": 18621
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.079904247273672e-05,
+      "loss": 0.8839,
+      "step": 18622
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0798192429781196e-05,
+      "loss": 1.1371,
+      "step": 18623
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0797342381021128e-05,
+      "loss": 0.8846,
+      "step": 18624
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0796492326462697e-05,
+      "loss": 0.7457,
+      "step": 18625
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0795642266112087e-05,
+      "loss": 0.8349,
+      "step": 18626
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0794792199975475e-05,
+      "loss": 1.1241,
+      "step": 18627
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0793942128059049e-05,
+      "loss": 1.0031,
+      "step": 18628
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.079309205036899e-05,
+      "loss": 0.9076,
+      "step": 18629
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0792241966911472e-05,
+      "loss": 0.7156,
+      "step": 18630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0791391877692686e-05,
+      "loss": 0.8723,
+      "step": 18631
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0790541782718806e-05,
+      "loss": 1.1684,
+      "step": 18632
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0789691681996021e-05,
+      "loss": 0.7245,
+      "step": 18633
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0788841575530513e-05,
+      "loss": 1.1316,
+      "step": 18634
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.078799146332846e-05,
+      "loss": 0.8978,
+      "step": 18635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0787141345396042e-05,
+      "loss": 1.1843,
+      "step": 18636
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0786291221739449e-05,
+      "loss": 1.1876,
+      "step": 18637
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0785441092364857e-05,
+      "loss": 1.048,
+      "step": 18638
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0784590957278452e-05,
+      "loss": 0.9102,
+      "step": 18639
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783740816486413e-05,
+      "loss": 0.9349,
+      "step": 18640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0782890669994924e-05,
+      "loss": 0.8607,
+      "step": 18641
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0782040517810164e-05,
+      "loss": 1.0286,
+      "step": 18642
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0781190359938323e-05,
+      "loss": 1.0287,
+      "step": 18643
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0780340196385578e-05,
+      "loss": 0.7868,
+      "step": 18644
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077949002715811e-05,
+      "loss": 1.0881,
+      "step": 18645
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0778639852262107e-05,
+      "loss": 0.6976,
+      "step": 18646
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0777789671703746e-05,
+      "loss": 0.7726,
+      "step": 18647
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0776939485489216e-05,
+      "loss": 0.8396,
+      "step": 18648
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0776089293624691e-05,
+      "loss": 1.1125,
+      "step": 18649
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0775239096116364e-05,
+      "loss": 0.8852,
+      "step": 18650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0774388892970404e-05,
+      "loss": 1.2878,
+      "step": 18651
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0773538684193008e-05,
+      "loss": 0.733,
+      "step": 18652
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0772688469790353e-05,
+      "loss": 0.8459,
+      "step": 18653
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077183824976862e-05,
+      "loss": 0.4406,
+      "step": 18654
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0770988024133994e-05,
+      "loss": 0.8774,
+      "step": 18655
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0770137792892656e-05,
+      "loss": 0.7914,
+      "step": 18656
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0769287556050794e-05,
+      "loss": 0.5713,
+      "step": 18657
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0768437313614584e-05,
+      "loss": 0.7127,
+      "step": 18658
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0767587065590214e-05,
+      "loss": 1.0068,
+      "step": 18659
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0766736811983864e-05,
+      "loss": 0.7529,
+      "step": 18660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0765886552801721e-05,
+      "loss": 1.0259,
+      "step": 18661
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0765036288049965e-05,
+      "loss": 1.2998,
+      "step": 18662
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.076418601773478e-05,
+      "loss": 0.972,
+      "step": 18663
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0763335741862351e-05,
+      "loss": 1.0336,
+      "step": 18664
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0762485460438858e-05,
+      "loss": 1.0522,
+      "step": 18665
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0761635173470485e-05,
+      "loss": 0.7601,
+      "step": 18666
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0760784880963417e-05,
+      "loss": 0.7318,
+      "step": 18667
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0759934582923839e-05,
+      "loss": 1.2323,
+      "step": 18668
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0759084279357932e-05,
+      "loss": 0.5899,
+      "step": 18669
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0758233970271879e-05,
+      "loss": 0.9699,
+      "step": 18670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0757383655671865e-05,
+      "loss": 0.6167,
+      "step": 18671
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0756533335564072e-05,
+      "loss": 0.7492,
+      "step": 18672
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0755683009954683e-05,
+      "loss": 1.013,
+      "step": 18673
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0754832678849887e-05,
+      "loss": 0.9776,
+      "step": 18674
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0753982342255863e-05,
+      "loss": 0.8901,
+      "step": 18675
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0753132000178791e-05,
+      "loss": 0.9442,
+      "step": 18676
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0752281652624862e-05,
+      "loss": 0.7785,
+      "step": 18677
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0751431299600261e-05,
+      "loss": 1.0126,
+      "step": 18678
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0750580941111166e-05,
+      "loss": 1.0386,
+      "step": 18679
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0749730577163761e-05,
+      "loss": 1.1368,
+      "step": 18680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0748880207764234e-05,
+      "loss": 1.3512,
+      "step": 18681
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0748029832918765e-05,
+      "loss": 0.9717,
+      "step": 18682
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0747179452633543e-05,
+      "loss": 1.1792,
+      "step": 18683
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0746329066914749e-05,
+      "loss": 0.9278,
+      "step": 18684
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0745478675768564e-05,
+      "loss": 1.1424,
+      "step": 18685
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0744628279201177e-05,
+      "loss": 1.306,
+      "step": 18686
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0743777877218771e-05,
+      "loss": 1.1389,
+      "step": 18687
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0742927469827526e-05,
+      "loss": 0.7198,
+      "step": 18688
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0742077057033638e-05,
+      "loss": 0.9849,
+      "step": 18689
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0741226638843276e-05,
+      "loss": 0.694,
+      "step": 18690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0740376215262633e-05,
+      "loss": 0.9529,
+      "step": 18691
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0739525786297893e-05,
+      "loss": 0.7804,
+      "step": 18692
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0738675351955238e-05,
+      "loss": 0.92,
+      "step": 18693
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0737824912240857e-05,
+      "loss": 0.536,
+      "step": 18694
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073697446716093e-05,
+      "loss": 0.9937,
+      "step": 18695
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0736124016721643e-05,
+      "loss": 1.2888,
+      "step": 18696
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0735273560929175e-05,
+      "loss": 1.0502,
+      "step": 18697
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0734423099789722e-05,
+      "loss": 1.0589,
+      "step": 18698
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0733572633309464e-05,
+      "loss": 0.9894,
+      "step": 18699
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073272216149458e-05,
+      "loss": 1.0997,
+      "step": 18700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0731871684351261e-05,
+      "loss": 0.9151,
+      "step": 18701
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073102120188569e-05,
+      "loss": 0.6491,
+      "step": 18702
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073017071410405e-05,
+      "loss": 0.9202,
+      "step": 18703
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072932022101253e-05,
+      "loss": 0.588,
+      "step": 18704
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072846972261731e-05,
+      "loss": 0.9557,
+      "step": 18705
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0727619218924577e-05,
+      "loss": 1.0582,
+      "step": 18706
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0726768709940519e-05,
+      "loss": 1.0174,
+      "step": 18707
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0725918195671317e-05,
+      "loss": 1.1408,
+      "step": 18708
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0725067676123155e-05,
+      "loss": 0.659,
+      "step": 18709
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0724217151302224e-05,
+      "loss": 0.8523,
+      "step": 18710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0723366621214704e-05,
+      "loss": 0.8881,
+      "step": 18711
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0722516085866779e-05,
+      "loss": 0.84,
+      "step": 18712
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072166554526464e-05,
+      "loss": 0.8578,
+      "step": 18713
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072081499941447e-05,
+      "loss": 0.9961,
+      "step": 18714
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071996444832245e-05,
+      "loss": 0.8518,
+      "step": 18715
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071911389199477e-05,
+      "loss": 0.8346,
+      "step": 18716
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0718263330437615e-05,
+      "loss": 1.1846,
+      "step": 18717
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.071741276365717e-05,
+      "loss": 1.0084,
+      "step": 18718
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0716562191659617e-05,
+      "loss": 1.1496,
+      "step": 18719
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0715711614451146e-05,
+      "loss": 0.7613,
+      "step": 18720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0714861032037938e-05,
+      "loss": 0.8431,
+      "step": 18721
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0714010444426182e-05,
+      "loss": 1.0555,
+      "step": 18722
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0713159851622064e-05,
+      "loss": 1.0438,
+      "step": 18723
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0712309253631767e-05,
+      "loss": 0.9969,
+      "step": 18724
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0711458650461482e-05,
+      "loss": 1.2347,
+      "step": 18725
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0710608042117387e-05,
+      "loss": 0.7746,
+      "step": 18726
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0709757428605672e-05,
+      "loss": 0.511,
+      "step": 18727
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.070890680993252e-05,
+      "loss": 1.2873,
+      "step": 18728
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0708056186104126e-05,
+      "loss": 0.9403,
+      "step": 18729
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0707205557126661e-05,
+      "loss": 1.4086,
+      "step": 18730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0706354923006321e-05,
+      "loss": 1.3339,
+      "step": 18731
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.070550428374929e-05,
+      "loss": 1.0787,
+      "step": 18732
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0704653639361755e-05,
+      "loss": 1.0164,
+      "step": 18733
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0703802989849901e-05,
+      "loss": 0.8924,
+      "step": 18734
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0702952335219912e-05,
+      "loss": 0.7911,
+      "step": 18735
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0702101675477974e-05,
+      "loss": 0.8236,
+      "step": 18736
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0701251010630275e-05,
+      "loss": 0.5394,
+      "step": 18737
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0700400340683e-05,
+      "loss": 1.0982,
+      "step": 18738
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0699549665642341e-05,
+      "loss": 1.1455,
+      "step": 18739
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0698698985514475e-05,
+      "loss": 1.2384,
+      "step": 18740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0697848300305593e-05,
+      "loss": 0.848,
+      "step": 18741
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0696997610021881e-05,
+      "loss": 0.8026,
+      "step": 18742
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0696146914669526e-05,
+      "loss": 0.826,
+      "step": 18743
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0695296214254714e-05,
+      "loss": 0.8223,
+      "step": 18744
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0694445508783628e-05,
+      "loss": 1.0277,
+      "step": 18745
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0693594798262456e-05,
+      "loss": 1.2442,
+      "step": 18746
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0692744082697389e-05,
+      "loss": 0.8951,
+      "step": 18747
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069189336209461e-05,
+      "loss": 0.5318,
+      "step": 18748
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0691042636460302e-05,
+      "loss": 0.9739,
+      "step": 18749
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0690191905800659e-05,
+      "loss": 0.9483,
+      "step": 18750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0689341170121863e-05,
+      "loss": 0.8991,
+      "step": 18751
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0688490429430099e-05,
+      "loss": 0.7828,
+      "step": 18752
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0687639683731559e-05,
+      "loss": 1.0963,
+      "step": 18753
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0686788933032429e-05,
+      "loss": 0.8516,
+      "step": 18754
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685938177338888e-05,
+      "loss": 0.9396,
+      "step": 18755
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685087416657131e-05,
+      "loss": 0.644,
+      "step": 18756
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0684236650993343e-05,
+      "loss": 0.9136,
+      "step": 18757
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0683385880353708e-05,
+      "loss": 0.9758,
+      "step": 18758
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0682535104744418e-05,
+      "loss": 0.7824,
+      "step": 18759
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0681684324171655e-05,
+      "loss": 0.9136,
+      "step": 18760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0680833538641607e-05,
+      "loss": 0.9665,
+      "step": 18761
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0679982748160462e-05,
+      "loss": 0.6463,
+      "step": 18762
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.067913195273441e-05,
+      "loss": 0.7531,
+      "step": 18763
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0678281152369632e-05,
+      "loss": 0.9279,
+      "step": 18764
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.067743034707232e-05,
+      "loss": 0.7969,
+      "step": 18765
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0676579536848658e-05,
+      "loss": 1.3116,
+      "step": 18766
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0675728721704834e-05,
+      "loss": 1.0127,
+      "step": 18767
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0674877901647037e-05,
+      "loss": 0.7022,
+      "step": 18768
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0674027076681454e-05,
+      "loss": 1.1009,
+      "step": 18769
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0673176246814268e-05,
+      "loss": 0.9022,
+      "step": 18770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0672325412051672e-05,
+      "loss": 0.9895,
+      "step": 18771
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0671474572399851e-05,
+      "loss": 0.9547,
+      "step": 18772
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.067062372786499e-05,
+      "loss": 0.8667,
+      "step": 18773
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0669772878453281e-05,
+      "loss": 0.7429,
+      "step": 18774
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066892202417091e-05,
+      "loss": 0.6459,
+      "step": 18775
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066807116502406e-05,
+      "loss": 1.0158,
+      "step": 18776
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0667220301018926e-05,
+      "loss": 0.8898,
+      "step": 18777
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0666369432161692e-05,
+      "loss": 0.7967,
+      "step": 18778
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0665518558458543e-05,
+      "loss": 0.8184,
+      "step": 18779
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066466767991567e-05,
+      "loss": 1.0155,
+      "step": 18780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0663816796539262e-05,
+      "loss": 0.9333,
+      "step": 18781
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.06629659083355e-05,
+      "loss": 0.6363,
+      "step": 18782
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066211501531058e-05,
+      "loss": 0.769,
+      "step": 18783
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0661264117470687e-05,
+      "loss": 1.1081,
+      "step": 18784
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0660413214822003e-05,
+      "loss": 0.7158,
+      "step": 18785
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0659562307370726e-05,
+      "loss": 0.9063,
+      "step": 18786
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0658711395123037e-05,
+      "loss": 0.7435,
+      "step": 18787
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0657860478085122e-05,
+      "loss": 0.8286,
+      "step": 18788
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0657009556263178e-05,
+      "loss": 0.8867,
+      "step": 18789
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0656158629663385e-05,
+      "loss": 0.6539,
+      "step": 18790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0655307698291932e-05,
+      "loss": 1.0223,
+      "step": 18791
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065445676215501e-05,
+      "loss": 1.2317,
+      "step": 18792
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0653605821258805e-05,
+      "loss": 0.7508,
+      "step": 18793
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0652754875609508e-05,
+      "loss": 1.2101,
+      "step": 18794
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0651903925213304e-05,
+      "loss": 0.8605,
+      "step": 18795
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0651052970076383e-05,
+      "loss": 0.7952,
+      "step": 18796
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0650202010204933e-05,
+      "loss": 1.0095,
+      "step": 18797
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.064935104560514e-05,
+      "loss": 0.6973,
+      "step": 18798
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0648500076283194e-05,
+      "loss": 1.0796,
+      "step": 18799
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0647649102245286e-05,
+      "loss": 0.9625,
+      "step": 18800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0646798123497597e-05,
+      "loss": 0.9343,
+      "step": 18801
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0645947140046323e-05,
+      "loss": 1.1076,
+      "step": 18802
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064509615189765e-05,
+      "loss": 1.004,
+      "step": 18803
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0644245159057766e-05,
+      "loss": 0.8247,
+      "step": 18804
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064339416153286e-05,
+      "loss": 0.9281,
+      "step": 18805
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064254315932912e-05,
+      "loss": 0.928,
+      "step": 18806
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0641692152452732e-05,
+      "loss": 0.9354,
+      "step": 18807
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0640841140909889e-05,
+      "loss": 0.8029,
+      "step": 18808
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0639990124706783e-05,
+      "loss": 0.9065,
+      "step": 18809
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0639139103849591e-05,
+      "loss": 0.9738,
+      "step": 18810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0638288078344512e-05,
+      "loss": 0.9118,
+      "step": 18811
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0637437048197732e-05,
+      "loss": 0.6838,
+      "step": 18812
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0636586013415436e-05,
+      "loss": 0.7665,
+      "step": 18813
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0635734974003818e-05,
+      "loss": 1.1116,
+      "step": 18814
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634883929969064e-05,
+      "loss": 0.9533,
+      "step": 18815
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634032881317362e-05,
+      "loss": 1.1662,
+      "step": 18816
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0633181828054905e-05,
+      "loss": 0.9149,
+      "step": 18817
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0632330770187877e-05,
+      "loss": 0.9275,
+      "step": 18818
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0631479707722472e-05,
+      "loss": 0.7839,
+      "step": 18819
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0630628640664875e-05,
+      "loss": 0.8165,
+      "step": 18820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0629777569021277e-05,
+      "loss": 0.8727,
+      "step": 18821
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0628926492797863e-05,
+      "loss": 1.0441,
+      "step": 18822
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0628075412000831e-05,
+      "loss": 0.937,
+      "step": 18823
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0627224326636365e-05,
+      "loss": 0.8623,
+      "step": 18824
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.062637323671065e-05,
+      "loss": 0.7053,
+      "step": 18825
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0625522142229883e-05,
+      "loss": 0.5319,
+      "step": 18826
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0624671043200248e-05,
+      "loss": 0.8017,
+      "step": 18827
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0623819939627933e-05,
+      "loss": 0.9974,
+      "step": 18828
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0622968831519135e-05,
+      "loss": 0.8773,
+      "step": 18829
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0622117718880036e-05,
+      "loss": 0.7382,
+      "step": 18830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0621266601716827e-05,
+      "loss": 0.9579,
+      "step": 18831
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.06204154800357e-05,
+      "loss": 0.8635,
+      "step": 18832
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0619564353842842e-05,
+      "loss": 0.8583,
+      "step": 18833
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0618713223144443e-05,
+      "loss": 0.631,
+      "step": 18834
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0617862087946694e-05,
+      "loss": 1.0014,
+      "step": 18835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.061701094825578e-05,
+      "loss": 0.9975,
+      "step": 18836
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0616159804077895e-05,
+      "loss": 0.8136,
+      "step": 18837
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.061530865541923e-05,
+      "loss": 0.8361,
+      "step": 18838
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0614457502285973e-05,
+      "loss": 0.8623,
+      "step": 18839
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0613606344684309e-05,
+      "loss": 0.7612,
+      "step": 18840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0612755182620434e-05,
+      "loss": 0.821,
+      "step": 18841
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0611904016100534e-05,
+      "loss": 0.9276,
+      "step": 18842
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.06110528451308e-05,
+      "loss": 0.6649,
+      "step": 18843
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0610201669717421e-05,
+      "loss": 0.8327,
+      "step": 18844
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060935048986659e-05,
+      "loss": 1.0933,
+      "step": 18845
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0608499305584492e-05,
+      "loss": 0.9405,
+      "step": 18846
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060764811687732e-05,
+      "loss": 0.8922,
+      "step": 18847
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0606796923751266e-05,
+      "loss": 0.9085,
+      "step": 18848
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0605945726212516e-05,
+      "loss": 0.8598,
+      "step": 18849
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0605094524267261e-05,
+      "loss": 1.0375,
+      "step": 18850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0604243317921692e-05,
+      "loss": 0.9743,
+      "step": 18851
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0603392107181998e-05,
+      "loss": 0.8979,
+      "step": 18852
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0602540892054368e-05,
+      "loss": 0.8364,
+      "step": 18853
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0601689672544997e-05,
+      "loss": 1.1362,
+      "step": 18854
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060083844866007e-05,
+      "loss": 0.8574,
+      "step": 18855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0599987220405779e-05,
+      "loss": 1.1178,
+      "step": 18856
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0599135987788314e-05,
+      "loss": 0.6446,
+      "step": 18857
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0598284750813868e-05,
+      "loss": 0.8603,
+      "step": 18858
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0597433509488625e-05,
+      "loss": 0.949,
+      "step": 18859
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0596582263818783e-05,
+      "loss": 0.9142,
+      "step": 18860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0595731013810526e-05,
+      "loss": 0.9681,
+      "step": 18861
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0594879759470047e-05,
+      "loss": 0.6185,
+      "step": 18862
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0594028500803536e-05,
+      "loss": 0.6895,
+      "step": 18863
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0593177237817188e-05,
+      "loss": 1.0081,
+      "step": 18864
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0592325970517185e-05,
+      "loss": 0.9812,
+      "step": 18865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0591474698909723e-05,
+      "loss": 0.8652,
+      "step": 18866
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0590623423000993e-05,
+      "loss": 0.738,
+      "step": 18867
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058977214279718e-05,
+      "loss": 0.9085,
+      "step": 18868
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0588920858304482e-05,
+      "loss": 1.1667,
+      "step": 18869
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0588069569529085e-05,
+      "loss": 1.1897,
+      "step": 18870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0587218276477178e-05,
+      "loss": 0.8016,
+      "step": 18871
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0586366979154957e-05,
+      "loss": 1.1069,
+      "step": 18872
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058551567756861e-05,
+      "loss": 1.1713,
+      "step": 18873
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058466437172433e-05,
+      "loss": 0.982,
+      "step": 18874
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0583813061628304e-05,
+      "loss": 0.6466,
+      "step": 18875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0582961747286724e-05,
+      "loss": 0.7314,
+      "step": 18876
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058211042870578e-05,
+      "loss": 0.8902,
+      "step": 18877
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0581259105891666e-05,
+      "loss": 0.8977,
+      "step": 18878
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0580407778850575e-05,
+      "loss": 0.7358,
+      "step": 18879
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0579556447588688e-05,
+      "loss": 0.8959,
+      "step": 18880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0578705112112205e-05,
+      "loss": 1.2019,
+      "step": 18881
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0577853772427314e-05,
+      "loss": 1.0943,
+      "step": 18882
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0577002428540203e-05,
+      "loss": 0.97,
+      "step": 18883
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0576151080457073e-05,
+      "loss": 1.0768,
+      "step": 18884
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0575299728184105e-05,
+      "loss": 0.9051,
+      "step": 18885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.057444837172749e-05,
+      "loss": 0.9501,
+      "step": 18886
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573597011093426e-05,
+      "loss": 1.0397,
+      "step": 18887
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05727456462881e-05,
+      "loss": 0.8069,
+      "step": 18888
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0571894277317707e-05,
+      "loss": 0.8583,
+      "step": 18889
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0571042904188433e-05,
+      "loss": 0.8496,
+      "step": 18890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0570191526906472e-05,
+      "loss": 0.8906,
+      "step": 18891
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0569340145478014e-05,
+      "loss": 1.1247,
+      "step": 18892
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.056848875990925e-05,
+      "loss": 1.1484,
+      "step": 18893
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0567637370206378e-05,
+      "loss": 1.0178,
+      "step": 18894
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0566785976375577e-05,
+      "loss": 1.3965,
+      "step": 18895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.056593457842305e-05,
+      "loss": 1.1335,
+      "step": 18896
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0565083176354982e-05,
+      "loss": 0.6482,
+      "step": 18897
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0564231770177566e-05,
+      "loss": 0.872,
+      "step": 18898
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0563380359896998e-05,
+      "loss": 1.2271,
+      "step": 18899
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0562528945519463e-05,
+      "loss": 0.9577,
+      "step": 18900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0561677527051154e-05,
+      "loss": 0.772,
+      "step": 18901
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0560826104498263e-05,
+      "loss": 1.0358,
+      "step": 18902
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0559974677866984e-05,
+      "loss": 1.0574,
+      "step": 18903
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0559123247163506e-05,
+      "loss": 0.9659,
+      "step": 18904
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0558271812394024e-05,
+      "loss": 1.198,
+      "step": 18905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0557420373564725e-05,
+      "loss": 0.6506,
+      "step": 18906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05565689306818e-05,
+      "loss": 0.9537,
+      "step": 18907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0555717483751449e-05,
+      "loss": 1.3407,
+      "step": 18908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.055486603277986e-05,
+      "loss": 0.6862,
+      "step": 18909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0554014577773219e-05,
+      "loss": 0.9924,
+      "step": 18910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0553163118737724e-05,
+      "loss": 0.8827,
+      "step": 18911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0552311655679565e-05,
+      "loss": 0.8771,
+      "step": 18912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0551460188604936e-05,
+      "loss": 0.9084,
+      "step": 18913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0550608717520028e-05,
+      "loss": 1.1453,
+      "step": 18914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0549757242431032e-05,
+      "loss": 1.1134,
+      "step": 18915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0548905763344136e-05,
+      "loss": 0.9493,
+      "step": 18916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0548054280265539e-05,
+      "loss": 1.1741,
+      "step": 18917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054720279320143e-05,
+      "loss": 0.7817,
+      "step": 18918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0546351302158001e-05,
+      "loss": 1.3634,
+      "step": 18919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0545499807141448e-05,
+      "loss": 0.6396,
+      "step": 18920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0544648308157959e-05,
+      "loss": 0.9563,
+      "step": 18921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0543796805213724e-05,
+      "loss": 0.7582,
+      "step": 18922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054294529831494e-05,
+      "loss": 1.0701,
+      "step": 18923
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05420937874678e-05,
+      "loss": 1.3203,
+      "step": 18924
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054124227267849e-05,
+      "loss": 0.978,
+      "step": 18925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0540390753953204e-05,
+      "loss": 0.7859,
+      "step": 18926
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0539539231298141e-05,
+      "loss": 0.7995,
+      "step": 18927
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0538687704719488e-05,
+      "loss": 0.9634,
+      "step": 18928
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053783617422344e-05,
+      "loss": 0.7112,
+      "step": 18929
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536984639816183e-05,
+      "loss": 0.7003,
+      "step": 18930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536133101503918e-05,
+      "loss": 0.6743,
+      "step": 18931
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053528155929283e-05,
+      "loss": 0.7721,
+      "step": 18932
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0534430013189117e-05,
+      "loss": 1.082,
+      "step": 18933
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053357846319897e-05,
+      "loss": 1.0342,
+      "step": 18934
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0532726909328581e-05,
+      "loss": 0.8734,
+      "step": 18935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0531875351584142e-05,
+      "loss": 0.8849,
+      "step": 18936
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0531023789971849e-05,
+      "loss": 0.9935,
+      "step": 18937
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0530172224497888e-05,
+      "loss": 0.9949,
+      "step": 18938
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0529320655168457e-05,
+      "loss": 0.7124,
+      "step": 18939
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.052846908198975e-05,
+      "loss": 0.813,
+      "step": 18940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0527617504967953e-05,
+      "loss": 1.0828,
+      "step": 18941
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0526765924109266e-05,
+      "loss": 1.1534,
+      "step": 18942
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525914339419878e-05,
+      "loss": 0.7443,
+      "step": 18943
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525062750905983e-05,
+      "loss": 1.1676,
+      "step": 18944
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0524211158573772e-05,
+      "loss": 0.9335,
+      "step": 18945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0523359562429441e-05,
+      "loss": 0.7935,
+      "step": 18946
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0522507962479178e-05,
+      "loss": 0.8218,
+      "step": 18947
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0521656358729183e-05,
+      "loss": 1.4673,
+      "step": 18948
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0520804751185645e-05,
+      "loss": 1.0375,
+      "step": 18949
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0519953139854752e-05,
+      "loss": 0.8856,
+      "step": 18950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0519101524742705e-05,
+      "loss": 0.853,
+      "step": 18951
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0518249905855695e-05,
+      "loss": 0.7707,
+      "step": 18952
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0517398283199911e-05,
+      "loss": 0.9884,
+      "step": 18953
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0516546656781554e-05,
+      "loss": 0.8282,
+      "step": 18954
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0515695026606812e-05,
+      "loss": 0.8951,
+      "step": 18955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514843392681873e-05,
+      "loss": 0.9547,
+      "step": 18956
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.051399175501294e-05,
+      "loss": 0.757,
+      "step": 18957
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0513140113606202e-05,
+      "loss": 0.7823,
+      "step": 18958
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0512288468467851e-05,
+      "loss": 0.9281,
+      "step": 18959
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0511436819604082e-05,
+      "loss": 0.8726,
+      "step": 18960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0510585167021088e-05,
+      "loss": 1.1209,
+      "step": 18961
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0509733510725057e-05,
+      "loss": 1.2254,
+      "step": 18962
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0508881850722192e-05,
+      "loss": 0.8609,
+      "step": 18963
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0508030187018685e-05,
+      "loss": 1.1007,
+      "step": 18964
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.050717851962072e-05,
+      "loss": 1.1922,
+      "step": 18965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0506326848534499e-05,
+      "loss": 1.0036,
+      "step": 18966
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0505475173766213e-05,
+      "loss": 0.9649,
+      "step": 18967
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0504623495322052e-05,
+      "loss": 0.891,
+      "step": 18968
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.050377181320822e-05,
+      "loss": 0.8792,
+      "step": 18969
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05029201274309e-05,
+      "loss": 0.8059,
+      "step": 18970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0502068437996285e-05,
+      "loss": 0.7578,
+      "step": 18971
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0501216744910577e-05,
+      "loss": 0.7127,
+      "step": 18972
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0500365048179964e-05,
+      "loss": 1.3242,
+      "step": 18973
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0499513347810641e-05,
+      "loss": 0.9016,
+      "step": 18974
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0498661643808801e-05,
+      "loss": 0.7878,
+      "step": 18975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0497809936180638e-05,
+      "loss": 0.8192,
+      "step": 18976
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0496958224932345e-05,
+      "loss": 0.9713,
+      "step": 18977
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049610651007012e-05,
+      "loss": 0.7968,
+      "step": 18978
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0495254791600153e-05,
+      "loss": 0.9243,
+      "step": 18979
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0494403069528634e-05,
+      "loss": 1.1647,
+      "step": 18980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0493551343861765e-05,
+      "loss": 0.8094,
+      "step": 18981
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0492699614605735e-05,
+      "loss": 0.753,
+      "step": 18982
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0491847881766736e-05,
+      "loss": 0.7017,
+      "step": 18983
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049099614535097e-05,
+      "loss": 0.4693,
+      "step": 18984
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049014440536462e-05,
+      "loss": 0.8531,
+      "step": 18985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0489292661813885e-05,
+      "loss": 1.0084,
+      "step": 18986
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0488440914704963e-05,
+      "loss": 1.1001,
+      "step": 18987
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0487589164044043e-05,
+      "loss": 0.8358,
+      "step": 18988
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0486737409837322e-05,
+      "loss": 0.7525,
+      "step": 18989
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0485885652090992e-05,
+      "loss": 0.8564,
+      "step": 18990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0485033890811248e-05,
+      "loss": 0.8866,
+      "step": 18991
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048418212600428e-05,
+      "loss": 0.7487,
+      "step": 18992
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048333035767629e-05,
+      "loss": 0.9378,
+      "step": 18993
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0482478585833467e-05,
+      "loss": 0.8686,
+      "step": 18994
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0481626810482004e-05,
+      "loss": 1.054,
+      "step": 18995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0480775031628098e-05,
+      "loss": 0.923,
+      "step": 18996
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0479923249277943e-05,
+      "loss": 0.9572,
+      "step": 18997
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0479071463437736e-05,
+      "loss": 0.9757,
+      "step": 18998
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0478219674113666e-05,
+      "loss": 0.8303,
+      "step": 18999
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0477367881311929e-05,
+      "loss": 0.8358,
+      "step": 19000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0476516085038719e-05,
+      "loss": 0.8742,
+      "step": 19001
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.047566428530023e-05,
+      "loss": 0.7041,
+      "step": 19002
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0474812482102659e-05,
+      "loss": 0.4689,
+      "step": 19003
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04739606754522e-05,
+      "loss": 0.9773,
+      "step": 19004
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0473108865355046e-05,
+      "loss": 0.8298,
+      "step": 19005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0472257051817392e-05,
+      "loss": 0.8525,
+      "step": 19006
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.047140523484543e-05,
+      "loss": 1.0019,
+      "step": 19007
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0470553414445356e-05,
+      "loss": 0.9446,
+      "step": 19008
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046970159062337e-05,
+      "loss": 0.6902,
+      "step": 19009
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046884976338566e-05,
+      "loss": 1.0231,
+      "step": 19010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0467997932738421e-05,
+      "loss": 1.1979,
+      "step": 19011
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046714609868785e-05,
+      "loss": 0.9716,
+      "step": 19012
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0466294261240142e-05,
+      "loss": 1.0301,
+      "step": 19013
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0465442420401488e-05,
+      "loss": 1.2321,
+      "step": 19014
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0464590576178086e-05,
+      "loss": 0.8996,
+      "step": 19015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0463738728576132e-05,
+      "loss": 1.0471,
+      "step": 19016
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0462886877601815e-05,
+      "loss": 0.6478,
+      "step": 19017
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0462035023261332e-05,
+      "loss": 1.0361,
+      "step": 19018
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0461183165560885e-05,
+      "loss": 0.9744,
+      "step": 19019
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0460331304506658e-05,
+      "loss": 0.9818,
+      "step": 19020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0459479440104851e-05,
+      "loss": 0.8129,
+      "step": 19021
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045862757236166e-05,
+      "loss": 0.7987,
+      "step": 19022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0457775701283277e-05,
+      "loss": 0.8627,
+      "step": 19023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456923826875902e-05,
+      "loss": 0.7879,
+      "step": 19024
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456071949145722e-05,
+      "loss": 0.8639,
+      "step": 19025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0455220068098937e-05,
+      "loss": 0.9291,
+      "step": 19026
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0454368183741741e-05,
+      "loss": 0.7943,
+      "step": 19027
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045351629608033e-05,
+      "loss": 0.815,
+      "step": 19028
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0452664405120897e-05,
+      "loss": 1.0994,
+      "step": 19029
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0451812510869641e-05,
+      "loss": 1.0321,
+      "step": 19030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0450960613332752e-05,
+      "loss": 0.703,
+      "step": 19031
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0450108712516425e-05,
+      "loss": 0.4913,
+      "step": 19032
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0449256808426859e-05,
+      "loss": 1.1483,
+      "step": 19033
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0448404901070253e-05,
+      "loss": 0.7706,
+      "step": 19034
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044755299045279e-05,
+      "loss": 1.0585,
+      "step": 19035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0446701076580674e-05,
+      "loss": 1.0036,
+      "step": 19036
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04458491594601e-05,
+      "loss": 0.8734,
+      "step": 19037
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0444997239097256e-05,
+      "loss": 1.0115,
+      "step": 19038
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044414531549835e-05,
+      "loss": 0.7356,
+      "step": 19039
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0443293388669567e-05,
+      "loss": 0.7937,
+      "step": 19040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0442441458617103e-05,
+      "loss": 1.1138,
+      "step": 19041
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0441589525347157e-05,
+      "loss": 1.495,
+      "step": 19042
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0440737588865924e-05,
+      "loss": 0.9387,
+      "step": 19043
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0439885649179597e-05,
+      "loss": 0.7997,
+      "step": 19044
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0439033706294374e-05,
+      "loss": 0.8545,
+      "step": 19045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043818176021645e-05,
+      "loss": 1.1496,
+      "step": 19046
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0437329810952015e-05,
+      "loss": 1.1627,
+      "step": 19047
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0436477858507272e-05,
+      "loss": 0.895,
+      "step": 19048
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0435625902888418e-05,
+      "loss": 1.2522,
+      "step": 19049
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0434773944101637e-05,
+      "loss": 1.2256,
+      "step": 19050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0433921982153135e-05,
+      "loss": 0.9604,
+      "step": 19051
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0433070017049103e-05,
+      "loss": 1.0782,
+      "step": 19052
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043221804879574e-05,
+      "loss": 0.9259,
+      "step": 19053
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0431366077399237e-05,
+      "loss": 1.1192,
+      "step": 19054
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0430514102865791e-05,
+      "loss": 0.875,
+      "step": 19055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04296621252016e-05,
+      "loss": 0.8006,
+      "step": 19056
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0428810144412858e-05,
+      "loss": 0.9605,
+      "step": 19057
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427958160505761e-05,
+      "loss": 0.9376,
+      "step": 19058
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427106173486505e-05,
+      "loss": 0.9449,
+      "step": 19059
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0426254183361286e-05,
+      "loss": 1.3094,
+      "step": 19060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04254021901363e-05,
+      "loss": 1.0305,
+      "step": 19061
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0424550193817742e-05,
+      "loss": 0.7369,
+      "step": 19062
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0423698194411804e-05,
+      "loss": 0.9142,
+      "step": 19063
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0422846191924691e-05,
+      "loss": 0.9118,
+      "step": 19064
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0421994186362591e-05,
+      "loss": 1.1005,
+      "step": 19065
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.04211421777317e-05,
+      "loss": 0.8772,
+      "step": 19066
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042029016603822e-05,
+      "loss": 0.885,
+      "step": 19067
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041943815128834e-05,
+      "loss": 0.8555,
+      "step": 19068
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0418586133488265e-05,
+      "loss": 0.8798,
+      "step": 19069
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041773411264418e-05,
+      "loss": 0.7376,
+      "step": 19070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0416882088762288e-05,
+      "loss": 0.7608,
+      "step": 19071
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.041603006184878e-05,
+      "loss": 0.9325,
+      "step": 19072
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0415178031909859e-05,
+      "loss": 0.4917,
+      "step": 19073
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0414325998951718e-05,
+      "loss": 0.8853,
+      "step": 19074
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413473962980549e-05,
+      "loss": 0.9804,
+      "step": 19075
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0412621924002553e-05,
+      "loss": 0.9837,
+      "step": 19076
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0411769882023925e-05,
+      "loss": 0.9153,
+      "step": 19077
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0410917837050857e-05,
+      "loss": 0.8963,
+      "step": 19078
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0410065789089554e-05,
+      "loss": 0.8599,
+      "step": 19079
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0409213738146207e-05,
+      "loss": 1.1005,
+      "step": 19080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0408361684227006e-05,
+      "loss": 1.2252,
+      "step": 19081
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0407509627338159e-05,
+      "loss": 0.8824,
+      "step": 19082
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0406657567485856e-05,
+      "loss": 0.9281,
+      "step": 19083
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0405805504676294e-05,
+      "loss": 0.9277,
+      "step": 19084
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0404953438915667e-05,
+      "loss": 0.9111,
+      "step": 19085
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0404101370210178e-05,
+      "loss": 0.6814,
+      "step": 19086
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0403249298566013e-05,
+      "loss": 1.0868,
+      "step": 19087
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0402397223989378e-05,
+      "loss": 1.0743,
+      "step": 19088
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040154514648647e-05,
+      "loss": 1.1344,
+      "step": 19089
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0400693066063474e-05,
+      "loss": 0.7585,
+      "step": 19090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0399840982726595e-05,
+      "loss": 0.8306,
+      "step": 19091
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039898889648203e-05,
+      "loss": 0.9223,
+      "step": 19092
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039813680733597e-05,
+      "loss": 0.7524,
+      "step": 19093
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039728471529462e-05,
+      "loss": 1.0014,
+      "step": 19094
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039643262036417e-05,
+      "loss": 0.6039,
+      "step": 19095
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0395580522550815e-05,
+      "loss": 0.858,
+      "step": 19096
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0394728421860755e-05,
+      "loss": 1.0109,
+      "step": 19097
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0393876318300187e-05,
+      "loss": 0.9938,
+      "step": 19098
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039302421187531e-05,
+      "loss": 0.8829,
+      "step": 19099
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0392172102592312e-05,
+      "loss": 0.9573,
+      "step": 19100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0391319990457398e-05,
+      "loss": 0.9549,
+      "step": 19101
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039046787547676e-05,
+      "loss": 0.9711,
+      "step": 19102
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.03896157576566e-05,
+      "loss": 0.9818,
+      "step": 19103
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0388763637003108e-05,
+      "loss": 0.8081,
+      "step": 19104
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038791151352248e-05,
+      "loss": 0.8453,
+      "step": 19105
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038705938722092e-05,
+      "loss": 0.7755,
+      "step": 19106
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0386207258104622e-05,
+      "loss": 0.898,
+      "step": 19107
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0385355126179782e-05,
+      "loss": 0.889,
+      "step": 19108
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0384502991452596e-05,
+      "loss": 1.0957,
+      "step": 19109
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0383650853929261e-05,
+      "loss": 0.8733,
+      "step": 19110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0382798713615974e-05,
+      "loss": 1.0129,
+      "step": 19111
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0381946570518932e-05,
+      "loss": 0.736,
+      "step": 19112
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0381094424644335e-05,
+      "loss": 1.0753,
+      "step": 19113
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0380242275998373e-05,
+      "loss": 0.9496,
+      "step": 19114
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0379390124587249e-05,
+      "loss": 0.9023,
+      "step": 19115
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0378537970417159e-05,
+      "loss": 0.8742,
+      "step": 19116
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0377685813494298e-05,
+      "loss": 0.7567,
+      "step": 19117
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0376833653824865e-05,
+      "loss": 0.8005,
+      "step": 19118
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0375981491415055e-05,
+      "loss": 0.7821,
+      "step": 19119
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0375129326271063e-05,
+      "loss": 0.8516,
+      "step": 19120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0374277158399092e-05,
+      "loss": 0.7028,
+      "step": 19121
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0373424987805336e-05,
+      "loss": 1.0359,
+      "step": 19122
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037257281449599e-05,
+      "loss": 0.9804,
+      "step": 19123
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371720638477255e-05,
+      "loss": 0.8799,
+      "step": 19124
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0370868459755325e-05,
+      "loss": 0.6945,
+      "step": 19125
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0370016278336398e-05,
+      "loss": 0.6852,
+      "step": 19126
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0369164094226673e-05,
+      "loss": 0.9171,
+      "step": 19127
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0368311907432346e-05,
+      "loss": 0.8928,
+      "step": 19128
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0367459717959612e-05,
+      "loss": 0.7541,
+      "step": 19129
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0366607525814671e-05,
+      "loss": 0.7365,
+      "step": 19130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.036575533100372e-05,
+      "loss": 1.0745,
+      "step": 19131
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0364903133532955e-05,
+      "loss": 1.0145,
+      "step": 19132
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0364050933408576e-05,
+      "loss": 0.9253,
+      "step": 19133
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0363198730636777e-05,
+      "loss": 0.8969,
+      "step": 19134
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0362346525223755e-05,
+      "loss": 0.7562,
+      "step": 19135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0361494317175706e-05,
+      "loss": 0.9513,
+      "step": 19136
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0360642106498835e-05,
+      "loss": 0.9145,
+      "step": 19137
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0359789893199335e-05,
+      "loss": 0.7976,
+      "step": 19138
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.03589376772834e-05,
+      "loss": 1.056,
+      "step": 19139
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0358085458757233e-05,
+      "loss": 1.0011,
+      "step": 19140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0357233237627026e-05,
+      "loss": 1.1763,
+      "step": 19141
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035638101389898e-05,
+      "loss": 0.6856,
+      "step": 19142
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0355528787579294e-05,
+      "loss": 0.631,
+      "step": 19143
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0354676558674162e-05,
+      "loss": 1.0033,
+      "step": 19144
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035382432718978e-05,
+      "loss": 1.2983,
+      "step": 19145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352972093132352e-05,
+      "loss": 0.7591,
+      "step": 19146
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352119856508069e-05,
+      "loss": 0.8914,
+      "step": 19147
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0351267617323133e-05,
+      "loss": 0.7945,
+      "step": 19148
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035041537558374e-05,
+      "loss": 0.6037,
+      "step": 19149
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0349563131296087e-05,
+      "loss": 1.0122,
+      "step": 19150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0348710884466371e-05,
+      "loss": 0.8912,
+      "step": 19151
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0347858635100792e-05,
+      "loss": 0.9278,
+      "step": 19152
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0347006383205546e-05,
+      "loss": 0.7637,
+      "step": 19153
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.034615412878683e-05,
+      "loss": 1.0844,
+      "step": 19154
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0345301871850843e-05,
+      "loss": 1.0404,
+      "step": 19155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0344449612403784e-05,
+      "loss": 0.7784,
+      "step": 19156
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0343597350451846e-05,
+      "loss": 1.1843,
+      "step": 19157
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0342745086001235e-05,
+      "loss": 0.7243,
+      "step": 19158
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.034189281905814e-05,
+      "loss": 0.7994,
+      "step": 19159
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0341040549628762e-05,
+      "loss": 0.8573,
+      "step": 19160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0340188277719303e-05,
+      "loss": 1.4205,
+      "step": 19161
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0339336003335954e-05,
+      "loss": 0.8577,
+      "step": 19162
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0338483726484919e-05,
+      "loss": 1.0753,
+      "step": 19163
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033763144717239e-05,
+      "loss": 0.846,
+      "step": 19164
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0336779165404567e-05,
+      "loss": 0.9018,
+      "step": 19165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0335926881187648e-05,
+      "loss": 1.2991,
+      "step": 19166
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0335074594527832e-05,
+      "loss": 0.7308,
+      "step": 19167
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0334222305431321e-05,
+      "loss": 1.0141,
+      "step": 19168
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0333370013904303e-05,
+      "loss": 0.9111,
+      "step": 19169
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0332517719952982e-05,
+      "loss": 0.9293,
+      "step": 19170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0331665423583558e-05,
+      "loss": 0.7815,
+      "step": 19171
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033081312480222e-05,
+      "loss": 0.7222,
+      "step": 19172
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.032996082361518e-05,
+      "loss": 0.8916,
+      "step": 19173
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0329108520028623e-05,
+      "loss": 0.9713,
+      "step": 19174
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0328256214048753e-05,
+      "loss": 1.0366,
+      "step": 19175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.032740390568177e-05,
+      "loss": 1.4218,
+      "step": 19176
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0326551594933866e-05,
+      "loss": 0.8805,
+      "step": 19177
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0325699281811245e-05,
+      "loss": 1.042,
+      "step": 19178
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0324846966320103e-05,
+      "loss": 1.1649,
+      "step": 19179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0323994648466637e-05,
+      "loss": 0.9387,
+      "step": 19180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0323142328257044e-05,
+      "loss": 0.8059,
+      "step": 19181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0322290005697527e-05,
+      "loss": 0.934,
+      "step": 19182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0321437680794282e-05,
+      "loss": 1.0753,
+      "step": 19183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0320585353553503e-05,
+      "loss": 0.9192,
+      "step": 19184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0319733023981392e-05,
+      "loss": 0.8348,
+      "step": 19185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031888069208415e-05,
+      "loss": 0.9363,
+      "step": 19186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0318028357867968e-05,
+      "loss": 0.8068,
+      "step": 19187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0317176021339053e-05,
+      "loss": 1.0092,
+      "step": 19188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0316323682503597e-05,
+      "loss": 1.0044,
+      "step": 19189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0315471341367797e-05,
+      "loss": 0.684,
+      "step": 19190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0314618997937857e-05,
+      "loss": 0.7585,
+      "step": 19191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0313766652219974e-05,
+      "loss": 1.1171,
+      "step": 19192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0312914304220342e-05,
+      "loss": 0.9875,
+      "step": 19193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0312061953945167e-05,
+      "loss": 0.5927,
+      "step": 19194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031120960140064e-05,
+      "loss": 1.1782,
+      "step": 19195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0310357246592961e-05,
+      "loss": 0.6106,
+      "step": 19196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030950488952833e-05,
+      "loss": 0.7139,
+      "step": 19197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0308652530212949e-05,
+      "loss": 1.1256,
+      "step": 19198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0307800168653008e-05,
+      "loss": 1.0034,
+      "step": 19199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030694780485471e-05,
+      "loss": 1.0438,
+      "step": 19200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0306095438824257e-05,
+      "loss": 1.028,
+      "step": 19201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0305243070567841e-05,
+      "loss": 0.7867,
+      "step": 19202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0304390700091664e-05,
+      "loss": 0.6227,
+      "step": 19203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0303538327401925e-05,
+      "loss": 1.2336,
+      "step": 19204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0302685952504824e-05,
+      "loss": 0.9714,
+      "step": 19205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0301833575406551e-05,
+      "loss": 1.0034,
+      "step": 19206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0300981196113313e-05,
+      "loss": 1.1075,
+      "step": 19207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030012881463131e-05,
+      "loss": 0.8194,
+      "step": 19208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0299276430966734e-05,
+      "loss": 0.9867,
+      "step": 19209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0298424045125785e-05,
+      "loss": 0.8151,
+      "step": 19210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0297571657114667e-05,
+      "loss": 0.8213,
+      "step": 19211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.029671926693957e-05,
+      "loss": 1.1198,
+      "step": 19212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0295866874606703e-05,
+      "loss": 0.7553,
+      "step": 19213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0295014480122256e-05,
+      "loss": 1.0265,
+      "step": 19214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0294162083492429e-05,
+      "loss": 0.83,
+      "step": 19215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0293309684723426e-05,
+      "loss": 1.019,
+      "step": 19216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0292457283821443e-05,
+      "loss": 0.7027,
+      "step": 19217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0291604880792676e-05,
+      "loss": 1.0983,
+      "step": 19218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0290752475643329e-05,
+      "loss": 1.2112,
+      "step": 19219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0289900068379595e-05,
+      "loss": 1.1523,
+      "step": 19220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0289047659007674e-05,
+      "loss": 0.9578,
+      "step": 19221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0288195247533768e-05,
+      "loss": 0.9021,
+      "step": 19222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0287342833964076e-05,
+      "loss": 0.7875,
+      "step": 19223
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0286490418304793e-05,
+      "loss": 0.9666,
+      "step": 19224
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0285638000562119e-05,
+      "loss": 0.9308,
+      "step": 19225
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0284785580742256e-05,
+      "loss": 1.0643,
+      "step": 19226
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0283933158851399e-05,
+      "loss": 0.6975,
+      "step": 19227
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028308073489575e-05,
+      "loss": 0.8015,
+      "step": 19228
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0282228308881506e-05,
+      "loss": 0.9524,
+      "step": 19229
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0281375880814864e-05,
+      "loss": 1.2096,
+      "step": 19230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0280523450702027e-05,
+      "loss": 0.7545,
+      "step": 19231
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0279671018549193e-05,
+      "loss": 1.2705,
+      "step": 19232
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0278818584362562e-05,
+      "loss": 1.0571,
+      "step": 19233
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0277966148148328e-05,
+      "loss": 0.7741,
+      "step": 19234
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0277113709912697e-05,
+      "loss": 1.1601,
+      "step": 19235
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0276261269661858e-05,
+      "loss": 0.8141,
+      "step": 19236
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027540882740202e-05,
+      "loss": 0.8725,
+      "step": 19237
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0274556383139383e-05,
+      "loss": 0.5632,
+      "step": 19238
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0273703936880135e-05,
+      "loss": 0.8096,
+      "step": 19239
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0272851488630484e-05,
+      "loss": 0.8595,
+      "step": 19240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0271999038396625e-05,
+      "loss": 0.775,
+      "step": 19241
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027114658618476e-05,
+      "loss": 0.8894,
+      "step": 19242
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0270294132001088e-05,
+      "loss": 0.8289,
+      "step": 19243
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0269441675851808e-05,
+      "loss": 0.5712,
+      "step": 19244
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0268589217743114e-05,
+      "loss": 0.8957,
+      "step": 19245
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0267736757681212e-05,
+      "loss": 0.8306,
+      "step": 19246
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02668842956723e-05,
+      "loss": 0.9105,
+      "step": 19247
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0266031831722574e-05,
+      "loss": 1.0344,
+      "step": 19248
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0265179365838233e-05,
+      "loss": 1.1348,
+      "step": 19249
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0264326898025483e-05,
+      "loss": 1.2119,
+      "step": 19250
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0263474428290512e-05,
+      "loss": 0.9132,
+      "step": 19251
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0262621956639532e-05,
+      "loss": 0.8278,
+      "step": 19252
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0261769483078734e-05,
+      "loss": 0.8795,
+      "step": 19253
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0260917007614316e-05,
+      "loss": 0.8979,
+      "step": 19254
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0260064530252485e-05,
+      "loss": 0.958,
+      "step": 19255
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0259212050999435e-05,
+      "loss": 1.027,
+      "step": 19256
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0258359569861364e-05,
+      "loss": 0.8476,
+      "step": 19257
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0257507086844476e-05,
+      "loss": 0.8713,
+      "step": 19258
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0256654601954967e-05,
+      "loss": 0.8719,
+      "step": 19259
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0255802115199034e-05,
+      "loss": 0.7288,
+      "step": 19260
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0254949626582883e-05,
+      "loss": 1.0121,
+      "step": 19261
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025409713611271e-05,
+      "loss": 0.8272,
+      "step": 19262
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0253244643794712e-05,
+      "loss": 0.7037,
+      "step": 19263
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0252392149635096e-05,
+      "loss": 0.5761,
+      "step": 19264
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0251539653640052e-05,
+      "loss": 0.7958,
+      "step": 19265
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0250687155815786e-05,
+      "loss": 1.0578,
+      "step": 19266
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0249834656168492e-05,
+      "loss": 1.1234,
+      "step": 19267
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0248982154704376e-05,
+      "loss": 0.7711,
+      "step": 19268
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024812965142963e-05,
+      "loss": 1.3321,
+      "step": 19269
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024727714635046e-05,
+      "loss": 0.9712,
+      "step": 19270
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0246424639473064e-05,
+      "loss": 0.8791,
+      "step": 19271
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0245572130803641e-05,
+      "loss": 0.9396,
+      "step": 19272
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024471962034839e-05,
+      "loss": 1.04,
+      "step": 19273
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024386710811351e-05,
+      "loss": 1.0197,
+      "step": 19274
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0243014594105201e-05,
+      "loss": 0.7833,
+      "step": 19275
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0242162078329663e-05,
+      "loss": 1.2181,
+      "step": 19276
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0241309560793095e-05,
+      "loss": 0.8468,
+      "step": 19277
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0240457041501698e-05,
+      "loss": 0.9653,
+      "step": 19278
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.023960452046167e-05,
+      "loss": 0.8923,
+      "step": 19279
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0238751997679212e-05,
+      "loss": 0.8371,
+      "step": 19280
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0237899473160524e-05,
+      "loss": 0.9568,
+      "step": 19281
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0237046946911802e-05,
+      "loss": 0.9103,
+      "step": 19282
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0236194418939252e-05,
+      "loss": 0.929,
+      "step": 19283
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0235341889249067e-05,
+      "loss": 1.0303,
+      "step": 19284
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0234489357847447e-05,
+      "loss": 0.8244,
+      "step": 19285
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0233636824740599e-05,
+      "loss": 0.9461,
+      "step": 19286
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0232784289934717e-05,
+      "loss": 0.6633,
+      "step": 19287
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0231931753436002e-05,
+      "loss": 0.9763,
+      "step": 19288
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0231079215250652e-05,
+      "loss": 1.0521,
+      "step": 19289
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0230226675384868e-05,
+      "loss": 1.0569,
+      "step": 19290
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0229374133844851e-05,
+      "loss": 0.9194,
+      "step": 19291
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02285215906368e-05,
+      "loss": 0.7883,
+      "step": 19292
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0227669045766917e-05,
+      "loss": 0.8955,
+      "step": 19293
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0226816499241396e-05,
+      "loss": 1.1079,
+      "step": 19294
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0225963951066442e-05,
+      "loss": 0.694,
+      "step": 19295
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0225111401248252e-05,
+      "loss": 1.039,
+      "step": 19296
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0224258849793026e-05,
+      "loss": 1.1983,
+      "step": 19297
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0223406296706969e-05,
+      "loss": 0.9403,
+      "step": 19298
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0222553741996273e-05,
+      "loss": 0.7092,
+      "step": 19299
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.022170118566714e-05,
+      "loss": 0.885,
+      "step": 19300
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0220848627725775e-05,
+      "loss": 1.0036,
+      "step": 19301
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0219996068178373e-05,
+      "loss": 0.9349,
+      "step": 19302
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0219143507031135e-05,
+      "loss": 1.2118,
+      "step": 19303
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021829094429026e-05,
+      "loss": 0.9915,
+      "step": 19304
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021743837996195e-05,
+      "loss": 1.2193,
+      "step": 19305
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0216585814052402e-05,
+      "loss": 0.5979,
+      "step": 19306
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0215733246567821e-05,
+      "loss": 0.9388,
+      "step": 19307
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0214880677514405e-05,
+      "loss": 0.8711,
+      "step": 19308
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0214028106898349e-05,
+      "loss": 0.6253,
+      "step": 19309
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021317553472586e-05,
+      "loss": 0.9571,
+      "step": 19310
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0212322961003132e-05,
+      "loss": 1.0168,
+      "step": 19311
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0211470385736368e-05,
+      "loss": 0.7537,
+      "step": 19312
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021061780893177e-05,
+      "loss": 1.1008,
+      "step": 19313
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0209765230595536e-05,
+      "loss": 0.9292,
+      "step": 19314
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0208912650733862e-05,
+      "loss": 0.8356,
+      "step": 19315
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0208060069352956e-05,
+      "loss": 0.8597,
+      "step": 19316
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0207207486459013e-05,
+      "loss": 0.9509,
+      "step": 19317
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0206354902058234e-05,
+      "loss": 0.6841,
+      "step": 19318
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0205502316156819e-05,
+      "loss": 0.768,
+      "step": 19319
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0204649728760969e-05,
+      "loss": 0.777,
+      "step": 19320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0203797139876884e-05,
+      "loss": 0.8229,
+      "step": 19321
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020294454951076e-05,
+      "loss": 0.8511,
+      "step": 19322
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0202091957668807e-05,
+      "loss": 0.6418,
+      "step": 19323
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0201239364357212e-05,
+      "loss": 1.104,
+      "step": 19324
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0200386769582187e-05,
+      "loss": 0.9419,
+      "step": 19325
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0199534173349925e-05,
+      "loss": 0.7921,
+      "step": 19326
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.019868157566663e-05,
+      "loss": 0.9059,
+      "step": 19327
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0197828976538497e-05,
+      "loss": 1.0331,
+      "step": 19328
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0196976375971732e-05,
+      "loss": 0.9953,
+      "step": 19329
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0196123773972536e-05,
+      "loss": 0.6979,
+      "step": 19330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0195271170547098e-05,
+      "loss": 1.0607,
+      "step": 19331
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0194418565701634e-05,
+      "loss": 1.0471,
+      "step": 19332
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0193565959442333e-05,
+      "loss": 0.7938,
+      "step": 19333
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01927133517754e-05,
+      "loss": 0.9981,
+      "step": 19334
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0191860742707034e-05,
+      "loss": 1.3772,
+      "step": 19335
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0191008132243437e-05,
+      "loss": 0.5902,
+      "step": 19336
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0190155520390803e-05,
+      "loss": 0.9054,
+      "step": 19337
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0189302907155342e-05,
+      "loss": 0.8626,
+      "step": 19338
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0188450292543248e-05,
+      "loss": 0.7454,
+      "step": 19339
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0187597676560718e-05,
+      "loss": 0.8805,
+      "step": 19340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0186745059213963e-05,
+      "loss": 0.8033,
+      "step": 19341
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0185892440509173e-05,
+      "loss": 0.9513,
+      "step": 19342
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0185039820452555e-05,
+      "loss": 0.8999,
+      "step": 19343
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0184187199050306e-05,
+      "loss": 0.492,
+      "step": 19344
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0183334576308628e-05,
+      "loss": 0.6239,
+      "step": 19345
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0182481952233719e-05,
+      "loss": 0.854,
+      "step": 19346
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0181629326831781e-05,
+      "loss": 0.9576,
+      "step": 19347
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0180776700109018e-05,
+      "loss": 0.5276,
+      "step": 19348
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0179924072071622e-05,
+      "loss": 0.6521,
+      "step": 19349
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0179071442725801e-05,
+      "loss": 0.854,
+      "step": 19350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0178218812077752e-05,
+      "loss": 1.0577,
+      "step": 19351
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0177366180133674e-05,
+      "loss": 0.6632,
+      "step": 19352
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0176513546899774e-05,
+      "loss": 0.9734,
+      "step": 19353
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0175660912382244e-05,
+      "loss": 0.8348,
+      "step": 19354
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0174808276587289e-05,
+      "loss": 0.9203,
+      "step": 19355
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017395563952111e-05,
+      "loss": 1.0239,
+      "step": 19356
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0173103001189904e-05,
+      "loss": 0.6863,
+      "step": 19357
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0172250361599875e-05,
+      "loss": 0.9091,
+      "step": 19358
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0171397720757224e-05,
+      "loss": 0.8906,
+      "step": 19359
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0170545078668148e-05,
+      "loss": 0.9821,
+      "step": 19360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0169692435338848e-05,
+      "loss": 0.8904,
+      "step": 19361
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0168839790775526e-05,
+      "loss": 0.9413,
+      "step": 19362
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0167987144984388e-05,
+      "loss": 0.9201,
+      "step": 19363
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0167134497971623e-05,
+      "loss": 0.9279,
+      "step": 19364
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0166281849743438e-05,
+      "loss": 0.9465,
+      "step": 19365
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0165429200306033e-05,
+      "loss": 1.0336,
+      "step": 19366
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0164576549665607e-05,
+      "loss": 0.8847,
+      "step": 19367
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0163723897828368e-05,
+      "loss": 1.3975,
+      "step": 19368
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0162871244800507e-05,
+      "loss": 0.9942,
+      "step": 19369
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0162018590588228e-05,
+      "loss": 0.7464,
+      "step": 19370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.016116593519773e-05,
+      "loss": 0.8613,
+      "step": 19371
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0160313278635219e-05,
+      "loss": 1.0302,
+      "step": 19372
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0159460620906889e-05,
+      "loss": 0.8918,
+      "step": 19373
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0158607962018948e-05,
+      "loss": 1.1219,
+      "step": 19374
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0157755301977588e-05,
+      "loss": 0.8727,
+      "step": 19375
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0156902640789013e-05,
+      "loss": 1.0846,
+      "step": 19376
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0156049978459428e-05,
+      "loss": 1.3293,
+      "step": 19377
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0155197314995032e-05,
+      "loss": 1.0612,
+      "step": 19378
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0154344650402017e-05,
+      "loss": 0.9765,
+      "step": 19379
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0153491984686595e-05,
+      "loss": 0.7809,
+      "step": 19380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0152639317854961e-05,
+      "loss": 0.6515,
+      "step": 19381
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015178664991332e-05,
+      "loss": 0.7314,
+      "step": 19382
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0150933980867868e-05,
+      "loss": 0.8478,
+      "step": 19383
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0150081310724807e-05,
+      "loss": 1.1084,
+      "step": 19384
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0149228639490333e-05,
+      "loss": 0.9417,
+      "step": 19385
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0148375967170658e-05,
+      "loss": 0.8998,
+      "step": 19386
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0147523293771975e-05,
+      "loss": 1.0763,
+      "step": 19387
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0146670619300488e-05,
+      "loss": 0.8212,
+      "step": 19388
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0145817943762392e-05,
+      "loss": 1.0937,
+      "step": 19389
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0144965267163895e-05,
+      "loss": 1.3528,
+      "step": 19390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0144112589511195e-05,
+      "loss": 0.8983,
+      "step": 19391
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0143259910810487e-05,
+      "loss": 1.2543,
+      "step": 19392
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0142407231067983e-05,
+      "loss": 0.9559,
+      "step": 19393
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0141554550289874e-05,
+      "loss": 0.6334,
+      "step": 19394
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0140701868482365e-05,
+      "loss": 0.8482,
+      "step": 19395
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0139849185651658e-05,
+      "loss": 0.5305,
+      "step": 19396
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0138996501803953e-05,
+      "loss": 0.9809,
+      "step": 19397
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0138143816945449e-05,
+      "loss": 0.7423,
+      "step": 19398
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0137291131082347e-05,
+      "loss": 0.7875,
+      "step": 19399
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0136438444220848e-05,
+      "loss": 0.9482,
+      "step": 19400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0135585756367154e-05,
+      "loss": 0.6776,
+      "step": 19401
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0134733067527467e-05,
+      "loss": 0.8001,
+      "step": 19402
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0133880377707983e-05,
+      "loss": 0.9109,
+      "step": 19403
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013302768691491e-05,
+      "loss": 0.9297,
+      "step": 19404
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0132174995154442e-05,
+      "loss": 0.7697,
+      "step": 19405
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0131322302432784e-05,
+      "loss": 0.7392,
+      "step": 19406
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0130469608756134e-05,
+      "loss": 1.0155,
+      "step": 19407
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0129616914130697e-05,
+      "loss": 0.7308,
+      "step": 19408
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0128764218562673e-05,
+      "loss": 0.9012,
+      "step": 19409
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0127911522058256e-05,
+      "loss": 0.8525,
+      "step": 19410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0127058824623654e-05,
+      "loss": 0.7396,
+      "step": 19411
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0126206126265068e-05,
+      "loss": 0.7037,
+      "step": 19412
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0125353426988695e-05,
+      "loss": 0.9374,
+      "step": 19413
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.012450072680074e-05,
+      "loss": 0.9395,
+      "step": 19414
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01236480257074e-05,
+      "loss": 1.0822,
+      "step": 19415
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0122795323714878e-05,
+      "loss": 1.0025,
+      "step": 19416
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0121942620829375e-05,
+      "loss": 0.6738,
+      "step": 19417
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0121089917057094e-05,
+      "loss": 0.634,
+      "step": 19418
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0120237212404231e-05,
+      "loss": 0.9808,
+      "step": 19419
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011938450687699e-05,
+      "loss": 0.9214,
+      "step": 19420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0118531800481573e-05,
+      "loss": 0.8911,
+      "step": 19421
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0117679093224176e-05,
+      "loss": 0.8295,
+      "step": 19422
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011682638511101e-05,
+      "loss": 0.7076,
+      "step": 19423
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0115973676148266e-05,
+      "loss": 0.9364,
+      "step": 19424
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0115120966342145e-05,
+      "loss": 0.9735,
+      "step": 19425
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0114268255698853e-05,
+      "loss": 0.8158,
+      "step": 19426
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0113415544224594e-05,
+      "loss": 0.9119,
+      "step": 19427
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011256283192556e-05,
+      "loss": 1.0383,
+      "step": 19428
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011171011880796e-05,
+      "loss": 0.8661,
+      "step": 19429
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0110857404877992e-05,
+      "loss": 0.9304,
+      "step": 19430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0110004690141853e-05,
+      "loss": 0.7153,
+      "step": 19431
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0109151974605749e-05,
+      "loss": 0.4964,
+      "step": 19432
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0108299258275882e-05,
+      "loss": 0.8038,
+      "step": 19433
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107446541158447e-05,
+      "loss": 0.9395,
+      "step": 19434
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0106593823259651e-05,
+      "loss": 0.9968,
+      "step": 19435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0105741104585693e-05,
+      "loss": 0.6307,
+      "step": 19436
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010488838514277e-05,
+      "loss": 0.6922,
+      "step": 19437
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0104035664937095e-05,
+      "loss": 0.7855,
+      "step": 19438
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0103182943974856e-05,
+      "loss": 0.8337,
+      "step": 19439
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0102330222262257e-05,
+      "loss": 1.0505,
+      "step": 19440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0101477499805506e-05,
+      "loss": 0.9527,
+      "step": 19441
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0100624776610798e-05,
+      "loss": 0.8464,
+      "step": 19442
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0099772052684334e-05,
+      "loss": 1.0727,
+      "step": 19443
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0098919328032317e-05,
+      "loss": 0.641,
+      "step": 19444
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009806660266095e-05,
+      "loss": 1.1027,
+      "step": 19445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0097213876576428e-05,
+      "loss": 0.7415,
+      "step": 19446
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0096361149784956e-05,
+      "loss": 1.0677,
+      "step": 19447
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0095508422292742e-05,
+      "loss": 0.9286,
+      "step": 19448
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094655694105972e-05,
+      "loss": 0.8648,
+      "step": 19449
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0093802965230859e-05,
+      "loss": 1.0224,
+      "step": 19450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00929502356736e-05,
+      "loss": 1.0125,
+      "step": 19451
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0092097505440395e-05,
+      "loss": 1.2109,
+      "step": 19452
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009124477453745e-05,
+      "loss": 1.0065,
+      "step": 19453
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009039204297096e-05,
+      "loss": 1.1459,
+      "step": 19454
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0089539310747127e-05,
+      "loss": 0.9945,
+      "step": 19455
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0088686577872157e-05,
+      "loss": 1.0509,
+      "step": 19456
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0087833844352251e-05,
+      "loss": 1.0154,
+      "step": 19457
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0086981110193603e-05,
+      "loss": 0.9729,
+      "step": 19458
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0086128375402422e-05,
+      "loss": 0.8535,
+      "step": 19459
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0085275639984905e-05,
+      "loss": 0.8487,
+      "step": 19460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0084422903947252e-05,
+      "loss": 0.8771,
+      "step": 19461
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0083570167295667e-05,
+      "loss": 0.9488,
+      "step": 19462
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0082717430036355e-05,
+      "loss": 0.9642,
+      "step": 19463
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081864692175506e-05,
+      "loss": 1.1288,
+      "step": 19464
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081011953719331e-05,
+      "loss": 1.1505,
+      "step": 19465
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0080159214674027e-05,
+      "loss": 1.0315,
+      "step": 19466
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0079306475045797e-05,
+      "loss": 0.9812,
+      "step": 19467
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0078453734840843e-05,
+      "loss": 0.8493,
+      "step": 19468
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0077600994065365e-05,
+      "loss": 0.6725,
+      "step": 19469
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.007674825272556e-05,
+      "loss": 0.7756,
+      "step": 19470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0075895510827636e-05,
+      "loss": 1.1133,
+      "step": 19471
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0075042768377792e-05,
+      "loss": 0.9305,
+      "step": 19472
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0074190025382227e-05,
+      "loss": 0.8138,
+      "step": 19473
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0073337281847144e-05,
+      "loss": 0.6673,
+      "step": 19474
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0072484537778747e-05,
+      "loss": 0.9829,
+      "step": 19475
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0071631793183233e-05,
+      "loss": 0.9764,
+      "step": 19476
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0070779048066798e-05,
+      "loss": 0.7777,
+      "step": 19477
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006992630243566e-05,
+      "loss": 1.2699,
+      "step": 19478
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0069073556296006e-05,
+      "loss": 0.6446,
+      "step": 19479
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0068220809654038e-05,
+      "loss": 0.7935,
+      "step": 19480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0067368062515965e-05,
+      "loss": 1.4193,
+      "step": 19481
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0066515314887983e-05,
+      "loss": 0.8567,
+      "step": 19482
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0065662566776295e-05,
+      "loss": 0.6983,
+      "step": 19483
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00648098181871e-05,
+      "loss": 0.4799,
+      "step": 19484
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0063957069126602e-05,
+      "loss": 0.5821,
+      "step": 19485
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0063104319601e-05,
+      "loss": 0.8928,
+      "step": 19486
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0062251569616497e-05,
+      "loss": 0.7335,
+      "step": 19487
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0061398819179295e-05,
+      "loss": 1.0365,
+      "step": 19488
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0060546068295591e-05,
+      "loss": 0.9196,
+      "step": 19489
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0059693316971591e-05,
+      "loss": 0.6789,
+      "step": 19490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0058840565213495e-05,
+      "loss": 1.0208,
+      "step": 19491
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00579878130275e-05,
+      "loss": 1.0695,
+      "step": 19492
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0057135060419819e-05,
+      "loss": 0.589,
+      "step": 19493
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.005628230739664e-05,
+      "loss": 1.031,
+      "step": 19494
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0055429553964169e-05,
+      "loss": 0.957,
+      "step": 19495
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0054576800128611e-05,
+      "loss": 0.7874,
+      "step": 19496
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0053724045896162e-05,
+      "loss": 0.747,
+      "step": 19497
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0052871291273027e-05,
+      "loss": 1.3958,
+      "step": 19498
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.005201853626541e-05,
+      "loss": 0.9356,
+      "step": 19499
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0051165780879503e-05,
+      "loss": 0.9562,
+      "step": 19500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0050313025121511e-05,
+      "loss": 1.0571,
+      "step": 19501
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0049460268997642e-05,
+      "loss": 1.0751,
+      "step": 19502
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0048607512514093e-05,
+      "loss": 0.7596,
+      "step": 19503
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0047754755677062e-05,
+      "loss": 0.9873,
+      "step": 19504
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0046901998492753e-05,
+      "loss": 0.7891,
+      "step": 19505
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0046049240967367e-05,
+      "loss": 0.7711,
+      "step": 19506
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0045196483107107e-05,
+      "loss": 0.9998,
+      "step": 19507
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0044343724918175e-05,
+      "loss": 0.7914,
+      "step": 19508
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004349096640677e-05,
+      "loss": 0.939,
+      "step": 19509
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.004263820757909e-05,
+      "loss": 1.2084,
+      "step": 19510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0041785448441343e-05,
+      "loss": 1.1081,
+      "step": 19511
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0040932688999727e-05,
+      "loss": 0.8595,
+      "step": 19512
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0040079929260448e-05,
+      "loss": 1.0337,
+      "step": 19513
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0039227169229698e-05,
+      "loss": 0.6396,
+      "step": 19514
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0038374408913684e-05,
+      "loss": 0.9815,
+      "step": 19515
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0037521648318607e-05,
+      "loss": 0.9275,
+      "step": 19516
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0036668887450672e-05,
+      "loss": 0.6769,
+      "step": 19517
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0035816126316075e-05,
+      "loss": 0.7981,
+      "step": 19518
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0034963364921018e-05,
+      "loss": 0.8735,
+      "step": 19519
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0034110603271705e-05,
+      "loss": 0.7178,
+      "step": 19520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0033257841374335e-05,
+      "loss": 1.2409,
+      "step": 19521
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003240507923511e-05,
+      "loss": 0.9611,
+      "step": 19522
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003155231686023e-05,
+      "loss": 0.7212,
+      "step": 19523
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0030699554255903e-05,
+      "loss": 0.8588,
+      "step": 19524
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002984679142832e-05,
+      "loss": 1.1329,
+      "step": 19525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0028994028383691e-05,
+      "loss": 1.1509,
+      "step": 19526
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0028141265128213e-05,
+      "loss": 1.1735,
+      "step": 19527
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002728850166809e-05,
+      "loss": 0.8802,
+      "step": 19528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002643573800952e-05,
+      "loss": 0.9114,
+      "step": 19529
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002558297415871e-05,
+      "loss": 0.8094,
+      "step": 19530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0024730210121855e-05,
+      "loss": 1.0577,
+      "step": 19531
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0023877445905156e-05,
+      "loss": 0.7844,
+      "step": 19532
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0023024681514826e-05,
+      "loss": 0.7957,
+      "step": 19533
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0022171916957051e-05,
+      "loss": 1.0369,
+      "step": 19534
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002131915223804e-05,
+      "loss": 0.6938,
+      "step": 19535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0020466387363996e-05,
+      "loss": 1.0846,
+      "step": 19536
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0019613622341117e-05,
+      "loss": 0.9763,
+      "step": 19537
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0018760857175606e-05,
+      "loss": 0.9761,
+      "step": 19538
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0017908091873665e-05,
+      "loss": 0.9967,
+      "step": 19539
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0017055326441493e-05,
+      "loss": 0.854,
+      "step": 19540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0016202560885292e-05,
+      "loss": 0.8084,
+      "step": 19541
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0015349795211264e-05,
+      "loss": 0.9521,
+      "step": 19542
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0014497029425612e-05,
+      "loss": 0.7462,
+      "step": 19543
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0013644263534537e-05,
+      "loss": 1.1758,
+      "step": 19544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0012791497544238e-05,
+      "loss": 0.8495,
+      "step": 19545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0011938731460918e-05,
+      "loss": 1.0709,
+      "step": 19546
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0011085965290777e-05,
+      "loss": 0.7193,
+      "step": 19547
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001023319904002e-05,
+      "loss": 0.8914,
+      "step": 19548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0009380432714846e-05,
+      "loss": 0.588,
+      "step": 19549
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0008527666321453e-05,
+      "loss": 0.9543,
+      "step": 19550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000767489986605e-05,
+      "loss": 0.9623,
+      "step": 19551
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000682213335483e-05,
+      "loss": 0.8745,
+      "step": 19552
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0005969366794002e-05,
+      "loss": 0.9179,
+      "step": 19553
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0005116600189764e-05,
+      "loss": 0.8421,
+      "step": 19554
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0004263833548317e-05,
+      "loss": 1.1335,
+      "step": 19555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0003411066875862e-05,
+      "loss": 0.878,
+      "step": 19556
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00025583001786e-05,
+      "loss": 0.8026,
+      "step": 19557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0001705533462737e-05,
+      "loss": 1.0867,
+      "step": 19558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0000852766734471e-05,
+      "loss": 0.8641,
+      "step": 19559
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1e-05,
+      "loss": 0.6934,
+      "step": 19560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.999147233265534e-06,
+      "loss": 0.8542,
+      "step": 19561
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998294466537264e-06,
+      "loss": 0.8787,
+      "step": 19562
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9974416998214e-06,
+      "loss": 0.7469,
+      "step": 19563
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99658893312414e-06,
+      "loss": 1.097,
+      "step": 19564
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.995736166451688e-06,
+      "loss": 1.0661,
+      "step": 19565
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99488339981024e-06,
+      "loss": 0.9291,
+      "step": 19566
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.994030633206003e-06,
+      "loss": 1.0683,
+      "step": 19567
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.993177866645172e-06,
+      "loss": 0.7627,
+      "step": 19568
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.992325100133956e-06,
+      "loss": 1.2104,
+      "step": 19569
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99147233367855e-06,
+      "loss": 0.9861,
+      "step": 19570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990619567285157e-06,
+      "loss": 0.843,
+      "step": 19571
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.989766800959982e-06,
+      "loss": 1.0809,
+      "step": 19572
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988914034709223e-06,
+      "loss": 0.7444,
+      "step": 19573
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988061268539084e-06,
+      "loss": 0.7688,
+      "step": 19574
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.987208502455767e-06,
+      "loss": 0.8706,
+      "step": 19575
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.986355736465468e-06,
+      "loss": 0.7471,
+      "step": 19576
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98550297057439e-06,
+      "loss": 0.7392,
+      "step": 19577
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.984650204788739e-06,
+      "loss": 0.8077,
+      "step": 19578
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.983797439114712e-06,
+      "loss": 0.9492,
+      "step": 19579
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.982944673558508e-06,
+      "loss": 0.9058,
+      "step": 19580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.982091908126338e-06,
+      "loss": 0.9163,
+      "step": 19581
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.981239142824395e-06,
+      "loss": 1.2512,
+      "step": 19582
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.980386377658884e-06,
+      "loss": 0.9151,
+      "step": 19583
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.979533612636003e-06,
+      "loss": 0.6822,
+      "step": 19584
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.978680847761965e-06,
+      "loss": 0.9059,
+      "step": 19585
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977828083042954e-06,
+      "loss": 0.8286,
+      "step": 19586
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.97697531848518e-06,
+      "loss": 0.9456,
+      "step": 19587
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.976122554094845e-06,
+      "loss": 0.9635,
+      "step": 19588
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.975269789878148e-06,
+      "loss": 0.8492,
+      "step": 19589
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.974417025841293e-06,
+      "loss": 0.8677,
+      "step": 19590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.973564261990481e-06,
+      "loss": 0.6739,
+      "step": 19591
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.972711498331913e-06,
+      "loss": 0.8656,
+      "step": 19592
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.971858734871787e-06,
+      "loss": 1.3255,
+      "step": 19593
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.97100597161631e-06,
+      "loss": 0.6246,
+      "step": 19594
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.970153208571682e-06,
+      "loss": 0.8226,
+      "step": 19595
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.969300445744102e-06,
+      "loss": 1.0223,
+      "step": 19596
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.968447683139773e-06,
+      "loss": 1.0526,
+      "step": 19597
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.967594920764893e-06,
+      "loss": 0.8815,
+      "step": 19598
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.966742158625668e-06,
+      "loss": 0.9687,
+      "step": 19599
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.965889396728298e-06,
+      "loss": 0.7371,
+      "step": 19600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.965036635078985e-06,
+      "loss": 0.9243,
+      "step": 19601
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.964183873683926e-06,
+      "loss": 0.8112,
+      "step": 19602
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96333111254933e-06,
+      "loss": 0.8978,
+      "step": 19603
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.962478351681393e-06,
+      "loss": 0.8854,
+      "step": 19604
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.961625591086321e-06,
+      "loss": 1.1154,
+      "step": 19605
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.960772830770307e-06,
+      "loss": 1.0327,
+      "step": 19606
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.959920070739557e-06,
+      "loss": 0.9274,
+      "step": 19607
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.959067311000275e-06,
+      "loss": 0.6971,
+      "step": 19608
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.958214551558659e-06,
+      "loss": 0.9707,
+      "step": 19609
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.957361792420912e-06,
+      "loss": 1.0343,
+      "step": 19610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.956509033593233e-06,
+      "loss": 0.8219,
+      "step": 19611
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.955656275081827e-06,
+      "loss": 1.3384,
+      "step": 19612
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.954803516892894e-06,
+      "loss": 0.8802,
+      "step": 19613
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953950759032632e-06,
+      "loss": 0.9738,
+      "step": 19614
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953098001507252e-06,
+      "loss": 0.8016,
+      "step": 19615
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.952245244322941e-06,
+      "loss": 0.6742,
+      "step": 19616
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.951392487485912e-06,
+      "loss": 0.743,
+      "step": 19617
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.950539731002362e-06,
+      "loss": 1.2183,
+      "step": 19618
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94968697487849e-06,
+      "loss": 0.8919,
+      "step": 19619
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9488342191205e-06,
+      "loss": 0.6328,
+      "step": 19620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.947981463734596e-06,
+      "loss": 0.9039,
+      "step": 19621
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.947128708726975e-06,
+      "loss": 1.1002,
+      "step": 19622
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.946275954103838e-06,
+      "loss": 1.2654,
+      "step": 19623
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94542319987139e-06,
+      "loss": 1.0862,
+      "step": 19624
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.944570446035833e-06,
+      "loss": 0.8074,
+      "step": 19625
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.943717692603365e-06,
+      "loss": 0.6837,
+      "step": 19626
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.942864939580186e-06,
+      "loss": 1.0571,
+      "step": 19627
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.942012186972501e-06,
+      "loss": 0.8329,
+      "step": 19628
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.941159434786508e-06,
+      "loss": 1.0092,
+      "step": 19629
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94030668302841e-06,
+      "loss": 1.0325,
+      "step": 19630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.939453931704412e-06,
+      "loss": 0.8785,
+      "step": 19631
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.938601180820707e-06,
+      "loss": 0.9966,
+      "step": 19632
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.937748430383503e-06,
+      "loss": 1.3698,
+      "step": 19633
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936895680399001e-06,
+      "loss": 0.8524,
+      "step": 19634
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936042930873403e-06,
+      "loss": 0.679,
+      "step": 19635
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.935190181812903e-06,
+      "loss": 1.0822,
+      "step": 19636
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.93433743322371e-06,
+      "loss": 0.9017,
+      "step": 19637
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.933484685112019e-06,
+      "loss": 0.7252,
+      "step": 19638
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.932631937484037e-06,
+      "loss": 0.7217,
+      "step": 19639
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.931779190345963e-06,
+      "loss": 1.2108,
+      "step": 19640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.930926443703996e-06,
+      "loss": 0.7693,
+      "step": 19641
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.930073697564342e-06,
+      "loss": 0.7397,
+      "step": 19642
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9292209519332e-06,
+      "loss": 1.1192,
+      "step": 19643
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.928368206816772e-06,
+      "loss": 0.9409,
+      "step": 19644
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.927515462221258e-06,
+      "loss": 1.2949,
+      "step": 19645
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.926662718152861e-06,
+      "loss": 1.1604,
+      "step": 19646
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.925809974617777e-06,
+      "loss": 1.1846,
+      "step": 19647
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.924957231622213e-06,
+      "loss": 0.9295,
+      "step": 19648
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.924104489172367e-06,
+      "loss": 1.1215,
+      "step": 19649
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.923251747274441e-06,
+      "loss": 0.8318,
+      "step": 19650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.922399005934639e-06,
+      "loss": 0.7079,
+      "step": 19651
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921546265159159e-06,
+      "loss": 1.0808,
+      "step": 19652
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.920693524954204e-06,
+      "loss": 0.8978,
+      "step": 19653
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.919840785325972e-06,
+      "loss": 1.0479,
+      "step": 19654
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918988046280674e-06,
+      "loss": 0.998,
+      "step": 19655
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918135307824497e-06,
+      "loss": 0.916,
+      "step": 19656
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91728256996365e-06,
+      "loss": 0.9586,
+      "step": 19657
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.916429832704334e-06,
+      "loss": 1.2611,
+      "step": 19658
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91557709605275e-06,
+      "loss": 0.9992,
+      "step": 19659
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.914724360015099e-06,
+      "loss": 0.8897,
+      "step": 19660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.913871624597581e-06,
+      "loss": 0.9713,
+      "step": 19661
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.913018889806399e-06,
+      "loss": 0.9764,
+      "step": 19662
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91216615564775e-06,
+      "loss": 0.6938,
+      "step": 19663
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.911313422127843e-06,
+      "loss": 0.622,
+      "step": 19664
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.910460689252876e-06,
+      "loss": 1.1003,
+      "step": 19665
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.909607957029046e-06,
+      "loss": 1.0247,
+      "step": 19666
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.908755225462557e-06,
+      "loss": 0.8612,
+      "step": 19667
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907902494559606e-06,
+      "loss": 0.7704,
+      "step": 19668
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907049764326405e-06,
+      "loss": 0.6915,
+      "step": 19669
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.906197034769144e-06,
+      "loss": 1.0504,
+      "step": 19670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.905344305894031e-06,
+      "loss": 0.7922,
+      "step": 19671
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.904491577707263e-06,
+      "loss": 0.7517,
+      "step": 19672
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.903638850215043e-06,
+      "loss": 0.7584,
+      "step": 19673
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.902786123423574e-06,
+      "loss": 0.9226,
+      "step": 19674
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.901933397339055e-06,
+      "loss": 1.0147,
+      "step": 19675
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.901080671967686e-06,
+      "loss": 0.8218,
+      "step": 19676
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.900227947315668e-06,
+      "loss": 0.8265,
+      "step": 19677
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.899375223389206e-06,
+      "loss": 0.7172,
+      "step": 19678
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.898522500194497e-06,
+      "loss": 0.6753,
+      "step": 19679
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.897669777737745e-06,
+      "loss": 0.9086,
+      "step": 19680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.896817056025146e-06,
+      "loss": 1.0148,
+      "step": 19681
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.895964335062908e-06,
+      "loss": 1.1812,
+      "step": 19682
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89511161485723e-06,
+      "loss": 0.8743,
+      "step": 19683
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.894258895414308e-06,
+      "loss": 1.1777,
+      "step": 19684
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.893406176740354e-06,
+      "loss": 1.0533,
+      "step": 19685
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892553458841556e-06,
+      "loss": 1.598,
+      "step": 19686
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.891700741724123e-06,
+      "loss": 0.7641,
+      "step": 19687
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.890848025394254e-06,
+      "loss": 1.0851,
+      "step": 19688
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.889995309858152e-06,
+      "loss": 0.9055,
+      "step": 19689
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.889142595122013e-06,
+      "loss": 0.9372,
+      "step": 19690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.888289881192043e-06,
+      "loss": 0.762,
+      "step": 19691
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.887437168074441e-06,
+      "loss": 1.1611,
+      "step": 19692
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.886584455775407e-06,
+      "loss": 0.805,
+      "step": 19693
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.885731744301146e-06,
+      "loss": 0.8335,
+      "step": 19694
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.884879033657859e-06,
+      "loss": 0.9491,
+      "step": 19695
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.884026323851741e-06,
+      "loss": 1.0726,
+      "step": 19696
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.883173614888997e-06,
+      "loss": 0.8693,
+      "step": 19697
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.882320906775828e-06,
+      "loss": 0.8051,
+      "step": 19698
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.88146819951843e-06,
+      "loss": 1.0144,
+      "step": 19699
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.880615493123013e-06,
+      "loss": 0.8781,
+      "step": 19700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.879762787595772e-06,
+      "loss": 0.8082,
+      "step": 19701
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.878910082942908e-06,
+      "loss": 0.9893,
+      "step": 19702
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.878057379170626e-06,
+      "loss": 0.9796,
+      "step": 19703
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.877204676285124e-06,
+      "loss": 1.2413,
+      "step": 19704
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.876351974292604e-06,
+      "loss": 1.0893,
+      "step": 19705
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.875499273199265e-06,
+      "loss": 0.9312,
+      "step": 19706
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87464657301131e-06,
+      "loss": 1.1476,
+      "step": 19707
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.873793873734935e-06,
+      "loss": 0.6786,
+      "step": 19708
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.872941175376348e-06,
+      "loss": 0.7352,
+      "step": 19709
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.872088477941748e-06,
+      "loss": 0.8614,
+      "step": 19710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.871235781437332e-06,
+      "loss": 1.07,
+      "step": 19711
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.870383085869305e-06,
+      "loss": 0.8928,
+      "step": 19712
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.869530391243867e-06,
+      "loss": 0.7587,
+      "step": 19713
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868677697567219e-06,
+      "loss": 0.3324,
+      "step": 19714
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.867825004845563e-06,
+      "loss": 1.0019,
+      "step": 19715
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866972313085096e-06,
+      "loss": 0.6618,
+      "step": 19716
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866119622292019e-06,
+      "loss": 0.8333,
+      "step": 19717
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.865266932472538e-06,
+      "loss": 1.0049,
+      "step": 19718
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.864414243632849e-06,
+      "loss": 1.195,
+      "step": 19719
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.863561555779154e-06,
+      "loss": 1.0435,
+      "step": 19720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.862708868917656e-06,
+      "loss": 0.9588,
+      "step": 19721
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.861856183054554e-06,
+      "loss": 0.8359,
+      "step": 19722
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.86100349819605e-06,
+      "loss": 1.1884,
+      "step": 19723
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.860150814348342e-06,
+      "loss": 1.0035,
+      "step": 19724
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.859298131517639e-06,
+      "loss": 0.8726,
+      "step": 19725
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.858445449710129e-06,
+      "loss": 0.7853,
+      "step": 19726
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.857592768932022e-06,
+      "loss": 0.8185,
+      "step": 19727
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.856740089189516e-06,
+      "loss": 1.0198,
+      "step": 19728
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85588741048881e-06,
+      "loss": 0.8121,
+      "step": 19729
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.855034732836107e-06,
+      "loss": 0.9119,
+      "step": 19730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85418205623761e-06,
+      "loss": 1.0026,
+      "step": 19731
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.853329380699518e-06,
+      "loss": 0.9748,
+      "step": 19732
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.852476706228026e-06,
+      "loss": 0.9927,
+      "step": 19733
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.851624032829342e-06,
+      "loss": 1.1664,
+      "step": 19734
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.850771360509668e-06,
+      "loss": 0.9545,
+      "step": 19735
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8499186892752e-06,
+      "loss": 0.985,
+      "step": 19736
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849066019132139e-06,
+      "loss": 0.7103,
+      "step": 19737
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.848213350086683e-06,
+      "loss": 0.9394,
+      "step": 19738
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.84736068214504e-06,
+      "loss": 0.8768,
+      "step": 19739
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.846508015313407e-06,
+      "loss": 0.8714,
+      "step": 19740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.845655349597985e-06,
+      "loss": 0.799,
+      "step": 19741
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.844802685004973e-06,
+      "loss": 0.8408,
+      "step": 19742
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.843950021540574e-06,
+      "loss": 0.6084,
+      "step": 19743
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.843097359210988e-06,
+      "loss": 0.9666,
+      "step": 19744
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.842244698022417e-06,
+      "loss": 0.8076,
+      "step": 19745
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841392037981059e-06,
+      "loss": 0.9024,
+      "step": 19746
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.840539379093114e-06,
+      "loss": 1.1054,
+      "step": 19747
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.839686721364784e-06,
+      "loss": 1.1398,
+      "step": 19748
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838834064802274e-06,
+      "loss": 0.8111,
+      "step": 19749
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.837981409411777e-06,
+      "loss": 1.0393,
+      "step": 19750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.837128755199496e-06,
+      "loss": 1.1701,
+      "step": 19751
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.836276102171635e-06,
+      "loss": 0.9289,
+      "step": 19752
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.835423450334393e-06,
+      "loss": 1.131,
+      "step": 19753
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.834570799693967e-06,
+      "loss": 0.8991,
+      "step": 19754
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.833718150256567e-06,
+      "loss": 1.138,
+      "step": 19755
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83286550202838e-06,
+      "loss": 0.7018,
+      "step": 19756
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.832012855015617e-06,
+      "loss": 0.7404,
+      "step": 19757
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.831160209224476e-06,
+      "loss": 0.8144,
+      "step": 19758
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830307564661154e-06,
+      "loss": 0.8318,
+      "step": 19759
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.829454921331854e-06,
+      "loss": 0.8707,
+      "step": 19760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82860227924278e-06,
+      "loss": 0.6672,
+      "step": 19761
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.827749638400126e-06,
+      "loss": 0.7257,
+      "step": 19762
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.826896998810096e-06,
+      "loss": 0.8035,
+      "step": 19763
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.826044360478892e-06,
+      "loss": 1.0209,
+      "step": 19764
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.825191723412714e-06,
+      "loss": 1.2056,
+      "step": 19765
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82433908761776e-06,
+      "loss": 0.8801,
+      "step": 19766
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823486453100231e-06,
+      "loss": 1.1249,
+      "step": 19767
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.822633819866329e-06,
+      "loss": 0.9568,
+      "step": 19768
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.821781187922251e-06,
+      "loss": 1.0583,
+      "step": 19769
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820928557274202e-06,
+      "loss": 0.8221,
+      "step": 19770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820075927928381e-06,
+      "loss": 0.9269,
+      "step": 19771
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.819223299890984e-06,
+      "loss": 1.0979,
+      "step": 19772
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.818370673168219e-06,
+      "loss": 0.6085,
+      "step": 19773
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.817518047766283e-06,
+      "loss": 0.782,
+      "step": 19774
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.816665423691377e-06,
+      "loss": 0.9638,
+      "step": 19775
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8158128009497e-06,
+      "loss": 0.8546,
+      "step": 19776
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.81496017954745e-06,
+      "loss": 0.9134,
+      "step": 19777
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.814107559490829e-06,
+      "loss": 1.0227,
+      "step": 19778
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.81325494078604e-06,
+      "loss": 0.8355,
+      "step": 19779
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812402323439283e-06,
+      "loss": 0.9079,
+      "step": 19780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.811549707456756e-06,
+      "loss": 0.8365,
+      "step": 19781
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810697092844661e-06,
+      "loss": 0.8439,
+      "step": 19782
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.809844479609199e-06,
+      "loss": 1.1617,
+      "step": 19783
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.808991867756568e-06,
+      "loss": 1.1786,
+      "step": 19784
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.808139257292971e-06,
+      "loss": 0.9125,
+      "step": 19785
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.807286648224606e-06,
+      "loss": 0.9877,
+      "step": 19786
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80643404055767e-06,
+      "loss": 0.8137,
+      "step": 19787
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80558143429837e-06,
+      "loss": 0.7658,
+      "step": 19788
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.804728829452905e-06,
+      "loss": 0.8894,
+      "step": 19789
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80387622602747e-06,
+      "loss": 0.7924,
+      "step": 19790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80302362402827e-06,
+      "loss": 0.9088,
+      "step": 19791
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.802171023461505e-06,
+      "loss": 1.2022,
+      "step": 19792
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.801318424333374e-06,
+      "loss": 0.6812,
+      "step": 19793
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.800465826650077e-06,
+      "loss": 0.5678,
+      "step": 19794
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79961323041782e-06,
+      "loss": 1.0572,
+      "step": 19795
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79876063564279e-06,
+      "loss": 0.8279,
+      "step": 19796
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.797908042331198e-06,
+      "loss": 1.2504,
+      "step": 19797
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.797055450489242e-06,
+      "loss": 0.8879,
+      "step": 19798
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79620286012312e-06,
+      "loss": 0.7017,
+      "step": 19799
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.795350271239034e-06,
+      "loss": 0.9844,
+      "step": 19800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.794497683843184e-06,
+      "loss": 0.559,
+      "step": 19801
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.793645097941768e-06,
+      "loss": 1.1044,
+      "step": 19802
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.792792513540989e-06,
+      "loss": 0.6911,
+      "step": 19803
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791939930647046e-06,
+      "loss": 0.666,
+      "step": 19804
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791087349266141e-06,
+      "loss": 0.8491,
+      "step": 19805
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79023476940447e-06,
+      "loss": 0.8318,
+      "step": 19806
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.789382191068235e-06,
+      "loss": 1.0693,
+      "step": 19807
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788529614263634e-06,
+      "loss": 0.8607,
+      "step": 19808
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.787677038996872e-06,
+      "loss": 1.06,
+      "step": 19809
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.786824465274146e-06,
+      "loss": 0.9326,
+      "step": 19810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.785971893101655e-06,
+      "loss": 0.8247,
+      "step": 19811
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.785119322485598e-06,
+      "loss": 1.1364,
+      "step": 19812
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78426675343218e-06,
+      "loss": 1.2038,
+      "step": 19813
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.783414185947598e-06,
+      "loss": 0.9435,
+      "step": 19814
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.782561620038055e-06,
+      "loss": 1.1325,
+      "step": 19815
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.781709055709744e-06,
+      "loss": 0.7762,
+      "step": 19816
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78085649296887e-06,
+      "loss": 0.8157,
+      "step": 19817
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78000393182163e-06,
+      "loss": 1.156,
+      "step": 19818
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77915137227423e-06,
+      "loss": 1.1619,
+      "step": 19819
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.778298814332863e-06,
+      "loss": 0.8314,
+      "step": 19820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77744625800373e-06,
+      "loss": 0.7597,
+      "step": 19821
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.776593703293035e-06,
+      "loss": 1.058,
+      "step": 19822
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.775741150206976e-06,
+      "loss": 0.7807,
+      "step": 19823
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77488859875175e-06,
+      "loss": 0.928,
+      "step": 19824
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.774036048933564e-06,
+      "loss": 0.6613,
+      "step": 19825
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77318350075861e-06,
+      "loss": 1.1124,
+      "step": 19826
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.772330954233086e-06,
+      "loss": 0.8234,
+      "step": 19827
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.771478409363201e-06,
+      "loss": 0.8629,
+      "step": 19828
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.770625866155152e-06,
+      "loss": 0.594,
+      "step": 19829
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.769773324615133e-06,
+      "loss": 0.8847,
+      "step": 19830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768920784749351e-06,
+      "loss": 0.5471,
+      "step": 19831
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768068246564003e-06,
+      "loss": 0.8255,
+      "step": 19832
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.767215710065287e-06,
+      "loss": 0.8651,
+      "step": 19833
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.766363175259403e-06,
+      "loss": 0.7714,
+      "step": 19834
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.765510642152558e-06,
+      "loss": 0.6068,
+      "step": 19835
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.764658110750938e-06,
+      "loss": 0.7224,
+      "step": 19836
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.763805581060755e-06,
+      "loss": 1.1441,
+      "step": 19837
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.762953053088203e-06,
+      "loss": 1.1227,
+      "step": 19838
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.76210052683948e-06,
+      "loss": 0.7515,
+      "step": 19839
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.76124800232079e-06,
+      "loss": 0.838,
+      "step": 19840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.760395479538333e-06,
+      "loss": 0.8754,
+      "step": 19841
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.759542958498304e-06,
+      "loss": 0.8591,
+      "step": 19842
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.758690439206906e-06,
+      "loss": 0.8046,
+      "step": 19843
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.757837921670339e-06,
+      "loss": 1.0247,
+      "step": 19844
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.756985405894802e-06,
+      "loss": 0.8129,
+      "step": 19845
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.756132891886495e-06,
+      "loss": 0.7718,
+      "step": 19846
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.755280379651615e-06,
+      "loss": 0.936,
+      "step": 19847
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75442786919636e-06,
+      "loss": 0.7649,
+      "step": 19848
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753575360526937e-06,
+      "loss": 0.8555,
+      "step": 19849
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.752722853649541e-06,
+      "loss": 1.206,
+      "step": 19850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751870348570372e-06,
+      "loss": 0.8918,
+      "step": 19851
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751017845295627e-06,
+      "loss": 0.915,
+      "step": 19852
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75016534383151e-06,
+      "loss": 0.512,
+      "step": 19853
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.749312844184217e-06,
+      "loss": 1.1613,
+      "step": 19854
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.748460346359953e-06,
+      "loss": 0.7165,
+      "step": 19855
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.74760785036491e-06,
+      "loss": 0.8397,
+      "step": 19856
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.74675535620529e-06,
+      "loss": 0.9069,
+      "step": 19857
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.745902863887293e-06,
+      "loss": 0.7941,
+      "step": 19858
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.74505037341712e-06,
+      "loss": 0.8947,
+      "step": 19859
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.744197884800968e-06,
+      "loss": 0.8089,
+      "step": 19860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.743345398045036e-06,
+      "loss": 0.7687,
+      "step": 19861
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.742492913155527e-06,
+      "loss": 1.3002,
+      "step": 19862
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.741640430138638e-06,
+      "loss": 1.0763,
+      "step": 19863
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.740787949000567e-06,
+      "loss": 1.1885,
+      "step": 19864
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73993546974752e-06,
+      "loss": 0.95,
+      "step": 19865
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.739082992385685e-06,
+      "loss": 0.7947,
+      "step": 19866
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.738230516921272e-06,
+      "loss": 0.8182,
+      "step": 19867
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.737378043360473e-06,
+      "loss": 1.1098,
+      "step": 19868
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73652557170949e-06,
+      "loss": 0.785,
+      "step": 19869
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73567310197452e-06,
+      "loss": 0.8123,
+      "step": 19870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.734820634161769e-06,
+      "loss": 0.9513,
+      "step": 19871
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73396816827743e-06,
+      "loss": 1.0528,
+      "step": 19872
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.733115704327702e-06,
+      "loss": 0.6364,
+      "step": 19873
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73226324231879e-06,
+      "loss": 0.9135,
+      "step": 19874
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.731410782256889e-06,
+      "loss": 1.0482,
+      "step": 19875
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.730558324148197e-06,
+      "loss": 0.4801,
+      "step": 19876
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729705867998917e-06,
+      "loss": 0.9143,
+      "step": 19877
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.728853413815243e-06,
+      "loss": 0.6684,
+      "step": 19878
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.728000961603377e-06,
+      "loss": 1.0497,
+      "step": 19879
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72714851136952e-06,
+      "loss": 1.193,
+      "step": 19880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.726296063119868e-06,
+      "loss": 0.9468,
+      "step": 19881
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72544361686062e-06,
+      "loss": 1.035,
+      "step": 19882
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72459117259798e-06,
+      "loss": 0.8233,
+      "step": 19883
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.723738730338142e-06,
+      "loss": 1.2135,
+      "step": 19884
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72288629008731e-06,
+      "loss": 0.7014,
+      "step": 19885
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.722033851851677e-06,
+      "loss": 0.9305,
+      "step": 19886
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.721181415637443e-06,
+      "loss": 0.9907,
+      "step": 19887
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72032898145081e-06,
+      "loss": 0.8601,
+      "step": 19888
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.719476549297975e-06,
+      "loss": 1.1117,
+      "step": 19889
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.718624119185138e-06,
+      "loss": 1.044,
+      "step": 19890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.717771691118498e-06,
+      "loss": 0.7822,
+      "step": 19891
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716919265104252e-06,
+      "loss": 1.0517,
+      "step": 19892
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716066841148603e-06,
+      "loss": 1.1259,
+      "step": 19893
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.715214419257748e-06,
+      "loss": 1.1328,
+      "step": 19894
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.714361999437886e-06,
+      "loss": 1.0169,
+      "step": 19895
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.713509581695213e-06,
+      "loss": 0.6269,
+      "step": 19896
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.712657166035929e-06,
+      "loss": 0.8014,
+      "step": 19897
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.711804752466235e-06,
+      "loss": 0.7175,
+      "step": 19898
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71095234099233e-06,
+      "loss": 0.9105,
+      "step": 19899
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.710099931620408e-06,
+      "loss": 0.8854,
+      "step": 19900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.709247524356675e-06,
+      "loss": 0.9913,
+      "step": 19901
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.708395119207326e-06,
+      "loss": 0.6568,
+      "step": 19902
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.707542716178559e-06,
+      "loss": 0.9464,
+      "step": 19903
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.706690315276573e-06,
+      "loss": 0.9376,
+      "step": 19904
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.705837916507575e-06,
+      "loss": 1.1172,
+      "step": 19905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.704985519877749e-06,
+      "loss": 1.1456,
+      "step": 19906
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.704133125393302e-06,
+      "loss": 1.239,
+      "step": 19907
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.703280733060433e-06,
+      "loss": 1.011,
+      "step": 19908
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.702428342885336e-06,
+      "loss": 0.8237,
+      "step": 19909
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.701575954874218e-06,
+      "loss": 0.9725,
+      "step": 19910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.700723569033271e-06,
+      "loss": 1.0995,
+      "step": 19911
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.699871185368694e-06,
+      "loss": 1.0243,
+      "step": 19912
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.699018803886685e-06,
+      "loss": 0.7692,
+      "step": 19913
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69816642459345e-06,
+      "loss": 0.7513,
+      "step": 19914
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.697314047495182e-06,
+      "loss": 0.9142,
+      "step": 19915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.696461672598078e-06,
+      "loss": 0.9313,
+      "step": 19916
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69560929990834e-06,
+      "loss": 0.8904,
+      "step": 19917
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69475692943216e-06,
+      "loss": 1.0932,
+      "step": 19918
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693904561175746e-06,
+      "loss": 0.9892,
+      "step": 19919
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693052195145292e-06,
+      "loss": 0.8659,
+      "step": 19920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.692199831346996e-06,
+      "loss": 0.8457,
+      "step": 19921
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.691347469787053e-06,
+      "loss": 0.8877,
+      "step": 19922
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.690495110471671e-06,
+      "loss": 1.0005,
+      "step": 19923
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68964275340704e-06,
+      "loss": 0.8589,
+      "step": 19924
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.688790398599365e-06,
+      "loss": 0.9366,
+      "step": 19925
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68793804605484e-06,
+      "loss": 0.8633,
+      "step": 19926
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68708569577966e-06,
+      "loss": 0.5703,
+      "step": 19927
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68623334778003e-06,
+      "loss": 0.7665,
+      "step": 19928
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.685381002062146e-06,
+      "loss": 0.8296,
+      "step": 19929
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.684528658632205e-06,
+      "loss": 1.1205,
+      "step": 19930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.683676317496407e-06,
+      "loss": 0.7971,
+      "step": 19931
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68282397866095e-06,
+      "loss": 0.811,
+      "step": 19932
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.681971642132034e-06,
+      "loss": 0.755,
+      "step": 19933
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.681119307915851e-06,
+      "loss": 1.0674,
+      "step": 19934
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.680266976018613e-06,
+      "loss": 1.2097,
+      "step": 19935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.679414646446502e-06,
+      "loss": 0.8667,
+      "step": 19936
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.678562319205725e-06,
+      "loss": 1.2394,
+      "step": 19937
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.677709994302478e-06,
+      "loss": 0.6954,
+      "step": 19938
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.67685767174296e-06,
+      "loss": 0.9139,
+      "step": 19939
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.676005351533366e-06,
+      "loss": 0.945,
+      "step": 19940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.6751530336799e-06,
+      "loss": 0.7468,
+      "step": 19941
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.674300718188756e-06,
+      "loss": 0.8023,
+      "step": 19942
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673448405066134e-06,
+      "loss": 0.5826,
+      "step": 19943
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.672596094318232e-06,
+      "loss": 0.7302,
+      "step": 19944
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67174378595125e-06,
+      "loss": 1.018,
+      "step": 19945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67089147997138e-06,
+      "loss": 0.6982,
+      "step": 19946
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.670039176384825e-06,
+      "loss": 1.1114,
+      "step": 19947
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.669186875197783e-06,
+      "loss": 0.8978,
+      "step": 19948
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.668334576416446e-06,
+      "loss": 0.908,
+      "step": 19949
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.66748228004702e-06,
+      "loss": 0.5709,
+      "step": 19950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.6666299860957e-06,
+      "loss": 1.0064,
+      "step": 19951
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.665777694568682e-06,
+      "loss": 0.8149,
+      "step": 19952
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.664925405472168e-06,
+      "loss": 1.0238,
+      "step": 19953
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.664073118812353e-06,
+      "loss": 0.9729,
+      "step": 19954
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.663220834595438e-06,
+      "loss": 0.7767,
+      "step": 19955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662368552827617e-06,
+      "loss": 0.5827,
+      "step": 19956
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.661516273515088e-06,
+      "loss": 1.1668,
+      "step": 19957
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.660663996664048e-06,
+      "loss": 0.9056,
+      "step": 19958
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.6598117222807e-06,
+      "loss": 0.9461,
+      "step": 19959
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.65895945037124e-06,
+      "loss": 0.7327,
+      "step": 19960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.658107180941861e-06,
+      "loss": 0.8075,
+      "step": 19961
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.657254913998766e-06,
+      "loss": 1.0254,
+      "step": 19962
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.656402649548154e-06,
+      "loss": 1.0485,
+      "step": 19963
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.655550387596218e-06,
+      "loss": 0.998,
+      "step": 19964
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.654698128149162e-06,
+      "loss": 0.9275,
+      "step": 19965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.653845871213175e-06,
+      "loss": 1.3628,
+      "step": 19966
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652993616794459e-06,
+      "loss": 0.6842,
+      "step": 19967
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652141364899213e-06,
+      "loss": 1.1493,
+      "step": 19968
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.651289115533632e-06,
+      "loss": 0.8819,
+      "step": 19969
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.650436868703916e-06,
+      "loss": 0.8051,
+      "step": 19970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.649584624416263e-06,
+      "loss": 0.917,
+      "step": 19971
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.64873238267687e-06,
+      "loss": 0.6709,
+      "step": 19972
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647880143491933e-06,
+      "loss": 1.2509,
+      "step": 19973
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.64702790686765e-06,
+      "loss": 1.0483,
+      "step": 19974
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.646175672810224e-06,
+      "loss": 1.0306,
+      "step": 19975
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.645323441325843e-06,
+      "loss": 0.9733,
+      "step": 19976
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.644471212420711e-06,
+      "loss": 1.0462,
+      "step": 19977
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.643618986101023e-06,
+      "loss": 0.8999,
+      "step": 19978
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.642766762372975e-06,
+      "loss": 0.8062,
+      "step": 19979
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.64191454124277e-06,
+      "loss": 0.8432,
+      "step": 19980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641062322716603e-06,
+      "loss": 0.9051,
+      "step": 19981
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.640210106800669e-06,
+      "loss": 0.9073,
+      "step": 19982
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.639357893501165e-06,
+      "loss": 0.6557,
+      "step": 19983
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.638505682824294e-06,
+      "loss": 1.0228,
+      "step": 19984
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63765347477625e-06,
+      "loss": 0.5998,
+      "step": 19985
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63680126936323e-06,
+      "loss": 0.7969,
+      "step": 19986
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63594906659143e-06,
+      "loss": 0.9202,
+      "step": 19987
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.635096866467047e-06,
+      "loss": 0.8365,
+      "step": 19988
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634244668996282e-06,
+      "loss": 0.8037,
+      "step": 19989
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.633392474185332e-06,
+      "loss": 0.9631,
+      "step": 19990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63254028204039e-06,
+      "loss": 1.0051,
+      "step": 19991
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631688092567655e-06,
+      "loss": 0.8022,
+      "step": 19992
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.630835905773329e-06,
+      "loss": 0.8804,
+      "step": 19993
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.629983721663604e-06,
+      "loss": 0.9276,
+      "step": 19994
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62913154024468e-06,
+      "loss": 1.0434,
+      "step": 19995
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62827936152275e-06,
+      "loss": 0.851,
+      "step": 19996
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.627427185504013e-06,
+      "loss": 1.0172,
+      "step": 19997
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.626575012194668e-06,
+      "loss": 0.8929,
+      "step": 19998
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62572284160091e-06,
+      "loss": 0.568,
+      "step": 19999
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62487067372894e-06,
+      "loss": 1.1393,
+      "step": 20000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.624018508584949e-06,
+      "loss": 0.769,
+      "step": 20001
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.623166346175139e-06,
+      "loss": 1.0751,
+      "step": 20002
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.622314186505704e-06,
+      "loss": 1.1539,
+      "step": 20003
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.621462029582841e-06,
+      "loss": 0.7044,
+      "step": 20004
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.620609875412754e-06,
+      "loss": 1.0992,
+      "step": 20005
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.619757724001628e-06,
+      "loss": 1.0579,
+      "step": 20006
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61890557535567e-06,
+      "loss": 0.7163,
+      "step": 20007
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61805342948107e-06,
+      "loss": 0.6891,
+      "step": 20008
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.617201286384031e-06,
+      "loss": 0.5838,
+      "step": 20009
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61634914607074e-06,
+      "loss": 0.6901,
+      "step": 20010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.615497008547408e-06,
+      "loss": 1.1034,
+      "step": 20011
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.614644873820222e-06,
+      "loss": 0.8762,
+      "step": 20012
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.613792741895378e-06,
+      "loss": 1.0255,
+      "step": 20013
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61294061277908e-06,
+      "loss": 0.8539,
+      "step": 20014
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.612088486477521e-06,
+      "loss": 0.6216,
+      "step": 20015
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.611236362996898e-06,
+      "loss": 0.7686,
+      "step": 20016
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610384242343407e-06,
+      "loss": 0.7625,
+      "step": 20017
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.609532124523245e-06,
+      "loss": 0.6044,
+      "step": 20018
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.608680009542604e-06,
+      "loss": 0.758,
+      "step": 20019
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60782789740769e-06,
+      "loss": 0.9718,
+      "step": 20020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.606975788124695e-06,
+      "loss": 0.7191,
+      "step": 20021
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.606123681699813e-06,
+      "loss": 0.845,
+      "step": 20022
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.605271578139246e-06,
+      "loss": 1.1163,
+      "step": 20023
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.604419477449187e-06,
+      "loss": 0.9054,
+      "step": 20024
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.603567379635836e-06,
+      "loss": 1.058,
+      "step": 20025
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.602715284705385e-06,
+      "loss": 0.9158,
+      "step": 20026
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.601863192664033e-06,
+      "loss": 0.7805,
+      "step": 20027
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.601011103517974e-06,
+      "loss": 0.7663,
+      "step": 20028
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.600159017273407e-06,
+      "loss": 0.8146,
+      "step": 20029
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59930693393653e-06,
+      "loss": 0.7756,
+      "step": 20030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.598454853513534e-06,
+      "loss": 0.6328,
+      "step": 20031
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.597602776010622e-06,
+      "loss": 0.8877,
+      "step": 20032
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.596750701433987e-06,
+      "loss": 1.0457,
+      "step": 20033
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.595898629789827e-06,
+      "loss": 0.7794,
+      "step": 20034
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.595046561084336e-06,
+      "loss": 1.0115,
+      "step": 20035
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.594194495323711e-06,
+      "loss": 1.0554,
+      "step": 20036
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.593342432514147e-06,
+      "loss": 1.1931,
+      "step": 20037
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.592490372661843e-06,
+      "loss": 0.8683,
+      "step": 20038
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591638315772995e-06,
+      "loss": 0.8847,
+      "step": 20039
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.590786261853798e-06,
+      "loss": 0.7273,
+      "step": 20040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.589934210910447e-06,
+      "loss": 0.7941,
+      "step": 20041
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.589082162949144e-06,
+      "loss": 0.5353,
+      "step": 20042
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.588230117976078e-06,
+      "loss": 0.7897,
+      "step": 20043
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.587378075997448e-06,
+      "loss": 0.8019,
+      "step": 20044
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.586526037019456e-06,
+      "loss": 0.6546,
+      "step": 20045
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.585674001048287e-06,
+      "loss": 0.705,
+      "step": 20046
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.584821968090145e-06,
+      "loss": 0.9601,
+      "step": 20047
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.583969938151223e-06,
+      "loss": 0.985,
+      "step": 20048
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.583117911237716e-06,
+      "loss": 0.8505,
+      "step": 20049
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.582265887355823e-06,
+      "loss": 1.0972,
+      "step": 20050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.581413866511739e-06,
+      "loss": 0.8595,
+      "step": 20051
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580561848711661e-06,
+      "loss": 0.7569,
+      "step": 20052
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.579709833961781e-06,
+      "loss": 0.697,
+      "step": 20053
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578857822268299e-06,
+      "loss": 0.9536,
+      "step": 20054
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578005813637414e-06,
+      "loss": 0.688,
+      "step": 20055
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.577153808075315e-06,
+      "loss": 1.0565,
+      "step": 20056
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.576301805588199e-06,
+      "loss": 0.6331,
+      "step": 20057
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.575449806182261e-06,
+      "loss": 1.0141,
+      "step": 20058
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.574597809863703e-06,
+      "loss": 1.0753,
+      "step": 20059
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.573745816638716e-06,
+      "loss": 0.9461,
+      "step": 20060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.572893826513498e-06,
+      "loss": 1.113,
+      "step": 20061
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57204183949424e-06,
+      "loss": 0.7907,
+      "step": 20062
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.571189855587142e-06,
+      "loss": 0.8921,
+      "step": 20063
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.570337874798403e-06,
+      "loss": 0.9507,
+      "step": 20064
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.569485897134214e-06,
+      "loss": 0.8357,
+      "step": 20065
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56863392260077e-06,
+      "loss": 0.7897,
+      "step": 20066
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.567781951204264e-06,
+      "loss": 0.9399,
+      "step": 20067
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5669299829509e-06,
+      "loss": 1.0612,
+      "step": 20068
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.566078017846869e-06,
+      "loss": 1.1129,
+      "step": 20069
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.565226055898366e-06,
+      "loss": 1.4097,
+      "step": 20070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.564374097111585e-06,
+      "loss": 0.9356,
+      "step": 20071
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.563522141492728e-06,
+      "loss": 1.1426,
+      "step": 20072
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.562670189047986e-06,
+      "loss": 1.3116,
+      "step": 20073
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.561818239783551e-06,
+      "loss": 0.768,
+      "step": 20074
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.560966293705631e-06,
+      "loss": 1.0817,
+      "step": 20075
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.560114350820405e-06,
+      "loss": 0.945,
+      "step": 20076
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.559262411134078e-06,
+      "loss": 0.7491,
+      "step": 20077
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.558410474652844e-06,
+      "loss": 1.0393,
+      "step": 20078
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5575585413829e-06,
+      "loss": 0.7491,
+      "step": 20079
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556706611330435e-06,
+      "loss": 0.9502,
+      "step": 20080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.555854684501652e-06,
+      "loss": 1.0874,
+      "step": 20081
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.555002760902742e-06,
+      "loss": 0.8452,
+      "step": 20082
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.554150840539902e-06,
+      "loss": 0.7484,
+      "step": 20083
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.553298923419327e-06,
+      "loss": 0.9468,
+      "step": 20084
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.552447009547214e-06,
+      "loss": 0.8586,
+      "step": 20085
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.551595098929752e-06,
+      "loss": 1.004,
+      "step": 20086
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.550743191573143e-06,
+      "loss": 0.7674,
+      "step": 20087
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.549891287483579e-06,
+      "loss": 0.9681,
+      "step": 20088
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.549039386667252e-06,
+      "loss": 0.7937,
+      "step": 20089
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.548187489130362e-06,
+      "loss": 0.7453,
+      "step": 20090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.547335594879104e-06,
+      "loss": 0.8421,
+      "step": 20091
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.546483703919671e-06,
+      "loss": 0.8444,
+      "step": 20092
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54563181625826e-06,
+      "loss": 0.8352,
+      "step": 20093
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.544779931901064e-06,
+      "loss": 1.0594,
+      "step": 20094
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543928050854283e-06,
+      "loss": 0.8432,
+      "step": 20095
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543076173124105e-06,
+      "loss": 1.0441,
+      "step": 20096
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.542224298716727e-06,
+      "loss": 1.0993,
+      "step": 20097
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.541372427638343e-06,
+      "loss": 0.742,
+      "step": 20098
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54052055989515e-06,
+      "loss": 0.873,
+      "step": 20099
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.539668695493344e-06,
+      "loss": 0.6021,
+      "step": 20100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.538816834439117e-06,
+      "loss": 1.0206,
+      "step": 20101
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537964976738668e-06,
+      "loss": 1.2285,
+      "step": 20102
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537113122398186e-06,
+      "loss": 1.0893,
+      "step": 20103
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.536261271423872e-06,
+      "loss": 0.9946,
+      "step": 20104
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535409423821919e-06,
+      "loss": 0.8387,
+      "step": 20105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.534557579598517e-06,
+      "loss": 0.6589,
+      "step": 20106
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.533705738759862e-06,
+      "loss": 0.8209,
+      "step": 20107
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532853901312153e-06,
+      "loss": 0.9551,
+      "step": 20108
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532002067261582e-06,
+      "loss": 0.6853,
+      "step": 20109
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.531150236614342e-06,
+      "loss": 0.8283,
+      "step": 20110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.530298409376631e-06,
+      "loss": 0.9056,
+      "step": 20111
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.529446585554644e-06,
+      "loss": 0.6765,
+      "step": 20112
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.528594765154573e-06,
+      "loss": 1.0302,
+      "step": 20113
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52774294818261e-06,
+      "loss": 0.9981,
+      "step": 20114
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52689113464496e-06,
+      "loss": 1.0242,
+      "step": 20115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.526039324547804e-06,
+      "loss": 0.7875,
+      "step": 20116
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.525187517897344e-06,
+      "loss": 0.9359,
+      "step": 20117
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.524335714699772e-06,
+      "loss": 0.6081,
+      "step": 20118
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.523483914961284e-06,
+      "loss": 0.9612,
+      "step": 20119
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.522632118688074e-06,
+      "loss": 0.9147,
+      "step": 20120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52178032588634e-06,
+      "loss": 0.8332,
+      "step": 20121
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.520928536562267e-06,
+      "loss": 0.562,
+      "step": 20122
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.520076750722055e-06,
+      "loss": 0.8798,
+      "step": 20123
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519224968371902e-06,
+      "loss": 0.647,
+      "step": 20124
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.518373189517999e-06,
+      "loss": 1.0655,
+      "step": 20125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.517521414166538e-06,
+      "loss": 0.8007,
+      "step": 20126
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.516669642323715e-06,
+      "loss": 0.7497,
+      "step": 20127
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.515817873995721e-06,
+      "loss": 0.8689,
+      "step": 20128
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.514966109188756e-06,
+      "loss": 0.8223,
+      "step": 20129
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.514114347909011e-06,
+      "loss": 0.9135,
+      "step": 20130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.513262590162681e-06,
+      "loss": 0.8919,
+      "step": 20131
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.512410835955957e-06,
+      "loss": 0.9659,
+      "step": 20132
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.511559085295037e-06,
+      "loss": 0.963,
+      "step": 20133
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.510707338186115e-06,
+      "loss": 0.5973,
+      "step": 20134
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509855594635385e-06,
+      "loss": 0.979,
+      "step": 20135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509003854649037e-06,
+      "loss": 0.901,
+      "step": 20136
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.508152118233265e-06,
+      "loss": 0.8884,
+      "step": 20137
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50730038539427e-06,
+      "loss": 0.8258,
+      "step": 20138
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.506448656138239e-06,
+      "loss": 0.8033,
+      "step": 20139
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.505596930471368e-06,
+      "loss": 1.1797,
+      "step": 20140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50474520839985e-06,
+      "loss": 1.2131,
+      "step": 20141
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.503893489929882e-06,
+      "loss": 0.7029,
+      "step": 20142
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.503041775067655e-06,
+      "loss": 0.8281,
+      "step": 20143
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502190063819362e-06,
+      "loss": 1.0722,
+      "step": 20144
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.501338356191204e-06,
+      "loss": 0.8711,
+      "step": 20145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.500486652189362e-06,
+      "loss": 1.0836,
+      "step": 20146
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49963495182004e-06,
+      "loss": 0.744,
+      "step": 20147
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.498783255089427e-06,
+      "loss": 0.9163,
+      "step": 20148
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.497931562003717e-06,
+      "loss": 0.8274,
+      "step": 20149
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.497079872569104e-06,
+      "loss": 1.1406,
+      "step": 20150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.496228186791784e-06,
+      "loss": 0.6783,
+      "step": 20151
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.495376504677947e-06,
+      "loss": 1.1528,
+      "step": 20152
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.494524826233788e-06,
+      "loss": 0.9212,
+      "step": 20153
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.493673151465501e-06,
+      "loss": 1.021,
+      "step": 20154
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.492821480379284e-06,
+      "loss": 0.6774,
+      "step": 20155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491969812981322e-06,
+      "loss": 1.4159,
+      "step": 20156
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49111814927781e-06,
+      "loss": 0.5282,
+      "step": 20157
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.490266489274945e-06,
+      "loss": 0.8303,
+      "step": 20158
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.489414832978917e-06,
+      "loss": 1.4515,
+      "step": 20159
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.488563180395922e-06,
+      "loss": 0.7334,
+      "step": 20160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.487711531532152e-06,
+      "loss": 1.0145,
+      "step": 20161
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4868598863938e-06,
+      "loss": 0.8091,
+      "step": 20162
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.48600824498706e-06,
+      "loss": 0.9923,
+      "step": 20163
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485156607318127e-06,
+      "loss": 0.7137,
+      "step": 20164
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.484304973393193e-06,
+      "loss": 0.8547,
+      "step": 20165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.483453343218451e-06,
+      "loss": 0.6808,
+      "step": 20166
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.482601716800092e-06,
+      "loss": 0.946,
+      "step": 20167
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.481750094144308e-06,
+      "loss": 1.3454,
+      "step": 20168
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.480898475257297e-06,
+      "loss": 0.849,
+      "step": 20169
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.480046860145251e-06,
+      "loss": 0.8715,
+      "step": 20170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47919524881436e-06,
+      "loss": 0.7247,
+      "step": 20171
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.478343641270819e-06,
+      "loss": 1.0093,
+      "step": 20172
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.477492037520824e-06,
+      "loss": 1.1143,
+      "step": 20173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.476640437570562e-06,
+      "loss": 1.0016,
+      "step": 20174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.475788841426232e-06,
+      "loss": 0.7968,
+      "step": 20175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474937249094022e-06,
+      "loss": 1.1003,
+      "step": 20176
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474085660580125e-06,
+      "loss": 0.8149,
+      "step": 20177
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.473234075890738e-06,
+      "loss": 0.9759,
+      "step": 20178
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47238249503205e-06,
+      "loss": 0.8312,
+      "step": 20179
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.471530918010254e-06,
+      "loss": 1.0759,
+      "step": 20180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.470679344831544e-06,
+      "loss": 0.7762,
+      "step": 20181
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.469827775502114e-06,
+      "loss": 1.2427,
+      "step": 20182
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468976210028156e-06,
+      "loss": 1.0163,
+      "step": 20183
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468124648415858e-06,
+      "loss": 0.8016,
+      "step": 20184
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.467273090671424e-06,
+      "loss": 1.028,
+      "step": 20185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.466421536801034e-06,
+      "loss": 0.9922,
+      "step": 20186
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465569986810886e-06,
+      "loss": 0.9925,
+      "step": 20187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.464718440707174e-06,
+      "loss": 0.9699,
+      "step": 20188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463866898496085e-06,
+      "loss": 0.7311,
+      "step": 20189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463015360183819e-06,
+      "loss": 0.8062,
+      "step": 20190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.462163825776565e-06,
+      "loss": 0.8317,
+      "step": 20191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.461312295280515e-06,
+      "loss": 0.8034,
+      "step": 20192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.460460768701859e-06,
+      "loss": 1.0442,
+      "step": 20193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.459609246046794e-06,
+      "loss": 0.9416,
+      "step": 20194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458757727321516e-06,
+      "loss": 0.618,
+      "step": 20195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457906212532206e-06,
+      "loss": 0.9165,
+      "step": 20196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457054701685064e-06,
+      "loss": 1.0267,
+      "step": 20197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.456203194786278e-06,
+      "loss": 0.907,
+      "step": 20198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455351691842046e-06,
+      "loss": 0.883,
+      "step": 20199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454500192858554e-06,
+      "loss": 0.9896,
+      "step": 20200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453648697842e-06,
+      "loss": 1.019,
+      "step": 20201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45279720679857e-06,
+      "loss": 0.7607,
+      "step": 20202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451945719734463e-06,
+      "loss": 0.9032,
+      "step": 20203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451094236655866e-06,
+      "loss": 0.8385,
+      "step": 20204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450242757568975e-06,
+      "loss": 1.0502,
+      "step": 20205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449391282479977e-06,
+      "loss": 0.8734,
+      "step": 20206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448539811395065e-06,
+      "loss": 0.7545,
+      "step": 20207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447688344320436e-06,
+      "loss": 0.8334,
+      "step": 20208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446836881262278e-06,
+      "loss": 0.7885,
+      "step": 20209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.445985422226784e-06,
+      "loss": 1.1662,
+      "step": 20210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.445133967220142e-06,
+      "loss": 0.7451,
+      "step": 20211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.444282516248553e-06,
+      "loss": 0.8734,
+      "step": 20212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4434310693182e-06,
+      "loss": 0.9934,
+      "step": 20213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442579626435277e-06,
+      "loss": 1.0693,
+      "step": 20214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.441728187605981e-06,
+      "loss": 0.9664,
+      "step": 20215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440876752836496e-06,
+      "loss": 0.8313,
+      "step": 20216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44002532213302e-06,
+      "loss": 1.1612,
+      "step": 20217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43917389550174e-06,
+      "loss": 0.8164,
+      "step": 20218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438322472948852e-06,
+      "loss": 0.9795,
+      "step": 20219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43747105448054e-06,
+      "loss": 0.9035,
+      "step": 20220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436619640103006e-06,
+      "loss": 1.2163,
+      "step": 20221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435768229822434e-06,
+      "loss": 0.8242,
+      "step": 20222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434916823645018e-06,
+      "loss": 0.8927,
+      "step": 20223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.43406542157695e-06,
+      "loss": 0.9046,
+      "step": 20224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.433214023624425e-06,
+      "loss": 1.094,
+      "step": 20225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.432362629793627e-06,
+      "loss": 0.6588,
+      "step": 20226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.431511240090753e-06,
+      "loss": 0.6742,
+      "step": 20227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.430659854521991e-06,
+      "loss": 0.9296,
+      "step": 20228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.429808473093532e-06,
+      "loss": 1.0422,
+      "step": 20229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428957095811569e-06,
+      "loss": 1.1193,
+      "step": 20230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428105722682296e-06,
+      "loss": 0.9652,
+      "step": 20231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.427254353711898e-06,
+      "loss": 0.855,
+      "step": 20232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426402988906574e-06,
+      "loss": 1.0007,
+      "step": 20233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42555162827251e-06,
+      "loss": 0.9535,
+      "step": 20234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.424700271815901e-06,
+      "loss": 0.9378,
+      "step": 20235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.423848919542932e-06,
+      "loss": 0.6506,
+      "step": 20236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4229975714598e-06,
+      "loss": 0.8169,
+      "step": 20237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42214622757269e-06,
+      "loss": 0.5877,
+      "step": 20238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.421294887887799e-06,
+      "loss": 0.8503,
+      "step": 20239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.420443552411315e-06,
+      "loss": 0.9098,
+      "step": 20240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41959222114943e-06,
+      "loss": 0.844,
+      "step": 20241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.418740894108334e-06,
+      "loss": 1.2441,
+      "step": 20242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.417889571294222e-06,
+      "loss": 0.8835,
+      "step": 20243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41703825271328e-06,
+      "loss": 1.1893,
+      "step": 20244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.416186938371703e-06,
+      "loss": 0.8493,
+      "step": 20245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.415335628275677e-06,
+      "loss": 0.9234,
+      "step": 20246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.414484322431393e-06,
+      "loss": 1.1375,
+      "step": 20247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.413633020845047e-06,
+      "loss": 0.6948,
+      "step": 20248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.412781723522827e-06,
+      "loss": 0.716,
+      "step": 20249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41193043047092e-06,
+      "loss": 0.7092,
+      "step": 20250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.411079141695522e-06,
+      "loss": 0.7514,
+      "step": 20251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.410227857202822e-06,
+      "loss": 0.8508,
+      "step": 20252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.409376576999012e-06,
+      "loss": 0.8176,
+      "step": 20253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.408525301090278e-06,
+      "loss": 0.8834,
+      "step": 20254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.40767402948282e-06,
+      "loss": 0.9393,
+      "step": 20255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.406822762182817e-06,
+      "loss": 0.743,
+      "step": 20256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.405971499196465e-06,
+      "loss": 0.7802,
+      "step": 20257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.405120240529956e-06,
+      "loss": 1.0466,
+      "step": 20258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.404268986189476e-06,
+      "loss": 0.7257,
+      "step": 20259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.403417736181219e-06,
+      "loss": 1.0309,
+      "step": 20260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402566490511376e-06,
+      "loss": 0.6307,
+      "step": 20261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.401715249186135e-06,
+      "loss": 1.0464,
+      "step": 20262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400864012211686e-06,
+      "loss": 0.9255,
+      "step": 20263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400012779594221e-06,
+      "loss": 1.1049,
+      "step": 20264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.399161551339933e-06,
+      "loss": 0.7542,
+      "step": 20265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.398310327455008e-06,
+      "loss": 0.868,
+      "step": 20266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.397459107945634e-06,
+      "loss": 0.7589,
+      "step": 20267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.396607892818005e-06,
+      "loss": 0.9972,
+      "step": 20268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39575668207831e-06,
+      "loss": 1.0532,
+      "step": 20269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.394905475732742e-06,
+      "loss": 1.2426,
+      "step": 20270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.394054273787487e-06,
+      "loss": 0.8722,
+      "step": 20271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.393203076248735e-06,
+      "loss": 0.8363,
+      "step": 20272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.39235188312268e-06,
+      "loss": 1.5306,
+      "step": 20273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.391500694415508e-06,
+      "loss": 1.3977,
+      "step": 20274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.390649510133414e-06,
+      "loss": 0.7581,
+      "step": 20275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.389798330282584e-06,
+      "loss": 0.8426,
+      "step": 20276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.388947154869203e-06,
+      "loss": 0.9534,
+      "step": 20277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38809598389947e-06,
+      "loss": 0.4451,
+      "step": 20278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38724481737957e-06,
+      "loss": 0.9706,
+      "step": 20279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.386393655315696e-06,
+      "loss": 0.8875,
+      "step": 20280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38554249771403e-06,
+      "loss": 0.9683,
+      "step": 20281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.384691344580772e-06,
+      "loss": 0.8273,
+      "step": 20282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.383840195922105e-06,
+      "loss": 0.9318,
+      "step": 20283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38298905174422e-06,
+      "loss": 0.8071,
+      "step": 20284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.382137912053313e-06,
+      "loss": 0.8348,
+      "step": 20285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38128677685556e-06,
+      "loss": 0.9241,
+      "step": 20286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.380435646157163e-06,
+      "loss": 0.9404,
+      "step": 20287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.379584519964304e-06,
+      "loss": 1.0502,
+      "step": 20288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.378733398283177e-06,
+      "loss": 0.8328,
+      "step": 20289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.377882281119967e-06,
+      "loss": 0.9152,
+      "step": 20290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.377031168480867e-06,
+      "loss": 0.8195,
+      "step": 20291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.376180060372067e-06,
+      "loss": 0.7419,
+      "step": 20292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.375328956799754e-06,
+      "loss": 1.0772,
+      "step": 20293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.374477857770118e-06,
+      "loss": 1.0967,
+      "step": 20294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.373626763289352e-06,
+      "loss": 0.9026,
+      "step": 20295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37277567336364e-06,
+      "loss": 1.177,
+      "step": 20296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.371924587999172e-06,
+      "loss": 1.2501,
+      "step": 20297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.371073507202138e-06,
+      "loss": 1.1252,
+      "step": 20298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.370222430978724e-06,
+      "loss": 0.8747,
+      "step": 20299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.369371359335128e-06,
+      "loss": 1.0826,
+      "step": 20300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.368520292277531e-06,
+      "loss": 0.6896,
+      "step": 20301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.367669229812123e-06,
+      "loss": 0.677,
+      "step": 20302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.366818171945097e-06,
+      "loss": 1.0704,
+      "step": 20303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36596711868264e-06,
+      "loss": 0.7925,
+      "step": 20304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36511607003094e-06,
+      "loss": 1.026,
+      "step": 20305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.364265025996187e-06,
+      "loss": 0.8791,
+      "step": 20306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36341398658457e-06,
+      "loss": 1.0126,
+      "step": 20307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.362562951802271e-06,
+      "loss": 1.1142,
+      "step": 20308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36171192165549e-06,
+      "loss": 0.5379,
+      "step": 20309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36086089615041e-06,
+      "loss": 1.0618,
+      "step": 20310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36000987529322e-06,
+      "loss": 1.0948,
+      "step": 20311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.359158859090109e-06,
+      "loss": 0.7042,
+      "step": 20312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.358307847547268e-06,
+      "loss": 0.9613,
+      "step": 20313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.357456840670884e-06,
+      "loss": 0.8646,
+      "step": 20314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.356605838467145e-06,
+      "loss": 0.9533,
+      "step": 20315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.35575484094224e-06,
+      "loss": 1.1166,
+      "step": 20316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354903848102352e-06,
+      "loss": 0.886,
+      "step": 20317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354052859953679e-06,
+      "loss": 0.8815,
+      "step": 20318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.353201876502406e-06,
+      "loss": 0.7872,
+      "step": 20319
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.352350897754719e-06,
+      "loss": 0.8122,
+      "step": 20320
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.351499923716808e-06,
+      "loss": 0.7066,
+      "step": 20321
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.350648954394862e-06,
+      "loss": 0.868,
+      "step": 20322
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34979798979507e-06,
+      "loss": 1.0371,
+      "step": 20323
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.348947029923617e-06,
+      "loss": 0.9519,
+      "step": 20324
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.3480960747867e-06,
+      "loss": 0.7999,
+      "step": 20325
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.347245124390496e-06,
+      "loss": 0.8184,
+      "step": 20326
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346394178741197e-06,
+      "loss": 0.6125,
+      "step": 20327
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.345543237844993e-06,
+      "loss": 0.8622,
+      "step": 20328
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34469230170807e-06,
+      "loss": 0.9709,
+      "step": 20329
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.343841370336619e-06,
+      "loss": 0.8074,
+      "step": 20330
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.342990443736825e-06,
+      "loss": 0.7215,
+      "step": 20331
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34213952191488e-06,
+      "loss": 0.4784,
+      "step": 20332
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.341288604876965e-06,
+      "loss": 0.9077,
+      "step": 20333
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.340437692629275e-06,
+      "loss": 1.0407,
+      "step": 20334
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.339586785177998e-06,
+      "loss": 0.8181,
+      "step": 20335
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.338735882529318e-06,
+      "loss": 0.6281,
+      "step": 20336
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337884984689424e-06,
+      "loss": 0.9752,
+      "step": 20337
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337034091664503e-06,
+      "loss": 0.6896,
+      "step": 20338
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.336183203460742e-06,
+      "loss": 0.9113,
+      "step": 20339
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.335332320084331e-06,
+      "loss": 0.9098,
+      "step": 20340
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33448144154146e-06,
+      "loss": 1.1509,
+      "step": 20341
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33363056783831e-06,
+      "loss": 0.7283,
+      "step": 20342
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.332779698981076e-06,
+      "loss": 1.0249,
+      "step": 20343
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33192883497594e-06,
+      "loss": 1.1908,
+      "step": 20344
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.331077975829094e-06,
+      "loss": 0.8081,
+      "step": 20345
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.330227121546722e-06,
+      "loss": 0.9102,
+      "step": 20346
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.329376272135013e-06,
+      "loss": 0.7255,
+      "step": 20347
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.328525427600152e-06,
+      "loss": 0.945,
+      "step": 20348
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.327674587948331e-06,
+      "loss": 0.9151,
+      "step": 20349
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.326823753185733e-06,
+      "loss": 0.8674,
+      "step": 20350
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325972923318547e-06,
+      "loss": 0.8879,
+      "step": 20351
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325122098352965e-06,
+      "loss": 0.4134,
+      "step": 20352
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.324271278295167e-06,
+      "loss": 1.2148,
+      "step": 20353
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.323420463151345e-06,
+      "loss": 1.1482,
+      "step": 20354
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.322569652927685e-06,
+      "loss": 0.8107,
+      "step": 20355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.321718847630372e-06,
+      "loss": 0.7185,
+      "step": 20356
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320868047265594e-06,
+      "loss": 0.673,
+      "step": 20357
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32001725183954e-06,
+      "loss": 1.0003,
+      "step": 20358
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.319166461358396e-06,
+      "loss": 0.7392,
+      "step": 20359
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.318315675828347e-06,
+      "loss": 0.7811,
+      "step": 20360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.317464895255585e-06,
+      "loss": 1.0108,
+      "step": 20361
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.316614119646294e-06,
+      "loss": 0.8373,
+      "step": 20362
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31576334900666e-06,
+      "loss": 1.0265,
+      "step": 20363
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31491258334287e-06,
+      "loss": 0.8228,
+      "step": 20364
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.314061822661117e-06,
+      "loss": 0.7312,
+      "step": 20365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.313211066967576e-06,
+      "loss": 0.9573,
+      "step": 20366
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.312360316268445e-06,
+      "loss": 0.7198,
+      "step": 20367
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.311509570569903e-06,
+      "loss": 1.0091,
+      "step": 20368
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31065882987814e-06,
+      "loss": 0.6909,
+      "step": 20369
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309808094199343e-06,
+      "loss": 0.9867,
+      "step": 20370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.3089573635397e-06,
+      "loss": 0.8654,
+      "step": 20371
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.308106637905394e-06,
+      "loss": 0.7224,
+      "step": 20372
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.307255917302613e-06,
+      "loss": 0.8847,
+      "step": 20373
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.306405201737544e-06,
+      "loss": 0.8824,
+      "step": 20374
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.305554491216377e-06,
+      "loss": 0.8553,
+      "step": 20375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.304703785745291e-06,
+      "loss": 0.9302,
+      "step": 20376
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30385308533048e-06,
+      "loss": 1.04,
+      "step": 20377
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30300238997812e-06,
+      "loss": 0.779,
+      "step": 20378
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.302151699694408e-06,
+      "loss": 0.8342,
+      "step": 20379
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.301301014485527e-06,
+      "loss": 0.8613,
+      "step": 20380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.300450334357662e-06,
+      "loss": 0.9799,
+      "step": 20381
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.299599659316997e-06,
+      "loss": 0.9145,
+      "step": 20382
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.298748989369727e-06,
+      "loss": 1.1119,
+      "step": 20383
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.297898324522028e-06,
+      "loss": 0.8522,
+      "step": 20384
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.297047664780093e-06,
+      "loss": 0.8738,
+      "step": 20385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.296197010150105e-06,
+      "loss": 0.8179,
+      "step": 20386
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.295346360638248e-06,
+      "loss": 0.7851,
+      "step": 20387
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294495716250711e-06,
+      "loss": 0.795,
+      "step": 20388
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.293645076993682e-06,
+      "loss": 0.9459,
+      "step": 20389
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.292794442873342e-06,
+      "loss": 0.9208,
+      "step": 20390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291943813895879e-06,
+      "loss": 0.7345,
+      "step": 20391
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291093190067479e-06,
+      "loss": 1.0797,
+      "step": 20392
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.29024257139433e-06,
+      "loss": 0.8966,
+      "step": 20393
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.289391957882614e-06,
+      "loss": 1.1186,
+      "step": 20394
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.288541349538525e-06,
+      "loss": 0.8154,
+      "step": 20395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.287690746368235e-06,
+      "loss": 0.9455,
+      "step": 20396
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28684014837794e-06,
+      "loss": 1.0109,
+      "step": 20397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285989555573822e-06,
+      "loss": 1.2378,
+      "step": 20398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285138967962067e-06,
+      "loss": 0.7447,
+      "step": 20399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.284288385548858e-06,
+      "loss": 1.0789,
+      "step": 20400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.283437808340387e-06,
+      "loss": 0.8154,
+      "step": 20401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.282587236342835e-06,
+      "loss": 0.8997,
+      "step": 20402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.281736669562387e-06,
+      "loss": 0.8481,
+      "step": 20403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28088610800523e-06,
+      "loss": 0.9306,
+      "step": 20404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280035551677553e-06,
+      "loss": 0.9258,
+      "step": 20405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.279185000585535e-06,
+      "loss": 1.0235,
+      "step": 20406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.278334454735362e-06,
+      "loss": 1.0171,
+      "step": 20407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.277483914133224e-06,
+      "loss": 0.8862,
+      "step": 20408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2766333787853e-06,
+      "loss": 0.7216,
+      "step": 20409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.27578284869778e-06,
+      "loss": 0.8989,
+      "step": 20410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.274932323876846e-06,
+      "loss": 0.8593,
+      "step": 20411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.274081804328685e-06,
+      "loss": 0.7906,
+      "step": 20412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.273231290059483e-06,
+      "loss": 1.1563,
+      "step": 20413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272380781075425e-06,
+      "loss": 0.9369,
+      "step": 20414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.271530277382695e-06,
+      "loss": 0.9231,
+      "step": 20415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.270679778987475e-06,
+      "loss": 1.2649,
+      "step": 20416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.269829285895954e-06,
+      "loss": 0.883,
+      "step": 20417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268978798114315e-06,
+      "loss": 0.8676,
+      "step": 20418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268128315648742e-06,
+      "loss": 0.7209,
+      "step": 20419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.267277838505423e-06,
+      "loss": 1.0968,
+      "step": 20420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26642736669054e-06,
+      "loss": 0.7893,
+      "step": 20421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.265576900210278e-06,
+      "loss": 0.7555,
+      "step": 20422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.264726439070823e-06,
+      "loss": 1.1924,
+      "step": 20423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26387598327836e-06,
+      "loss": 0.7555,
+      "step": 20424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.263025532839077e-06,
+      "loss": 1.0492,
+      "step": 20425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.262175087759148e-06,
+      "loss": 0.9022,
+      "step": 20426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.261324648044763e-06,
+      "loss": 0.5797,
+      "step": 20427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.260474213702109e-06,
+      "loss": 0.8557,
+      "step": 20428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.259623784737369e-06,
+      "loss": 1.1054,
+      "step": 20429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.258773361156725e-06,
+      "loss": 0.8265,
+      "step": 20430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.257922942966366e-06,
+      "loss": 0.8941,
+      "step": 20431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.257072530172474e-06,
+      "loss": 0.7539,
+      "step": 20432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.256222122781232e-06,
+      "loss": 0.9027,
+      "step": 20433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.255371720798824e-06,
+      "loss": 0.9174,
+      "step": 20434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25452132423144e-06,
+      "loss": 0.6337,
+      "step": 20435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.253670933085256e-06,
+      "loss": 0.6815,
+      "step": 20436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.252820547366462e-06,
+      "loss": 1.0788,
+      "step": 20437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.251970167081238e-06,
+      "loss": 0.7226,
+      "step": 20438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.251119792235767e-06,
+      "loss": 0.6349,
+      "step": 20439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.25026942283624e-06,
+      "loss": 0.8416,
+      "step": 20440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.249419058888838e-06,
+      "loss": 1.1627,
+      "step": 20441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.248568700399742e-06,
+      "loss": 1.0585,
+      "step": 20442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.247718347375136e-06,
+      "loss": 0.8508,
+      "step": 20443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24686799982121e-06,
+      "loss": 1.0657,
+      "step": 20444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.246017657744142e-06,
+      "loss": 0.9101,
+      "step": 20445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.245167321150118e-06,
+      "loss": 1.0133,
+      "step": 20446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.244316990045322e-06,
+      "loss": 0.7344,
+      "step": 20447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.243466664435931e-06,
+      "loss": 0.6844,
+      "step": 20448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24261634432814e-06,
+      "loss": 0.9197,
+      "step": 20449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.241766029728124e-06,
+      "loss": 0.6989,
+      "step": 20450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.240915720642071e-06,
+      "loss": 0.797,
+      "step": 20451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.240065417076161e-06,
+      "loss": 0.6059,
+      "step": 20452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.239215119036583e-06,
+      "loss": 0.6779,
+      "step": 20453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.238364826529516e-06,
+      "loss": 1.0302,
+      "step": 20454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.237514539561147e-06,
+      "loss": 0.7248,
+      "step": 20455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236664258137654e-06,
+      "loss": 1.2526,
+      "step": 20456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.235813982265223e-06,
+      "loss": 0.9227,
+      "step": 20457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.234963711950038e-06,
+      "loss": 0.794,
+      "step": 20458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.234113447198282e-06,
+      "loss": 0.93,
+      "step": 20459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.233263188016138e-06,
+      "loss": 0.7287,
+      "step": 20460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.232412934409787e-06,
+      "loss": 0.8483,
+      "step": 20461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.231562686385418e-06,
+      "loss": 0.5684,
+      "step": 20462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.23071244394921e-06,
+      "loss": 0.6948,
+      "step": 20463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.229862207107342e-06,
+      "loss": 0.7529,
+      "step": 20464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22901197586601e-06,
+      "loss": 0.9577,
+      "step": 20465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.228161750231382e-06,
+      "loss": 0.7245,
+      "step": 20466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22731153020965e-06,
+      "loss": 1.0885,
+      "step": 20467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.226461315806995e-06,
+      "loss": 0.6759,
+      "step": 20468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.225611107029598e-06,
+      "loss": 0.7089,
+      "step": 20469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.224760903883641e-06,
+      "loss": 1.0296,
+      "step": 20470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22391070637531e-06,
+      "loss": 1.0262,
+      "step": 20471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.223060514510788e-06,
+      "loss": 0.8134,
+      "step": 20472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.222210328296254e-06,
+      "loss": 0.8871,
+      "step": 20473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.221360147737894e-06,
+      "loss": 0.9418,
+      "step": 20474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.220509972841893e-06,
+      "loss": 1.1066,
+      "step": 20475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.219659803614427e-06,
+      "loss": 1.007,
+      "step": 20476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218809640061682e-06,
+      "loss": 1.0221,
+      "step": 20477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21795948218984e-06,
+      "loss": 0.8949,
+      "step": 20478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21710933000508e-06,
+      "loss": 0.6749,
+      "step": 20479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21625918351359e-06,
+      "loss": 1.1867,
+      "step": 20480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.215409042721553e-06,
+      "loss": 0.8957,
+      "step": 20481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214558907635144e-06,
+      "loss": 0.9426,
+      "step": 20482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.213708778260553e-06,
+      "loss": 0.996,
+      "step": 20483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21285865460396e-06,
+      "loss": 0.9726,
+      "step": 20484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.212008536671545e-06,
+      "loss": 1.0829,
+      "step": 20485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211158424469492e-06,
+      "loss": 0.8298,
+      "step": 20486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.210308318003982e-06,
+      "loss": 0.9034,
+      "step": 20487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.209458217281195e-06,
+      "loss": 0.9286,
+      "step": 20488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20860812230732e-06,
+      "loss": 0.9826,
+      "step": 20489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.207758033088533e-06,
+      "loss": 0.9513,
+      "step": 20490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.206907949631015e-06,
+      "loss": 0.8078,
+      "step": 20491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.206057871940953e-06,
+      "loss": 0.658,
+      "step": 20492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.205207800024525e-06,
+      "loss": 1.0011,
+      "step": 20493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.204357733887918e-06,
+      "loss": 0.9593,
+      "step": 20494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.203507673537306e-06,
+      "loss": 0.8065,
+      "step": 20495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.202657618978876e-06,
+      "loss": 0.824,
+      "step": 20496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.201807570218807e-06,
+      "loss": 0.8042,
+      "step": 20497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.200957527263283e-06,
+      "loss": 0.7533,
+      "step": 20498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.200107490118486e-06,
+      "loss": 0.9658,
+      "step": 20499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19925745879059e-06,
+      "loss": 1.0191,
+      "step": 20500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19840743328579e-06,
+      "loss": 1.0733,
+      "step": 20501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.197557413610258e-06,
+      "loss": 1.0518,
+      "step": 20502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.196707399770178e-06,
+      "loss": 0.9181,
+      "step": 20503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19585739177173e-06,
+      "loss": 0.8559,
+      "step": 20504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.195007389621098e-06,
+      "loss": 0.6528,
+      "step": 20505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.194157393324458e-06,
+      "loss": 0.534,
+      "step": 20506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.193307402887997e-06,
+      "loss": 0.9007,
+      "step": 20507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.192457418317895e-06,
+      "loss": 0.799,
+      "step": 20508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19160743962033e-06,
+      "loss": 1.1783,
+      "step": 20509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190757466801488e-06,
+      "loss": 1.0232,
+      "step": 20510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.189907499867546e-06,
+      "loss": 0.8123,
+      "step": 20511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.189057538824687e-06,
+      "loss": 0.5885,
+      "step": 20512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.188207583679091e-06,
+      "loss": 0.8409,
+      "step": 20513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.187357634436944e-06,
+      "loss": 0.8979,
+      "step": 20514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.186507691104417e-06,
+      "loss": 0.8308,
+      "step": 20515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.185657753687699e-06,
+      "loss": 0.6646,
+      "step": 20516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.184807822192969e-06,
+      "loss": 1.1288,
+      "step": 20517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.183957896626404e-06,
+      "loss": 0.9447,
+      "step": 20518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18310797699419e-06,
+      "loss": 0.8008,
+      "step": 20519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.182258063302504e-06,
+      "loss": 0.9493,
+      "step": 20520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.18140815555753e-06,
+      "loss": 0.7111,
+      "step": 20521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.180558253765443e-06,
+      "loss": 0.6944,
+      "step": 20522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17970835793243e-06,
+      "loss": 0.822,
+      "step": 20523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.178858468064674e-06,
+      "loss": 1.0338,
+      "step": 20524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.178008584168343e-06,
+      "loss": 1.0954,
+      "step": 20525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17715870624963e-06,
+      "loss": 0.8637,
+      "step": 20526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176308834314704e-06,
+      "loss": 1.008,
+      "step": 20527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.175458968369757e-06,
+      "loss": 0.7156,
+      "step": 20528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.174609108420962e-06,
+      "loss": 0.785,
+      "step": 20529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.173759254474501e-06,
+      "loss": 0.9332,
+      "step": 20530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.172909406536554e-06,
+      "loss": 1.0507,
+      "step": 20531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.172059564613302e-06,
+      "loss": 0.9196,
+      "step": 20532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171209728710928e-06,
+      "loss": 1.1485,
+      "step": 20533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.170359898835608e-06,
+      "loss": 0.9582,
+      "step": 20534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16951007499352e-06,
+      "loss": 0.893,
+      "step": 20535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.168660257190846e-06,
+      "loss": 1.0312,
+      "step": 20536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.167810445433771e-06,
+      "loss": 0.4868,
+      "step": 20537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166960639728468e-06,
+      "loss": 1.1769,
+      "step": 20538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166110840081122e-06,
+      "loss": 0.7305,
+      "step": 20539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.165261046497908e-06,
+      "loss": 0.9248,
+      "step": 20540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16441125898501e-06,
+      "loss": 1.0875,
+      "step": 20541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.163561477548605e-06,
+      "loss": 0.5345,
+      "step": 20542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.162711702194873e-06,
+      "loss": 0.9904,
+      "step": 20543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16186193293e-06,
+      "loss": 1.0949,
+      "step": 20544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.161012169760155e-06,
+      "loss": 0.8231,
+      "step": 20545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.160162412691525e-06,
+      "loss": 1.0176,
+      "step": 20546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.159312661730285e-06,
+      "loss": 0.9634,
+      "step": 20547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.158462916882618e-06,
+      "loss": 1.1609,
+      "step": 20548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1576131781547e-06,
+      "loss": 0.8581,
+      "step": 20549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.156763445552714e-06,
+      "loss": 0.8839,
+      "step": 20550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15591371908284e-06,
+      "loss": 0.744,
+      "step": 20551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15506399875125e-06,
+      "loss": 0.7363,
+      "step": 20552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.154214284564132e-06,
+      "loss": 0.7046,
+      "step": 20553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.153364576527664e-06,
+      "loss": 1.0025,
+      "step": 20554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.152514874648021e-06,
+      "loss": 0.9619,
+      "step": 20555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.151665178931384e-06,
+      "loss": 0.7862,
+      "step": 20556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15081548938393e-06,
+      "loss": 1.222,
+      "step": 20557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.14996580601184e-06,
+      "loss": 0.8612,
+      "step": 20558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.149116128821294e-06,
+      "loss": 1.1437,
+      "step": 20559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.148266457818473e-06,
+      "loss": 0.7711,
+      "step": 20560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.147416793009547e-06,
+      "loss": 0.6268,
+      "step": 20561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.146567134400704e-06,
+      "loss": 1.0652,
+      "step": 20562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.14571748199812e-06,
+      "loss": 0.9495,
+      "step": 20563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.144867835807976e-06,
+      "loss": 1.0279,
+      "step": 20564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.144018195836445e-06,
+      "loss": 0.9552,
+      "step": 20565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.143168562089708e-06,
+      "loss": 0.6751,
+      "step": 20566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.142318934573944e-06,
+      "loss": 0.9126,
+      "step": 20567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.141469313295333e-06,
+      "loss": 0.9504,
+      "step": 20568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.140619698260054e-06,
+      "loss": 1.2251,
+      "step": 20569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.13977008947428e-06,
+      "loss": 0.9381,
+      "step": 20570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138920486944196e-06,
+      "loss": 1.3527,
+      "step": 20571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138070890675979e-06,
+      "loss": 1.1227,
+      "step": 20572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137221300675805e-06,
+      "loss": 1.1002,
+      "step": 20573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.136371716949856e-06,
+      "loss": 1.0154,
+      "step": 20574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.135522139504305e-06,
+      "loss": 1.1702,
+      "step": 20575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.13467256834533e-06,
+      "loss": 1.0744,
+      "step": 20576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.133823003479117e-06,
+      "loss": 0.8828,
+      "step": 20577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132973444911838e-06,
+      "loss": 0.8363,
+      "step": 20578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132123892649671e-06,
+      "loss": 0.5824,
+      "step": 20579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.131274346698797e-06,
+      "loss": 0.9001,
+      "step": 20580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.130424807065392e-06,
+      "loss": 0.7713,
+      "step": 20581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.129575273755634e-06,
+      "loss": 0.7444,
+      "step": 20582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.1287257467757e-06,
+      "loss": 0.8178,
+      "step": 20583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127876226131775e-06,
+      "loss": 0.8951,
+      "step": 20584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127026711830025e-06,
+      "loss": 0.5121,
+      "step": 20585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.126177203876635e-06,
+      "loss": 1.0946,
+      "step": 20586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.125327702277782e-06,
+      "loss": 1.0703,
+      "step": 20587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12447820703964e-06,
+      "loss": 0.7395,
+      "step": 20588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123628718168394e-06,
+      "loss": 0.7048,
+      "step": 20589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.122779235670217e-06,
+      "loss": 0.7661,
+      "step": 20590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121929759551285e-06,
+      "loss": 1.0321,
+      "step": 20591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121080289817776e-06,
+      "loss": 0.8909,
+      "step": 20592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.120230826475871e-06,
+      "loss": 1.0376,
+      "step": 20593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.119381369531748e-06,
+      "loss": 0.8034,
+      "step": 20594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.118531918991578e-06,
+      "loss": 1.1367,
+      "step": 20595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.117682474861542e-06,
+      "loss": 1.0563,
+      "step": 20596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.116833037147815e-06,
+      "loss": 0.7371,
+      "step": 20597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11598360585658e-06,
+      "loss": 0.9077,
+      "step": 20598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.115134180994007e-06,
+      "loss": 0.8122,
+      "step": 20599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11428476256628e-06,
+      "loss": 1.2468,
+      "step": 20600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.113435350579568e-06,
+      "loss": 0.822,
+      "step": 20601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.112585945040055e-06,
+      "loss": 0.7967,
+      "step": 20602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.111736545953917e-06,
+      "loss": 0.6617,
+      "step": 20603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11088715332733e-06,
+      "loss": 1.0143,
+      "step": 20604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11003776716647e-06,
+      "loss": 1.0201,
+      "step": 20605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10918838747751e-06,
+      "loss": 0.8947,
+      "step": 20606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.108339014266636e-06,
+      "loss": 1.0784,
+      "step": 20607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.107489647540017e-06,
+      "loss": 1.1983,
+      "step": 20608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.106640287303833e-06,
+      "loss": 0.6596,
+      "step": 20609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.105790933564259e-06,
+      "loss": 0.8431,
+      "step": 20610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.104941586327472e-06,
+      "loss": 0.8888,
+      "step": 20611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.104092245599652e-06,
+      "loss": 0.9158,
+      "step": 20612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10324291138697e-06,
+      "loss": 1.431,
+      "step": 20613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10239358369561e-06,
+      "loss": 1.0969,
+      "step": 20614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10154426253174e-06,
+      "loss": 0.8266,
+      "step": 20615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.10069494790154e-06,
+      "loss": 0.7109,
+      "step": 20616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.099845639811185e-06,
+      "loss": 1.1484,
+      "step": 20617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.098996338266854e-06,
+      "loss": 1.0335,
+      "step": 20618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09814704327472e-06,
+      "loss": 0.9602,
+      "step": 20619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.097297754840962e-06,
+      "loss": 0.792,
+      "step": 20620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.096448472971756e-06,
+      "loss": 1.0142,
+      "step": 20621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.095599197673275e-06,
+      "loss": 0.8303,
+      "step": 20622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.094749928951698e-06,
+      "loss": 0.8982,
+      "step": 20623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.093900666813202e-06,
+      "loss": 0.814,
+      "step": 20624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09305141126396e-06,
+      "loss": 0.7906,
+      "step": 20625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.092202162310149e-06,
+      "loss": 1.1896,
+      "step": 20626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.091352919957945e-06,
+      "loss": 1.3475,
+      "step": 20627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09050368421352e-06,
+      "loss": 0.8628,
+      "step": 20628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.089654455083055e-06,
+      "loss": 0.8646,
+      "step": 20629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.088805232572725e-06,
+      "loss": 0.9894,
+      "step": 20630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.087956016688702e-06,
+      "loss": 1.348,
+      "step": 20631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.087106807437167e-06,
+      "loss": 1.0338,
+      "step": 20632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086257604824291e-06,
+      "loss": 0.963,
+      "step": 20633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.085408408856256e-06,
+      "loss": 1.0787,
+      "step": 20634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.084559219539228e-06,
+      "loss": 0.9364,
+      "step": 20635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.083710036879387e-06,
+      "loss": 0.7774,
+      "step": 20636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.082860860882908e-06,
+      "loss": 0.8214,
+      "step": 20637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.082011691555968e-06,
+      "loss": 1.0484,
+      "step": 20638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08116252890474e-06,
+      "loss": 0.7769,
+      "step": 20639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.080313372935399e-06,
+      "loss": 1.2322,
+      "step": 20640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.079464223654123e-06,
+      "loss": 0.8729,
+      "step": 20641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078615081067086e-06,
+      "loss": 0.9981,
+      "step": 20642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.077765945180463e-06,
+      "loss": 0.8829,
+      "step": 20643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.076916816000428e-06,
+      "loss": 1.1446,
+      "step": 20644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.076067693533156e-06,
+      "loss": 0.8591,
+      "step": 20645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07521857778482e-06,
+      "loss": 0.8194,
+      "step": 20646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0743694687616e-06,
+      "loss": 0.8479,
+      "step": 20647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.073520366469667e-06,
+      "loss": 1.1133,
+      "step": 20648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.072671270915195e-06,
+      "loss": 0.784,
+      "step": 20649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.071822182104362e-06,
+      "loss": 0.601,
+      "step": 20650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.070973100043344e-06,
+      "loss": 1.0568,
+      "step": 20651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07012402473831e-06,
+      "loss": 1.059,
+      "step": 20652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.069274956195435e-06,
+      "loss": 0.8694,
+      "step": 20653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.068425894420903e-06,
+      "loss": 0.9724,
+      "step": 20654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.067576839420876e-06,
+      "loss": 0.7333,
+      "step": 20655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.066727791201534e-06,
+      "loss": 1.0243,
+      "step": 20656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065878749769052e-06,
+      "loss": 1.0971,
+      "step": 20657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065029715129604e-06,
+      "loss": 1.104,
+      "step": 20658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.064180687289361e-06,
+      "loss": 1.0115,
+      "step": 20659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.063331666254503e-06,
+      "loss": 0.9978,
+      "step": 20660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0624826520312e-06,
+      "loss": 0.9072,
+      "step": 20661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.061633644625626e-06,
+      "loss": 0.6997,
+      "step": 20662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.060784644043959e-06,
+      "loss": 1.2051,
+      "step": 20663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.059935650292371e-06,
+      "loss": 0.9054,
+      "step": 20664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.059086663377034e-06,
+      "loss": 0.9764,
+      "step": 20665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.058237683304123e-06,
+      "loss": 1.1115,
+      "step": 20666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05738871007981e-06,
+      "loss": 1.0117,
+      "step": 20667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.056539743710273e-06,
+      "loss": 0.9883,
+      "step": 20668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.055690784201684e-06,
+      "loss": 0.8868,
+      "step": 20669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.054841831560216e-06,
+      "loss": 0.8462,
+      "step": 20670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.053992885792041e-06,
+      "loss": 0.8713,
+      "step": 20671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.053143946903338e-06,
+      "loss": 1.0828,
+      "step": 20672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.052295014900276e-06,
+      "loss": 1.106,
+      "step": 20673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.051446089789033e-06,
+      "loss": 0.9141,
+      "step": 20674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.050597171575776e-06,
+      "loss": 0.8499,
+      "step": 20675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.049748260266682e-06,
+      "loss": 0.6881,
+      "step": 20676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.048899355867924e-06,
+      "loss": 0.9978,
+      "step": 20677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.048050458385676e-06,
+      "loss": 0.9408,
+      "step": 20678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.047201567826112e-06,
+      "loss": 0.5829,
+      "step": 20679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0463526841954e-06,
+      "loss": 0.6881,
+      "step": 20680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04550380749972e-06,
+      "loss": 0.932,
+      "step": 20681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044654937745244e-06,
+      "loss": 1.2761,
+      "step": 20682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04380607493814e-06,
+      "loss": 0.9453,
+      "step": 20683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.042957219084591e-06,
+      "loss": 0.8043,
+      "step": 20684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.042108370190757e-06,
+      "loss": 0.712,
+      "step": 20685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04125952826282e-06,
+      "loss": 0.7296,
+      "step": 20686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.040410693306948e-06,
+      "loss": 0.7657,
+      "step": 20687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.039561865329317e-06,
+      "loss": 0.6816,
+      "step": 20688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.038713044336097e-06,
+      "loss": 0.8066,
+      "step": 20689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037864230333465e-06,
+      "loss": 0.7389,
+      "step": 20690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037015423327591e-06,
+      "loss": 0.879,
+      "step": 20691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036166623324645e-06,
+      "loss": 0.8976,
+      "step": 20692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.035317830330806e-06,
+      "loss": 1.0473,
+      "step": 20693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.034469044352242e-06,
+      "loss": 0.991,
+      "step": 20694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.033620265395127e-06,
+      "loss": 0.8681,
+      "step": 20695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.032771493465632e-06,
+      "loss": 0.9254,
+      "step": 20696
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031922728569929e-06,
+      "loss": 0.6584,
+      "step": 20697
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031073970714191e-06,
+      "loss": 0.8583,
+      "step": 20698
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.030225219904592e-06,
+      "loss": 0.9409,
+      "step": 20699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.029376476147303e-06,
+      "loss": 1.24,
+      "step": 20700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028527739448495e-06,
+      "loss": 0.9319,
+      "step": 20701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.027679009814341e-06,
+      "loss": 0.7318,
+      "step": 20702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.026830287251014e-06,
+      "loss": 0.495,
+      "step": 20703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.025981571764688e-06,
+      "loss": 0.7897,
+      "step": 20704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02513286336153e-06,
+      "loss": 0.7235,
+      "step": 20705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.024284162047713e-06,
+      "loss": 0.9782,
+      "step": 20706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.023435467829409e-06,
+      "loss": 1.1661,
+      "step": 20707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.022586780712792e-06,
+      "loss": 0.8561,
+      "step": 20708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.021738100704033e-06,
+      "loss": 0.8657,
+      "step": 20709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.020889427809301e-06,
+      "loss": 0.764,
+      "step": 20710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.020040762034773e-06,
+      "loss": 1.0792,
+      "step": 20711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.019192103386615e-06,
+      "loss": 0.7861,
+      "step": 20712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.018343451871002e-06,
+      "loss": 0.7365,
+      "step": 20713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.017494807494106e-06,
+      "loss": 0.7919,
+      "step": 20714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.016646170262096e-06,
+      "loss": 1.4598,
+      "step": 20715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.015797540181143e-06,
+      "loss": 0.579,
+      "step": 20716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01494891725742e-06,
+      "loss": 0.9346,
+      "step": 20717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.014100301497099e-06,
+      "loss": 0.7539,
+      "step": 20718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.013251692906346e-06,
+      "loss": 0.9242,
+      "step": 20719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.012403091491339e-06,
+      "loss": 0.9727,
+      "step": 20720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.011554497258248e-06,
+      "loss": 0.8354,
+      "step": 20721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.010705910213242e-06,
+      "loss": 0.7428,
+      "step": 20722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00985733036249e-06,
+      "loss": 0.9504,
+      "step": 20723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.009008757712171e-06,
+      "loss": 0.668,
+      "step": 20724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.008160192268444e-06,
+      "loss": 0.7703,
+      "step": 20725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00731163403749e-06,
+      "loss": 0.9866,
+      "step": 20726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.006463083025475e-06,
+      "loss": 0.8226,
+      "step": 20727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.005614539238568e-06,
+      "loss": 0.875,
+      "step": 20728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.004766002682946e-06,
+      "loss": 0.8391,
+      "step": 20729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003917473364774e-06,
+      "loss": 0.9608,
+      "step": 20730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003068951290226e-06,
+      "loss": 0.8408,
+      "step": 20731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00222043646547e-06,
+      "loss": 1.0042,
+      "step": 20732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.001371928896678e-06,
+      "loss": 1.0532,
+      "step": 20733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000523428590024e-06,
+      "loss": 1.0613,
+      "step": 20734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.999674935551672e-06,
+      "loss": 0.9496,
+      "step": 20735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.998826449787796e-06,
+      "loss": 1.1779,
+      "step": 20736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99797797130456e-06,
+      "loss": 1.0528,
+      "step": 20737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.997129500108145e-06,
+      "loss": 0.8918,
+      "step": 20738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996281036204714e-06,
+      "loss": 0.8821,
+      "step": 20739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.995432579600439e-06,
+      "loss": 1.2612,
+      "step": 20740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994584130301489e-06,
+      "loss": 0.4662,
+      "step": 20741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.993735688314036e-06,
+      "loss": 0.6846,
+      "step": 20742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.992887253644247e-06,
+      "loss": 0.4402,
+      "step": 20743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.992038826298298e-06,
+      "loss": 0.821,
+      "step": 20744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991190406282352e-06,
+      "loss": 0.7989,
+      "step": 20745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99034199360258e-06,
+      "loss": 1.136,
+      "step": 20746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.989493588265154e-06,
+      "loss": 1.0733,
+      "step": 20747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.988645190276243e-06,
+      "loss": 0.9127,
+      "step": 20748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.987796799642018e-06,
+      "loss": 1.2839,
+      "step": 20749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.986948416368642e-06,
+      "loss": 1.1311,
+      "step": 20750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.986100040462295e-06,
+      "loss": 0.9141,
+      "step": 20751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.985251671929139e-06,
+      "loss": 0.3131,
+      "step": 20752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.984403310775343e-06,
+      "loss": 0.7272,
+      "step": 20753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.983554957007088e-06,
+      "loss": 1.2797,
+      "step": 20754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.982706610630525e-06,
+      "loss": 1.0593,
+      "step": 20755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.981858271651836e-06,
+      "loss": 0.8172,
+      "step": 20756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.981009940077186e-06,
+      "loss": 0.8002,
+      "step": 20757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.980161615912746e-06,
+      "loss": 0.8619,
+      "step": 20758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97931329916468e-06,
+      "loss": 0.7541,
+      "step": 20759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.978464989839165e-06,
+      "loss": 0.9936,
+      "step": 20760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.977616687942366e-06,
+      "loss": 0.8352,
+      "step": 20761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.976768393480449e-06,
+      "loss": 0.8126,
+      "step": 20762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.975920106459588e-06,
+      "loss": 1.1288,
+      "step": 20763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.975071826885952e-06,
+      "loss": 0.9181,
+      "step": 20764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.974223554765707e-06,
+      "loss": 0.8019,
+      "step": 20765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97337529010502e-06,
+      "loss": 1.2304,
+      "step": 20766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.972527032910062e-06,
+      "loss": 1.1572,
+      "step": 20767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.971678783187e-06,
+      "loss": 0.7318,
+      "step": 20768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.970830540942006e-06,
+      "loss": 1.124,
+      "step": 20769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.969982306181246e-06,
+      "loss": 1.1524,
+      "step": 20770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.969134078910889e-06,
+      "loss": 0.7719,
+      "step": 20771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968285859137103e-06,
+      "loss": 0.9643,
+      "step": 20772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.967437646866058e-06,
+      "loss": 0.8824,
+      "step": 20773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.966589442103924e-06,
+      "loss": 0.8634,
+      "step": 20774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.965741244856864e-06,
+      "loss": 0.9342,
+      "step": 20775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.964893055131047e-06,
+      "loss": 0.7966,
+      "step": 20776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.964044872932641e-06,
+      "loss": 0.9381,
+      "step": 20777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.963196698267819e-06,
+      "loss": 0.9413,
+      "step": 20778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.962348531142745e-06,
+      "loss": 0.942,
+      "step": 20779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.961500371563585e-06,
+      "loss": 1.1459,
+      "step": 20780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.960652219536512e-06,
+      "loss": 0.9481,
+      "step": 20781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.959804075067693e-06,
+      "loss": 0.8408,
+      "step": 20782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.958955938163293e-06,
+      "loss": 0.9976,
+      "step": 20783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.958107808829483e-06,
+      "loss": 0.8167,
+      "step": 20784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.957259687072428e-06,
+      "loss": 0.9605,
+      "step": 20785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.956411572898292e-06,
+      "loss": 0.9805,
+      "step": 20786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95556346631325e-06,
+      "loss": 0.7792,
+      "step": 20787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.954715367323468e-06,
+      "loss": 0.7222,
+      "step": 20788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.953867275935108e-06,
+      "loss": 0.9727,
+      "step": 20789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.953019192154344e-06,
+      "loss": 0.8085,
+      "step": 20790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95217111598734e-06,
+      "loss": 0.8256,
+      "step": 20791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.951323047440265e-06,
+      "loss": 0.887,
+      "step": 20792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.950474986519283e-06,
+      "loss": 0.9761,
+      "step": 20793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94962693323057e-06,
+      "loss": 0.758,
+      "step": 20794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94877888758028e-06,
+      "loss": 0.765,
+      "step": 20795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.947930849574588e-06,
+      "loss": 1.1056,
+      "step": 20796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94708281921966e-06,
+      "loss": 0.8852,
+      "step": 20797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94623479652166e-06,
+      "loss": 1.0796,
+      "step": 20798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.945386781486762e-06,
+      "loss": 0.6786,
+      "step": 20799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.944538774121125e-06,
+      "loss": 0.942,
+      "step": 20800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943690774430922e-06,
+      "loss": 0.8569,
+      "step": 20801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.942842782422314e-06,
+      "loss": 0.9981,
+      "step": 20802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.941994798101474e-06,
+      "loss": 0.8054,
+      "step": 20803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.941146821474566e-06,
+      "loss": 0.734,
+      "step": 20804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.940298852547753e-06,
+      "loss": 0.8781,
+      "step": 20805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.939450891327207e-06,
+      "loss": 1.009,
+      "step": 20806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938602937819087e-06,
+      "loss": 1.066,
+      "step": 20807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.937754992029569e-06,
+      "loss": 0.9865,
+      "step": 20808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.936907053964811e-06,
+      "loss": 0.7736,
+      "step": 20809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.936059123630986e-06,
+      "loss": 0.9916,
+      "step": 20810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.935211201034255e-06,
+      "loss": 0.7834,
+      "step": 20811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.934363286180787e-06,
+      "loss": 1.132,
+      "step": 20812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.933515379076749e-06,
+      "loss": 0.9241,
+      "step": 20813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.932667479728307e-06,
+      "loss": 0.7018,
+      "step": 20814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.931819588141623e-06,
+      "loss": 0.7903,
+      "step": 20815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.930971704322863e-06,
+      "loss": 0.8805,
+      "step": 20816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.9301238282782e-06,
+      "loss": 0.8655,
+      "step": 20817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.929275960013794e-06,
+      "loss": 0.8155,
+      "step": 20818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.928428099535811e-06,
+      "loss": 0.6977,
+      "step": 20819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927580246850418e-06,
+      "loss": 0.7097,
+      "step": 20820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.926732401963781e-06,
+      "loss": 0.7188,
+      "step": 20821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.925884564882066e-06,
+      "loss": 0.9089,
+      "step": 20822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.925036735611435e-06,
+      "loss": 0.769,
+      "step": 20823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.924188914158063e-06,
+      "loss": 0.7503,
+      "step": 20824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.923341100528102e-06,
+      "loss": 0.7543,
+      "step": 20825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922493294727727e-06,
+      "loss": 1.0674,
+      "step": 20826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.9216454967631e-06,
+      "loss": 0.9424,
+      "step": 20827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.920797706640387e-06,
+      "loss": 0.9885,
+      "step": 20828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.919949924365749e-06,
+      "loss": 0.6364,
+      "step": 20829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.919102149945358e-06,
+      "loss": 0.8688,
+      "step": 20830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.918254383385377e-06,
+      "loss": 0.8826,
+      "step": 20831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.917406624691967e-06,
+      "loss": 0.8928,
+      "step": 20832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.9165588738713e-06,
+      "loss": 1.0874,
+      "step": 20833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.915711130929539e-06,
+      "loss": 0.8163,
+      "step": 20834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914863395872844e-06,
+      "loss": 0.833,
+      "step": 20835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914015668707383e-06,
+      "loss": 0.9045,
+      "step": 20836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.913167949439321e-06,
+      "loss": 0.7799,
+      "step": 20837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912320238074821e-06,
+      "loss": 0.6662,
+      "step": 20838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.91147253462005e-06,
+      "loss": 0.9103,
+      "step": 20839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.910624839081171e-06,
+      "loss": 0.5343,
+      "step": 20840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90977715146435e-06,
+      "loss": 0.8602,
+      "step": 20841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.908929471775749e-06,
+      "loss": 0.7559,
+      "step": 20842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.908081800021534e-06,
+      "loss": 0.7114,
+      "step": 20843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.907234136207875e-06,
+      "loss": 0.8889,
+      "step": 20844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.906386480340927e-06,
+      "loss": 1.1352,
+      "step": 20845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.905538832426857e-06,
+      "loss": 0.7772,
+      "step": 20846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.904691192471829e-06,
+      "loss": 0.8459,
+      "step": 20847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90384356048201e-06,
+      "loss": 1.1297,
+      "step": 20848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.902995936463564e-06,
+      "loss": 0.6793,
+      "step": 20849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90214832042265e-06,
+      "loss": 0.805,
+      "step": 20850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901300712365436e-06,
+      "loss": 0.9961,
+      "step": 20851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.900453112298087e-06,
+      "loss": 0.9245,
+      "step": 20852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.899605520226764e-06,
+      "loss": 0.7716,
+      "step": 20853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898757936157635e-06,
+      "loss": 1.1922,
+      "step": 20854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89791036009686e-06,
+      "loss": 0.9075,
+      "step": 20855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.897062792050596e-06,
+      "loss": 0.7584,
+      "step": 20856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89621523202502e-06,
+      "loss": 0.8021,
+      "step": 20857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.89536768002629e-06,
+      "loss": 0.9967,
+      "step": 20858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.894520136060567e-06,
+      "loss": 0.8603,
+      "step": 20859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.893672600134013e-06,
+      "loss": 1.047,
+      "step": 20860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8928250722528e-06,
+      "loss": 0.799,
+      "step": 20861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.891977552423083e-06,
+      "loss": 0.9366,
+      "step": 20862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.891130040651028e-06,
+      "loss": 0.8811,
+      "step": 20863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890282536942804e-06,
+      "loss": 0.9848,
+      "step": 20864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.889435041304565e-06,
+      "loss": 1.1324,
+      "step": 20865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.888587553742476e-06,
+      "loss": 0.9656,
+      "step": 20866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.887740074262703e-06,
+      "loss": 0.7693,
+      "step": 20867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.886892602871408e-06,
+      "loss": 0.9852,
+      "step": 20868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.886045139574752e-06,
+      "loss": 1.0682,
+      "step": 20869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8851976843789e-06,
+      "loss": 0.9518,
+      "step": 20870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.884350237290016e-06,
+      "loss": 0.7208,
+      "step": 20871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.88350279831426e-06,
+      "loss": 0.9874,
+      "step": 20872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.882655367457794e-06,
+      "loss": 0.7143,
+      "step": 20873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.881807944726786e-06,
+      "loss": 0.8019,
+      "step": 20874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.880960530127394e-06,
+      "loss": 0.7529,
+      "step": 20875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.880113123665779e-06,
+      "loss": 0.7885,
+      "step": 20876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.879265725348107e-06,
+      "loss": 1.0382,
+      "step": 20877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.878418335180538e-06,
+      "loss": 1.0467,
+      "step": 20878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.877570953169236e-06,
+      "loss": 1.1719,
+      "step": 20879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.876723579320363e-06,
+      "loss": 0.7769,
+      "step": 20880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.875876213640078e-06,
+      "loss": 0.9041,
+      "step": 20881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87502885613455e-06,
+      "loss": 0.6575,
+      "step": 20882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.874181506809936e-06,
+      "loss": 0.8861,
+      "step": 20883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8733341656724e-06,
+      "loss": 0.8567,
+      "step": 20884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.872486832728103e-06,
+      "loss": 0.614,
+      "step": 20885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.871639507983206e-06,
+      "loss": 0.9302,
+      "step": 20886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87079219144387e-06,
+      "loss": 1.0118,
+      "step": 20887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869944883116261e-06,
+      "loss": 0.659,
+      "step": 20888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869097583006538e-06,
+      "loss": 0.748,
+      "step": 20889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86825029112086e-06,
+      "loss": 1.1722,
+      "step": 20890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867403007465395e-06,
+      "loss": 0.8133,
+      "step": 20891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8665557320463e-06,
+      "loss": 0.8679,
+      "step": 20892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.865708464869738e-06,
+      "loss": 0.6538,
+      "step": 20893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.864861205941872e-06,
+      "loss": 0.8029,
+      "step": 20894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86401395526886e-06,
+      "loss": 0.5622,
+      "step": 20895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.863166712856862e-06,
+      "loss": 0.8734,
+      "step": 20896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.862319478712044e-06,
+      "loss": 0.8517,
+      "step": 20897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.861472252840564e-06,
+      "loss": 0.6428,
+      "step": 20898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.860625035248584e-06,
+      "loss": 0.7717,
+      "step": 20899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.859777825942265e-06,
+      "loss": 0.9749,
+      "step": 20900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858930624927772e-06,
+      "loss": 0.9271,
+      "step": 20901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85808343221126e-06,
+      "loss": 1.1374,
+      "step": 20902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85723624779889e-06,
+      "loss": 1.2441,
+      "step": 20903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85638907169683e-06,
+      "loss": 0.8843,
+      "step": 20904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85554190391123e-06,
+      "loss": 0.7339,
+      "step": 20905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85469474444826e-06,
+      "loss": 0.7569,
+      "step": 20906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853847593314077e-06,
+      "loss": 0.7162,
+      "step": 20907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853000450514839e-06,
+      "loss": 0.7992,
+      "step": 20908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85215331605671e-06,
+      "loss": 1.1507,
+      "step": 20909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85130618994585e-06,
+      "loss": 0.919,
+      "step": 20910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.850459072188419e-06,
+      "loss": 1.0413,
+      "step": 20911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.849611962790575e-06,
+      "loss": 0.9219,
+      "step": 20912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.848764861758483e-06,
+      "loss": 1.0719,
+      "step": 20913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.847917769098305e-06,
+      "loss": 0.9212,
+      "step": 20914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.847070684816192e-06,
+      "loss": 1.1067,
+      "step": 20915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84622360891831e-06,
+      "loss": 0.6196,
+      "step": 20916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.845376541410815e-06,
+      "loss": 1.0935,
+      "step": 20917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.844529482299873e-06,
+      "loss": 0.8626,
+      "step": 20918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84368243159164e-06,
+      "loss": 1.018,
+      "step": 20919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.842835389292279e-06,
+      "loss": 0.8733,
+      "step": 20920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.841988355407943e-06,
+      "loss": 0.7858,
+      "step": 20921
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8411413299448e-06,
+      "loss": 0.7867,
+      "step": 20922
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.840294312909005e-06,
+      "loss": 0.6354,
+      "step": 20923
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.839447304306721e-06,
+      "loss": 0.7543,
+      "step": 20924
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.838600304144102e-06,
+      "loss": 1.089,
+      "step": 20925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83775331242731e-06,
+      "loss": 0.7897,
+      "step": 20926
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.836906329162505e-06,
+      "loss": 1.246,
+      "step": 20927
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83605935435585e-06,
+      "loss": 1.064,
+      "step": 20928
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.835212388013498e-06,
+      "loss": 0.7142,
+      "step": 20929
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.834365430141608e-06,
+      "loss": 1.054,
+      "step": 20930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.833518480746346e-06,
+      "loss": 0.784,
+      "step": 20931
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.832671539833865e-06,
+      "loss": 1.1197,
+      "step": 20932
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.831824607410326e-06,
+      "loss": 0.9028,
+      "step": 20933
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830977683481894e-06,
+      "loss": 0.7341,
+      "step": 20934
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830130768054717e-06,
+      "loss": 0.6999,
+      "step": 20935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.829283861134959e-06,
+      "loss": 0.538,
+      "step": 20936
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82843696272878e-06,
+      "loss": 0.7758,
+      "step": 20937
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.827590072842337e-06,
+      "loss": 0.9059,
+      "step": 20938
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.826743191481787e-06,
+      "loss": 0.99,
+      "step": 20939
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.825896318653294e-06,
+      "loss": 0.7898,
+      "step": 20940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.825049454363011e-06,
+      "loss": 1.4131,
+      "step": 20941
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8242025986171e-06,
+      "loss": 0.9266,
+      "step": 20942
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.823355751421718e-06,
+      "loss": 0.9644,
+      "step": 20943
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.822508912783025e-06,
+      "loss": 1.1015,
+      "step": 20944
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.821662082707177e-06,
+      "loss": 0.9387,
+      "step": 20945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.820815261200334e-06,
+      "loss": 0.8218,
+      "step": 20946
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819968448268653e-06,
+      "loss": 0.9267,
+      "step": 20947
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819121643918291e-06,
+      "loss": 0.939,
+      "step": 20948
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.818274848155409e-06,
+      "loss": 1.0459,
+      "step": 20949
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.817428060986162e-06,
+      "loss": 1.0465,
+      "step": 20950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.816581282416708e-06,
+      "loss": 0.8242,
+      "step": 20951
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81573451245321e-06,
+      "loss": 0.9476,
+      "step": 20952
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.814887751101821e-06,
+      "loss": 0.9802,
+      "step": 20953
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.814040998368702e-06,
+      "loss": 0.6726,
+      "step": 20954
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.813194254260006e-06,
+      "loss": 0.9722,
+      "step": 20955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.812347518781894e-06,
+      "loss": 0.6867,
+      "step": 20956
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81150079194052e-06,
+      "loss": 0.8539,
+      "step": 20957
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.810654073742047e-06,
+      "loss": 0.7374,
+      "step": 20958
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80980736419263e-06,
+      "loss": 1.0204,
+      "step": 20959
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.808960663298423e-06,
+      "loss": 0.9013,
+      "step": 20960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.808113971065588e-06,
+      "loss": 0.7471,
+      "step": 20961
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.807267287500281e-06,
+      "loss": 1.2792,
+      "step": 20962
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80642061260866e-06,
+      "loss": 0.8957,
+      "step": 20963
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.805573946396883e-06,
+      "loss": 0.5491,
+      "step": 20964
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.804727288871101e-06,
+      "loss": 0.9687,
+      "step": 20965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.803880640037472e-06,
+      "loss": 0.7263,
+      "step": 20966
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80303399990216e-06,
+      "loss": 1.1166,
+      "step": 20967
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.802187368471317e-06,
+      "loss": 0.7946,
+      "step": 20968
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8013407457511e-06,
+      "loss": 1.0645,
+      "step": 20969
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.800494131747667e-06,
+      "loss": 0.6255,
+      "step": 20970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.799647526467174e-06,
+      "loss": 0.9804,
+      "step": 20971
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.798800929915776e-06,
+      "loss": 0.8966,
+      "step": 20972
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79795434209963e-06,
+      "loss": 0.9281,
+      "step": 20973
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7971077630249e-06,
+      "loss": 0.6158,
+      "step": 20974
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79626119269773e-06,
+      "loss": 0.8955,
+      "step": 20975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.795414631124283e-06,
+      "loss": 1.019,
+      "step": 20976
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.794568078310716e-06,
+      "loss": 0.6333,
+      "step": 20977
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.793721534263181e-06,
+      "loss": 0.8866,
+      "step": 20978
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79287499898784e-06,
+      "loss": 1.0104,
+      "step": 20979
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.792028472490844e-06,
+      "loss": 0.9569,
+      "step": 20980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791181954778353e-06,
+      "loss": 0.6803,
+      "step": 20981
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.790335445856519e-06,
+      "loss": 0.7318,
+      "step": 20982
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.789488945731501e-06,
+      "loss": 1.1798,
+      "step": 20983
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.788642454409457e-06,
+      "loss": 0.8176,
+      "step": 20984
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.787795971896536e-06,
+      "loss": 0.6089,
+      "step": 20985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.786949498198897e-06,
+      "loss": 0.8121,
+      "step": 20986
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.786103033322697e-06,
+      "loss": 0.9947,
+      "step": 20987
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785256577274089e-06,
+      "loss": 0.7987,
+      "step": 20988
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.784410130059232e-06,
+      "loss": 1.0803,
+      "step": 20989
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.78356369168428e-06,
+      "loss": 0.6237,
+      "step": 20990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.782717262155384e-06,
+      "loss": 0.8044,
+      "step": 20991
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.781870841478708e-06,
+      "loss": 0.6682,
+      "step": 20992
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7810244296604e-06,
+      "loss": 0.9047,
+      "step": 20993
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.780178026706621e-06,
+      "loss": 0.7392,
+      "step": 20994
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77933163262352e-06,
+      "loss": 0.6077,
+      "step": 20995
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.778485247417253e-06,
+      "loss": 0.6187,
+      "step": 20996
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77763887109398e-06,
+      "loss": 0.6756,
+      "step": 20997
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.776792503659853e-06,
+      "loss": 1.4646,
+      "step": 20998
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.775946145121024e-06,
+      "loss": 0.8452,
+      "step": 20999
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.775099795483651e-06,
+      "loss": 0.7794,
+      "step": 21000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77425345475389e-06,
+      "loss": 0.9383,
+      "step": 21001
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.773407122937893e-06,
+      "loss": 1.2124,
+      "step": 21002
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.772560800041814e-06,
+      "loss": 0.874,
+      "step": 21003
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.771714486071816e-06,
+      "loss": 0.6133,
+      "step": 21004
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77086818103404e-06,
+      "loss": 1.007,
+      "step": 21005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77002188493465e-06,
+      "loss": 0.7177,
+      "step": 21006
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.769175597779796e-06,
+      "loss": 0.9917,
+      "step": 21007
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.768329319575634e-06,
+      "loss": 1.1162,
+      "step": 21008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.767483050328316e-06,
+      "loss": 0.822,
+      "step": 21009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.766636790044002e-06,
+      "loss": 0.8704,
+      "step": 21010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76579053872884e-06,
+      "loss": 1.014,
+      "step": 21011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.764944296388985e-06,
+      "loss": 0.9683,
+      "step": 21012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.764098063030596e-06,
+      "loss": 0.9344,
+      "step": 21013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763251838659822e-06,
+      "loss": 0.7761,
+      "step": 21014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.762405623282817e-06,
+      "loss": 0.8922,
+      "step": 21015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.761559416905736e-06,
+      "loss": 0.7839,
+      "step": 21016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.760713219534734e-06,
+      "loss": 0.8892,
+      "step": 21017
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.759867031175958e-06,
+      "loss": 1.0044,
+      "step": 21018
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.759020851835572e-06,
+      "loss": 0.8987,
+      "step": 21019
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.758174681519722e-06,
+      "loss": 1.1094,
+      "step": 21020
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.757328520234561e-06,
+      "loss": 0.6385,
+      "step": 21021
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.756482367986248e-06,
+      "loss": 0.9171,
+      "step": 21022
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.755636224780933e-06,
+      "loss": 0.8606,
+      "step": 21023
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.754790090624772e-06,
+      "loss": 0.9877,
+      "step": 21024
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.753943965523914e-06,
+      "loss": 1.0379,
+      "step": 21025
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.753097849484514e-06,
+      "loss": 0.9928,
+      "step": 21026
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.752251742512722e-06,
+      "loss": 1.0495,
+      "step": 21027
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.751405644614696e-06,
+      "loss": 1.1096,
+      "step": 21028
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.750559555796588e-06,
+      "loss": 0.8878,
+      "step": 21029
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.749713476064547e-06,
+      "loss": 0.8786,
+      "step": 21030
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.748867405424729e-06,
+      "loss": 0.9381,
+      "step": 21031
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.748021343883288e-06,
+      "loss": 1.0464,
+      "step": 21032
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.747175291446373e-06,
+      "loss": 0.6984,
+      "step": 21033
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.746329248120142e-06,
+      "loss": 1.0084,
+      "step": 21034
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.745483213910742e-06,
+      "loss": 1.0026,
+      "step": 21035
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.744637188824326e-06,
+      "loss": 1.0244,
+      "step": 21036
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.743791172867047e-06,
+      "loss": 0.9324,
+      "step": 21037
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74294516604506e-06,
+      "loss": 0.9805,
+      "step": 21038
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.742099168364513e-06,
+      "loss": 1.3403,
+      "step": 21039
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.741253179831563e-06,
+      "loss": 0.9751,
+      "step": 21040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.740407200452359e-06,
+      "loss": 0.7721,
+      "step": 21041
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.739561230233054e-06,
+      "loss": 0.9708,
+      "step": 21042
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.738715269179799e-06,
+      "loss": 0.749,
+      "step": 21043
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73786931729875e-06,
+      "loss": 1.1232,
+      "step": 21044
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.737023374596051e-06,
+      "loss": 1.0495,
+      "step": 21045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73617744107786e-06,
+      "loss": 0.691,
+      "step": 21046
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735331516750328e-06,
+      "loss": 0.7569,
+      "step": 21047
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.734485601619603e-06,
+      "loss": 0.9917,
+      "step": 21048
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73363969569184e-06,
+      "loss": 0.823,
+      "step": 21049
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.73279379897319e-06,
+      "loss": 1.2118,
+      "step": 21050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.731947911469805e-06,
+      "loss": 1.0275,
+      "step": 21051
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.731102033187834e-06,
+      "loss": 0.8992,
+      "step": 21052
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.730256164133431e-06,
+      "loss": 0.842,
+      "step": 21053
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.729410304312748e-06,
+      "loss": 1.0811,
+      "step": 21054
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.728564453731934e-06,
+      "loss": 0.6683,
+      "step": 21055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72771861239714e-06,
+      "loss": 0.8044,
+      "step": 21056
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.726872780314515e-06,
+      "loss": 0.6605,
+      "step": 21057
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.726026957490215e-06,
+      "loss": 0.8163,
+      "step": 21058
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.725181143930387e-06,
+      "loss": 0.8046,
+      "step": 21059
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.724335339641185e-06,
+      "loss": 0.7664,
+      "step": 21060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.723489544628753e-06,
+      "loss": 0.7736,
+      "step": 21061
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.722643758899252e-06,
+      "loss": 0.9178,
+      "step": 21062
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.721797982458826e-06,
+      "loss": 0.6736,
+      "step": 21063
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.720952215313631e-06,
+      "loss": 0.8204,
+      "step": 21064
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72010645746981e-06,
+      "loss": 1.0606,
+      "step": 21065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.719260708933514e-06,
+      "loss": 0.5653,
+      "step": 21066
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.7184149697109e-06,
+      "loss": 0.5794,
+      "step": 21067
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.717569239808115e-06,
+      "loss": 0.9312,
+      "step": 21068
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.716723519231307e-06,
+      "loss": 0.7667,
+      "step": 21069
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.715877807986627e-06,
+      "loss": 1.2548,
+      "step": 21070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.715032106080227e-06,
+      "loss": 0.7169,
+      "step": 21071
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.714186413518258e-06,
+      "loss": 0.8967,
+      "step": 21072
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.713340730306867e-06,
+      "loss": 0.7273,
+      "step": 21073
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71249505645221e-06,
+      "loss": 0.9725,
+      "step": 21074
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.711649391960424e-06,
+      "loss": 0.9463,
+      "step": 21075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71080373683767e-06,
+      "loss": 0.9471,
+      "step": 21076
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709958091090096e-06,
+      "loss": 0.9063,
+      "step": 21077
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709112454723848e-06,
+      "loss": 0.893,
+      "step": 21078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.708266827745077e-06,
+      "loss": 0.5415,
+      "step": 21079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.707421210159935e-06,
+      "loss": 0.949,
+      "step": 21080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.706575601974568e-06,
+      "loss": 0.9043,
+      "step": 21081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705730003195126e-06,
+      "loss": 0.8573,
+      "step": 21082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70488441382776e-06,
+      "loss": 1.3361,
+      "step": 21083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.704038833878621e-06,
+      "loss": 0.7427,
+      "step": 21084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.703193263353854e-06,
+      "loss": 0.8456,
+      "step": 21085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.702347702259609e-06,
+      "loss": 1.0041,
+      "step": 21086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.701502150602036e-06,
+      "loss": 0.789,
+      "step": 21087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.700656608387282e-06,
+      "loss": 0.9262,
+      "step": 21088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699811075621498e-06,
+      "loss": 0.9363,
+      "step": 21089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.698965552310834e-06,
+      "loss": 0.8491,
+      "step": 21090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.698120038461434e-06,
+      "loss": 0.8109,
+      "step": 21091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.697274534079452e-06,
+      "loss": 0.8065,
+      "step": 21092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.696429039171033e-06,
+      "loss": 1.04,
+      "step": 21093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.695583553742331e-06,
+      "loss": 0.8435,
+      "step": 21094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.694738077799487e-06,
+      "loss": 1.0098,
+      "step": 21095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.693892611348653e-06,
+      "loss": 0.9512,
+      "step": 21096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.693047154395975e-06,
+      "loss": 0.7106,
+      "step": 21097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692201706947603e-06,
+      "loss": 0.8329,
+      "step": 21098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69135626900969e-06,
+      "loss": 1.043,
+      "step": 21099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.690510840588373e-06,
+      "loss": 0.7304,
+      "step": 21100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68966542168981e-06,
+      "loss": 0.8075,
+      "step": 21101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688820012320147e-06,
+      "loss": 1.1309,
+      "step": 21102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.687974612485529e-06,
+      "loss": 1.0317,
+      "step": 21103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.687129222192108e-06,
+      "loss": 0.9319,
+      "step": 21104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.686283841446027e-06,
+      "loss": 0.9473,
+      "step": 21105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.685438470253433e-06,
+      "loss": 0.8818,
+      "step": 21106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.684593108620479e-06,
+      "loss": 1.0576,
+      "step": 21107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.683747756553311e-06,
+      "loss": 0.7103,
+      "step": 21108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.682902414058072e-06,
+      "loss": 1.2749,
+      "step": 21109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.682057081140915e-06,
+      "loss": 0.8814,
+      "step": 21110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.681211757807985e-06,
+      "loss": 0.8182,
+      "step": 21111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.680366444065431e-06,
+      "loss": 0.9615,
+      "step": 21112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.679521139919397e-06,
+      "loss": 0.6755,
+      "step": 21113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.678675845376036e-06,
+      "loss": 0.6478,
+      "step": 21114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.677830560441486e-06,
+      "loss": 0.8667,
+      "step": 21115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.676985285121902e-06,
+      "loss": 0.7931,
+      "step": 21116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.676140019423427e-06,
+      "loss": 0.8714,
+      "step": 21117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.675294763352206e-06,
+      "loss": 0.8806,
+      "step": 21118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.674449516914392e-06,
+      "loss": 0.8124,
+      "step": 21119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.673604280116127e-06,
+      "loss": 1.0188,
+      "step": 21120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.67275905296356e-06,
+      "loss": 1.0619,
+      "step": 21121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671913835462833e-06,
+      "loss": 1.0315,
+      "step": 21122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671068627620099e-06,
+      "loss": 1.1029,
+      "step": 21123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.670223429441505e-06,
+      "loss": 0.8494,
+      "step": 21124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.669378240933189e-06,
+      "loss": 0.7449,
+      "step": 21125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668533062101303e-06,
+      "loss": 0.881,
+      "step": 21126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.667687892951992e-06,
+      "loss": 0.8426,
+      "step": 21127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.666842733491404e-06,
+      "loss": 0.7306,
+      "step": 21128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.665997583725684e-06,
+      "loss": 0.8845,
+      "step": 21129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.665152443660977e-06,
+      "loss": 1.1897,
+      "step": 21130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.664307313303427e-06,
+      "loss": 0.8389,
+      "step": 21131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663462192659186e-06,
+      "loss": 0.8183,
+      "step": 21132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.662617081734394e-06,
+      "loss": 0.6256,
+      "step": 21133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.661771980535204e-06,
+      "loss": 0.8185,
+      "step": 21134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660926889067753e-06,
+      "loss": 1.2663,
+      "step": 21135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.66008180733819e-06,
+      "loss": 0.7665,
+      "step": 21136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.659236735352661e-06,
+      "loss": 0.8227,
+      "step": 21137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.658391673117312e-06,
+      "loss": 1.0994,
+      "step": 21138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.657546620638288e-06,
+      "loss": 1.4248,
+      "step": 21139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.656701577921733e-06,
+      "loss": 0.921,
+      "step": 21140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.655856544973793e-06,
+      "loss": 0.6846,
+      "step": 21141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.655011521800615e-06,
+      "loss": 0.5646,
+      "step": 21142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.654166508408341e-06,
+      "loss": 0.8009,
+      "step": 21143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.653321504803123e-06,
+      "loss": 0.8211,
+      "step": 21144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.652476510991094e-06,
+      "loss": 0.7899,
+      "step": 21145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.651631526978407e-06,
+      "loss": 0.7283,
+      "step": 21146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.650786552771206e-06,
+      "loss": 0.5595,
+      "step": 21147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649941588375637e-06,
+      "loss": 0.9348,
+      "step": 21148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649096633797838e-06,
+      "loss": 0.8791,
+      "step": 21149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.648251689043961e-06,
+      "loss": 1.0436,
+      "step": 21150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.647406754120148e-06,
+      "loss": 0.9626,
+      "step": 21151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.646561829032541e-06,
+      "loss": 0.7257,
+      "step": 21152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.645716913787288e-06,
+      "loss": 1.3102,
+      "step": 21153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.644872008390537e-06,
+      "loss": 1.0343,
+      "step": 21154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.644027112848422e-06,
+      "loss": 0.9789,
+      "step": 21155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.643182227167094e-06,
+      "loss": 0.8217,
+      "step": 21156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.642337351352693e-06,
+      "loss": 0.7298,
+      "step": 21157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.641492485411367e-06,
+      "loss": 0.7586,
+      "step": 21158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64064762934926e-06,
+      "loss": 0.9022,
+      "step": 21159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.639802783172513e-06,
+      "loss": 0.8064,
+      "step": 21160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63895794688727e-06,
+      "loss": 0.8967,
+      "step": 21161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.638113120499676e-06,
+      "loss": 0.6954,
+      "step": 21162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.637268304015878e-06,
+      "loss": 0.607,
+      "step": 21163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.636423497442017e-06,
+      "loss": 0.8208,
+      "step": 21164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.635578700784232e-06,
+      "loss": 1.0555,
+      "step": 21165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634733914048672e-06,
+      "loss": 1.1942,
+      "step": 21166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.633889137241476e-06,
+      "loss": 0.9912,
+      "step": 21167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.633044370368792e-06,
+      "loss": 0.6729,
+      "step": 21168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.632199613436762e-06,
+      "loss": 0.9557,
+      "step": 21169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.631354866451524e-06,
+      "loss": 1.0573,
+      "step": 21170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63051012941923e-06,
+      "loss": 0.8465,
+      "step": 21171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.629665402346017e-06,
+      "loss": 1.0742,
+      "step": 21172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.628820685238029e-06,
+      "loss": 0.7842,
+      "step": 21173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.627975978101412e-06,
+      "loss": 1.103,
+      "step": 21174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.627131280942305e-06,
+      "loss": 0.9487,
+      "step": 21175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.626286593766848e-06,
+      "loss": 0.804,
+      "step": 21176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62544191658119e-06,
+      "loss": 0.6521,
+      "step": 21177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62459724939147e-06,
+      "loss": 0.7344,
+      "step": 21178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.623752592203831e-06,
+      "loss": 0.9335,
+      "step": 21179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.622907945024418e-06,
+      "loss": 1.3736,
+      "step": 21180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62206330785937e-06,
+      "loss": 1.4568,
+      "step": 21181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.621218680714832e-06,
+      "loss": 0.7921,
+      "step": 21182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.620374063596943e-06,
+      "loss": 0.8207,
+      "step": 21183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.619529456511853e-06,
+      "loss": 0.913,
+      "step": 21184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61868485946569e-06,
+      "loss": 1.0996,
+      "step": 21185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.617840272464609e-06,
+      "loss": 0.9135,
+      "step": 21186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.616995695514745e-06,
+      "loss": 0.9804,
+      "step": 21187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.616151128622241e-06,
+      "loss": 1.1443,
+      "step": 21188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.615306571793242e-06,
+      "loss": 0.7756,
+      "step": 21189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.614462025033886e-06,
+      "loss": 0.8503,
+      "step": 21190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.613617488350317e-06,
+      "loss": 1.0917,
+      "step": 21191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.612772961748673e-06,
+      "loss": 0.6751,
+      "step": 21192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.611928445235101e-06,
+      "loss": 0.7709,
+      "step": 21193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61108393881574e-06,
+      "loss": 1.215,
+      "step": 21194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61023944249673e-06,
+      "loss": 0.6994,
+      "step": 21195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.609394956284213e-06,
+      "loss": 1.009,
+      "step": 21196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.608550480184327e-06,
+      "loss": 0.8724,
+      "step": 21197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60770601420322e-06,
+      "loss": 0.8882,
+      "step": 21198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.606861558347028e-06,
+      "loss": 0.8073,
+      "step": 21199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.606017112621895e-06,
+      "loss": 0.7647,
+      "step": 21200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.605172677033957e-06,
+      "loss": 1.0829,
+      "step": 21201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604328251589361e-06,
+      "loss": 0.7846,
+      "step": 21202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.603483836294245e-06,
+      "loss": 0.7486,
+      "step": 21203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.602639431154752e-06,
+      "loss": 0.7529,
+      "step": 21204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.601795036177017e-06,
+      "loss": 0.9071,
+      "step": 21205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600950651367184e-06,
+      "loss": 0.8729,
+      "step": 21206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600106276731392e-06,
+      "loss": 0.6972,
+      "step": 21207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.599261912275784e-06,
+      "loss": 0.9434,
+      "step": 21208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5984175580065e-06,
+      "loss": 0.754,
+      "step": 21209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597573213929677e-06,
+      "loss": 0.8675,
+      "step": 21210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.596728880051459e-06,
+      "loss": 0.8969,
+      "step": 21211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595884556377984e-06,
+      "loss": 1.0,
+      "step": 21212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.59504024291539e-06,
+      "loss": 0.7192,
+      "step": 21213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.594195939669826e-06,
+      "loss": 1.2179,
+      "step": 21214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.59335164664742e-06,
+      "loss": 1.0964,
+      "step": 21215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.592507363854316e-06,
+      "loss": 0.8983,
+      "step": 21216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.591663091296658e-06,
+      "loss": 0.8469,
+      "step": 21217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.590818828980581e-06,
+      "loss": 1.0174,
+      "step": 21218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589974576912224e-06,
+      "loss": 0.85,
+      "step": 21219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589130335097732e-06,
+      "loss": 0.7686,
+      "step": 21220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58828610354324e-06,
+      "loss": 0.6049,
+      "step": 21221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.587441882254885e-06,
+      "loss": 1.066,
+      "step": 21222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.586597671238811e-06,
+      "loss": 0.864,
+      "step": 21223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58575347050116e-06,
+      "loss": 1.0491,
+      "step": 21224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.584909280048064e-06,
+      "loss": 0.749,
+      "step": 21225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.584065099885666e-06,
+      "loss": 0.7358,
+      "step": 21226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.583220930020102e-06,
+      "loss": 0.7338,
+      "step": 21227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58237677045751e-06,
+      "loss": 0.7794,
+      "step": 21228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.581532621204035e-06,
+      "loss": 0.9021,
+      "step": 21229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.580688482265813e-06,
+      "loss": 0.801,
+      "step": 21230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.579844353648979e-06,
+      "loss": 0.7295,
+      "step": 21231
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.579000235359675e-06,
+      "loss": 0.8797,
+      "step": 21232
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.578156127404042e-06,
+      "loss": 1.0674,
+      "step": 21233
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.577312029788217e-06,
+      "loss": 0.5623,
+      "step": 21234
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.576467942518334e-06,
+      "loss": 0.9317,
+      "step": 21235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.575623865600534e-06,
+      "loss": 0.7487,
+      "step": 21236
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.574779799040954e-06,
+      "loss": 0.6015,
+      "step": 21237
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.573935742845735e-06,
+      "loss": 0.7112,
+      "step": 21238
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.573091697021015e-06,
+      "loss": 0.6191,
+      "step": 21239
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.572247661572926e-06,
+      "loss": 0.7125,
+      "step": 21240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.571403636507615e-06,
+      "loss": 0.8825,
+      "step": 21241
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.570559621831213e-06,
+      "loss": 1.0774,
+      "step": 21242
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.569715617549863e-06,
+      "loss": 0.9045,
+      "step": 21243
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.568871623669701e-06,
+      "loss": 1.0205,
+      "step": 21244
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.568027640196862e-06,
+      "loss": 0.8174,
+      "step": 21245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567183667137481e-06,
+      "loss": 0.8615,
+      "step": 21246
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.566339704497702e-06,
+      "loss": 1.1871,
+      "step": 21247
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.56549575228366e-06,
+      "loss": 0.9234,
+      "step": 21248
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.56465181050149e-06,
+      "loss": 0.6404,
+      "step": 21249
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.563807879157336e-06,
+      "loss": 1.0363,
+      "step": 21250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.562963958257328e-06,
+      "loss": 0.7001,
+      "step": 21251
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.562120047807605e-06,
+      "loss": 0.9249,
+      "step": 21252
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.561276147814305e-06,
+      "loss": 0.6914,
+      "step": 21253
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.560432258283569e-06,
+      "loss": 0.8959,
+      "step": 21254
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.559588379221525e-06,
+      "loss": 0.8727,
+      "step": 21255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.558744510634316e-06,
+      "loss": 0.9596,
+      "step": 21256
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.557900652528075e-06,
+      "loss": 1.0706,
+      "step": 21257
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.557056804908941e-06,
+      "loss": 1.24,
+      "step": 21258
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.556212967783051e-06,
+      "loss": 1.1956,
+      "step": 21259
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55536914115654e-06,
+      "loss": 0.8616,
+      "step": 21260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.554525325035545e-06,
+      "loss": 0.8429,
+      "step": 21261
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.553681519426203e-06,
+      "loss": 0.8374,
+      "step": 21262
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.552837724334648e-06,
+      "loss": 0.8564,
+      "step": 21263
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.551993939767022e-06,
+      "loss": 0.8154,
+      "step": 21264
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.551150165729455e-06,
+      "loss": 1.3483,
+      "step": 21265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.550306402228084e-06,
+      "loss": 0.6845,
+      "step": 21266
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.549462649269044e-06,
+      "loss": 0.9315,
+      "step": 21267
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.548618906858475e-06,
+      "loss": 0.9223,
+      "step": 21268
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547775175002509e-06,
+      "loss": 1.0206,
+      "step": 21269
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546931453707285e-06,
+      "loss": 0.9978,
+      "step": 21270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546087742978935e-06,
+      "loss": 0.4906,
+      "step": 21271
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.545244042823597e-06,
+      "loss": 0.8664,
+      "step": 21272
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.544400353247405e-06,
+      "loss": 0.5483,
+      "step": 21273
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5435566742565e-06,
+      "loss": 0.8326,
+      "step": 21274
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54271300585701e-06,
+      "loss": 1.301,
+      "step": 21275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54186934805507e-06,
+      "loss": 0.9267,
+      "step": 21276
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54102570085682e-06,
+      "loss": 0.4491,
+      "step": 21277
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.540182064268394e-06,
+      "loss": 0.706,
+      "step": 21278
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.539338438295925e-06,
+      "loss": 1.1659,
+      "step": 21279
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.538494822945546e-06,
+      "loss": 0.8237,
+      "step": 21280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5376512182234e-06,
+      "loss": 0.9355,
+      "step": 21281
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.536807624135614e-06,
+      "loss": 0.8072,
+      "step": 21282
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.535964040688324e-06,
+      "loss": 0.8215,
+      "step": 21283
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.535120467887675e-06,
+      "loss": 0.6543,
+      "step": 21284
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.534276905739783e-06,
+      "loss": 1.0077,
+      "step": 21285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.533433354250797e-06,
+      "loss": 0.6976,
+      "step": 21286
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.532589813426844e-06,
+      "loss": 0.4675,
+      "step": 21287
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531746283274061e-06,
+      "loss": 0.7421,
+      "step": 21288
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.530902763798583e-06,
+      "loss": 0.7252,
+      "step": 21289
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.530059255006542e-06,
+      "loss": 1.1187,
+      "step": 21290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529215756904075e-06,
+      "loss": 0.7848,
+      "step": 21291
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.528372269497311e-06,
+      "loss": 1.0095,
+      "step": 21292
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52752879279239e-06,
+      "loss": 0.8749,
+      "step": 21293
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.526685326795445e-06,
+      "loss": 0.8704,
+      "step": 21294
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.525841871512607e-06,
+      "loss": 0.7932,
+      "step": 21295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52499842695001e-06,
+      "loss": 0.8072,
+      "step": 21296
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.524154993113787e-06,
+      "loss": 0.9311,
+      "step": 21297
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.523311570010072e-06,
+      "loss": 0.9115,
+      "step": 21298
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.522468157645001e-06,
+      "loss": 0.777,
+      "step": 21299
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.521624756024706e-06,
+      "loss": 0.5736,
+      "step": 21300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.520781365155317e-06,
+      "loss": 1.0508,
+      "step": 21301
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.519937985042974e-06,
+      "loss": 0.9607,
+      "step": 21302
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.519094615693804e-06,
+      "loss": 1.147,
+      "step": 21303
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518251257113945e-06,
+      "loss": 0.8513,
+      "step": 21304
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.517407909309527e-06,
+      "loss": 0.7973,
+      "step": 21305
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.516564572286681e-06,
+      "loss": 0.83,
+      "step": 21306
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.515721246051543e-06,
+      "loss": 0.7895,
+      "step": 21307
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.514877930610245e-06,
+      "loss": 0.7897,
+      "step": 21308
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.514034625968918e-06,
+      "loss": 0.5646,
+      "step": 21309
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.513191332133696e-06,
+      "loss": 0.7207,
+      "step": 21310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.512348049110714e-06,
+      "loss": 0.6435,
+      "step": 21311
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.511504776906103e-06,
+      "loss": 0.9156,
+      "step": 21312
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510661515525995e-06,
+      "loss": 0.7828,
+      "step": 21313
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.509818264976521e-06,
+      "loss": 0.7435,
+      "step": 21314
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.508975025263814e-06,
+      "loss": 0.7036,
+      "step": 21315
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.508131796394003e-06,
+      "loss": 0.9584,
+      "step": 21316
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.507288578373227e-06,
+      "loss": 0.9245,
+      "step": 21317
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.506445371207612e-06,
+      "loss": 1.2339,
+      "step": 21318
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.505602174903292e-06,
+      "loss": 1.0577,
+      "step": 21319
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.5047589894664e-06,
+      "loss": 0.4835,
+      "step": 21320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.503915814903066e-06,
+      "loss": 0.6778,
+      "step": 21321
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.503072651219425e-06,
+      "loss": 1.0087,
+      "step": 21322
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.502229498421602e-06,
+      "loss": 1.1334,
+      "step": 21323
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.501386356515737e-06,
+      "loss": 0.856,
+      "step": 21324
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.500543225507954e-06,
+      "loss": 0.6998,
+      "step": 21325
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.499700105404388e-06,
+      "loss": 0.6364,
+      "step": 21326
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.498856996211167e-06,
+      "loss": 0.9623,
+      "step": 21327
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.498013897934424e-06,
+      "loss": 0.9378,
+      "step": 21328
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.497170810580293e-06,
+      "loss": 0.6978,
+      "step": 21329
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.496327734154905e-06,
+      "loss": 1.0399,
+      "step": 21330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.495484668664386e-06,
+      "loss": 1.1902,
+      "step": 21331
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.494641614114869e-06,
+      "loss": 0.8369,
+      "step": 21332
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.493798570512487e-06,
+      "loss": 1.0057,
+      "step": 21333
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49295553786337e-06,
+      "loss": 0.9373,
+      "step": 21334
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.492112516173646e-06,
+      "loss": 0.7627,
+      "step": 21335
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.491269505449449e-06,
+      "loss": 0.7238,
+      "step": 21336
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.490426505696903e-06,
+      "loss": 0.8465,
+      "step": 21337
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.489583516922147e-06,
+      "loss": 0.8446,
+      "step": 21338
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.488740539131306e-06,
+      "loss": 1.0747,
+      "step": 21339
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.487897572330514e-06,
+      "loss": 0.7456,
+      "step": 21340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.487054616525893e-06,
+      "loss": 0.8635,
+      "step": 21341
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.486211671723584e-06,
+      "loss": 0.7339,
+      "step": 21342
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48536873792971e-06,
+      "loss": 0.9234,
+      "step": 21343
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.484525815150407e-06,
+      "loss": 0.9448,
+      "step": 21344
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.483682903391796e-06,
+      "loss": 0.9052,
+      "step": 21345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.482840002660011e-06,
+      "loss": 1.0712,
+      "step": 21346
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.481997112961184e-06,
+      "loss": 0.9621,
+      "step": 21347
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.481154234301441e-06,
+      "loss": 0.9953,
+      "step": 21348
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.480311366686915e-06,
+      "loss": 0.6087,
+      "step": 21349
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47946851012373e-06,
+      "loss": 0.909,
+      "step": 21350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.478625664618023e-06,
+      "loss": 0.6771,
+      "step": 21351
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.477782830175917e-06,
+      "loss": 0.7943,
+      "step": 21352
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.476940006803543e-06,
+      "loss": 0.8376,
+      "step": 21353
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.476097194507034e-06,
+      "loss": 0.8633,
+      "step": 21354
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.475254393292513e-06,
+      "loss": 0.8896,
+      "step": 21355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.474411603166109e-06,
+      "loss": 1.1741,
+      "step": 21356
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.473568824133956e-06,
+      "loss": 0.6738,
+      "step": 21357
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.472726056202178e-06,
+      "loss": 0.7585,
+      "step": 21358
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.471883299376905e-06,
+      "loss": 0.9675,
+      "step": 21359
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.471040553664269e-06,
+      "loss": 0.7843,
+      "step": 21360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.470197819070395e-06,
+      "loss": 1.0811,
+      "step": 21361
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.469355095601413e-06,
+      "loss": 0.6699,
+      "step": 21362
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.468512383263447e-06,
+      "loss": 0.7603,
+      "step": 21363
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.467669682062637e-06,
+      "loss": 0.8448,
+      "step": 21364
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.466826992005095e-06,
+      "loss": 0.8806,
+      "step": 21365
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.465984313096962e-06,
+      "loss": 0.7466,
+      "step": 21366
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46514164534436e-06,
+      "loss": 0.9923,
+      "step": 21367
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.464298988753417e-06,
+      "loss": 1.0605,
+      "step": 21368
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.463456343330264e-06,
+      "loss": 0.8155,
+      "step": 21369
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.462613709081026e-06,
+      "loss": 1.3052,
+      "step": 21370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.461771086011833e-06,
+      "loss": 0.5975,
+      "step": 21371
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46092847412881e-06,
+      "loss": 1.2245,
+      "step": 21372
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46008587343809e-06,
+      "loss": 0.7512,
+      "step": 21373
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.459243283945794e-06,
+      "loss": 0.7047,
+      "step": 21374
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.458400705658051e-06,
+      "loss": 0.8408,
+      "step": 21375
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45755813858099e-06,
+      "loss": 0.8379,
+      "step": 21376
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.456715582720737e-06,
+      "loss": 0.9266,
+      "step": 21377
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45587303808342e-06,
+      "loss": 1.1173,
+      "step": 21378
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.455030504675167e-06,
+      "loss": 1.0305,
+      "step": 21379
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.454187982502102e-06,
+      "loss": 0.7818,
+      "step": 21380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.453345471570353e-06,
+      "loss": 0.9562,
+      "step": 21381
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.452502971886049e-06,
+      "loss": 0.7806,
+      "step": 21382
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.451660483455315e-06,
+      "loss": 0.8475,
+      "step": 21383
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.450818006284281e-06,
+      "loss": 0.7972,
+      "step": 21384
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.449975540379067e-06,
+      "loss": 0.7032,
+      "step": 21385
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.449133085745801e-06,
+      "loss": 0.6596,
+      "step": 21386
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.448290642390615e-06,
+      "loss": 1.0511,
+      "step": 21387
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.44744821031963e-06,
+      "loss": 1.0406,
+      "step": 21388
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.446605789538975e-06,
+      "loss": 1.1682,
+      "step": 21389
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.445763380054773e-06,
+      "loss": 0.8841,
+      "step": 21390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.444920981873155e-06,
+      "loss": 0.8355,
+      "step": 21391
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.444078595000243e-06,
+      "loss": 0.8486,
+      "step": 21392
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.443236219442163e-06,
+      "loss": 0.9041,
+      "step": 21393
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.442393855205048e-06,
+      "loss": 1.036,
+      "step": 21394
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.441551502295012e-06,
+      "loss": 0.8865,
+      "step": 21395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.440709160718188e-06,
+      "loss": 0.8262,
+      "step": 21396
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.439866830480701e-06,
+      "loss": 0.8901,
+      "step": 21397
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.439024511588673e-06,
+      "loss": 0.9502,
+      "step": 21398
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.438182204048234e-06,
+      "loss": 0.7767,
+      "step": 21399
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.437339907865506e-06,
+      "loss": 0.7884,
+      "step": 21400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.436497623046617e-06,
+      "loss": 0.9412,
+      "step": 21401
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43565534959769e-06,
+      "loss": 1.0042,
+      "step": 21402
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.434813087524856e-06,
+      "loss": 0.8666,
+      "step": 21403
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43397083683423e-06,
+      "loss": 0.8737,
+      "step": 21404
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.433128597531943e-06,
+      "loss": 0.5922,
+      "step": 21405
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43228636962412e-06,
+      "loss": 1.0732,
+      "step": 21406
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.431444153116883e-06,
+      "loss": 1.0147,
+      "step": 21407
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.430601948016357e-06,
+      "loss": 1.1316,
+      "step": 21408
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42975975432867e-06,
+      "loss": 0.8078,
+      "step": 21409
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.428917572059945e-06,
+      "loss": 0.8071,
+      "step": 21410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.428075401216302e-06,
+      "loss": 1.0778,
+      "step": 21411
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.427233241803874e-06,
+      "loss": 0.7912,
+      "step": 21412
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42639109382878e-06,
+      "loss": 0.8158,
+      "step": 21413
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.425548957297146e-06,
+      "loss": 0.8404,
+      "step": 21414
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.424706832215092e-06,
+      "loss": 0.7701,
+      "step": 21415
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.423864718588745e-06,
+      "loss": 0.8514,
+      "step": 21416
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.423022616424227e-06,
+      "loss": 0.6426,
+      "step": 21417
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.422180525727666e-06,
+      "loss": 1.1075,
+      "step": 21418
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.421338446505182e-06,
+      "loss": 1.0253,
+      "step": 21419
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.420496378762901e-06,
+      "loss": 0.9019,
+      "step": 21420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.419654322506945e-06,
+      "loss": 0.7705,
+      "step": 21421
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41881227774344e-06,
+      "loss": 0.519,
+      "step": 21422
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.417970244478509e-06,
+      "loss": 0.8088,
+      "step": 21423
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.417128222718273e-06,
+      "loss": 0.7348,
+      "step": 21424
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.416286212468855e-06,
+      "loss": 1.1077,
+      "step": 21425
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.415444213736378e-06,
+      "loss": 0.8824,
+      "step": 21426
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.414602226526969e-06,
+      "loss": 0.9021,
+      "step": 21427
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413760250846747e-06,
+      "loss": 1.0505,
+      "step": 21428
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.412918286701836e-06,
+      "loss": 0.9249,
+      "step": 21429
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.412076334098363e-06,
+      "loss": 1.0846,
+      "step": 21430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.411234393042444e-06,
+      "loss": 0.9276,
+      "step": 21431
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410392463540207e-06,
+      "loss": 0.8178,
+      "step": 21432
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.409550545597774e-06,
+      "loss": 0.922,
+      "step": 21433
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.408708639221265e-06,
+      "loss": 1.0506,
+      "step": 21434
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.407866744416801e-06,
+      "loss": 0.7772,
+      "step": 21435
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.407024861190508e-06,
+      "loss": 1.0492,
+      "step": 21436
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.406182989548507e-06,
+      "loss": 1.1179,
+      "step": 21437
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.405341129496919e-06,
+      "loss": 0.7024,
+      "step": 21438
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.404499281041869e-06,
+      "loss": 1.1452,
+      "step": 21439
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.403657444189478e-06,
+      "loss": 0.6945,
+      "step": 21440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.402815618945868e-06,
+      "loss": 0.9396,
+      "step": 21441
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.401973805317157e-06,
+      "loss": 0.7925,
+      "step": 21442
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.401132003309475e-06,
+      "loss": 0.8391,
+      "step": 21443
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.400290212928934e-06,
+      "loss": 0.7709,
+      "step": 21444
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.399448434181661e-06,
+      "loss": 0.9475,
+      "step": 21445
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.398606667073778e-06,
+      "loss": 0.7481,
+      "step": 21446
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.397764911611404e-06,
+      "loss": 0.9453,
+      "step": 21447
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396923167800661e-06,
+      "loss": 1.4864,
+      "step": 21448
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396081435647673e-06,
+      "loss": 0.7253,
+      "step": 21449
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.395239715158558e-06,
+      "loss": 0.797,
+      "step": 21450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.394398006339435e-06,
+      "loss": 0.856,
+      "step": 21451
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.393556309196431e-06,
+      "loss": 0.7258,
+      "step": 21452
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392714623735667e-06,
+      "loss": 0.7822,
+      "step": 21453
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.391872949963258e-06,
+      "loss": 0.969,
+      "step": 21454
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.391031287885326e-06,
+      "loss": 1.0891,
+      "step": 21455
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.390189637507992e-06,
+      "loss": 0.8406,
+      "step": 21456
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.38934799883738e-06,
+      "loss": 0.8377,
+      "step": 21457
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.38850637187961e-06,
+      "loss": 0.7158,
+      "step": 21458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.387664756640797e-06,
+      "loss": 0.9651,
+      "step": 21459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.386823153127065e-06,
+      "loss": 1.016,
+      "step": 21460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.385981561344537e-06,
+      "loss": 1.0313,
+      "step": 21461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.385139981299328e-06,
+      "loss": 0.9605,
+      "step": 21462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.384298412997564e-06,
+      "loss": 0.9161,
+      "step": 21463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.383456856445359e-06,
+      "loss": 0.4731,
+      "step": 21464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.382615311648833e-06,
+      "loss": 0.9992,
+      "step": 21465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381773778614111e-06,
+      "loss": 0.7193,
+      "step": 21466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.380932257347309e-06,
+      "loss": 1.016,
+      "step": 21467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.380090747854546e-06,
+      "loss": 0.9826,
+      "step": 21468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.379249250141943e-06,
+      "loss": 0.9519,
+      "step": 21469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37840776421562e-06,
+      "loss": 0.7927,
+      "step": 21470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.377566290081698e-06,
+      "loss": 0.568,
+      "step": 21471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37672482774629e-06,
+      "loss": 0.6617,
+      "step": 21472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.375883377215527e-06,
+      "loss": 0.8297,
+      "step": 21473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.375041938495513e-06,
+      "loss": 0.4484,
+      "step": 21474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.374200511592378e-06,
+      "loss": 0.8282,
+      "step": 21475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.373359096512236e-06,
+      "loss": 0.7996,
+      "step": 21476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.372517693261207e-06,
+      "loss": 0.9213,
+      "step": 21477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.371676301845408e-06,
+      "loss": 0.7353,
+      "step": 21478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.370834922270963e-06,
+      "loss": 0.9375,
+      "step": 21479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.369993554543987e-06,
+      "loss": 0.9038,
+      "step": 21480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.369152198670597e-06,
+      "loss": 1.0518,
+      "step": 21481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.368310854656914e-06,
+      "loss": 0.9972,
+      "step": 21482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36746952250906e-06,
+      "loss": 0.7701,
+      "step": 21483
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.366628202233145e-06,
+      "loss": 0.9829,
+      "step": 21484
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.365786893835292e-06,
+      "loss": 0.9931,
+      "step": 21485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.364945597321619e-06,
+      "loss": 1.013,
+      "step": 21486
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36410431269824e-06,
+      "loss": 1.1023,
+      "step": 21487
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.363263039971277e-06,
+      "loss": 0.7572,
+      "step": 21488
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.362421779146849e-06,
+      "loss": 1.0245,
+      "step": 21489
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.361580530231068e-06,
+      "loss": 1.077,
+      "step": 21490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.360739293230058e-06,
+      "loss": 0.7792,
+      "step": 21491
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359898068149934e-06,
+      "loss": 0.8637,
+      "step": 21492
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359056854996816e-06,
+      "loss": 0.8455,
+      "step": 21493
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.358215653776817e-06,
+      "loss": 0.7633,
+      "step": 21494
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.357374464496056e-06,
+      "loss": 0.8087,
+      "step": 21495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.356533287160646e-06,
+      "loss": 0.9728,
+      "step": 21496
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.355692121776712e-06,
+      "loss": 0.9399,
+      "step": 21497
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.354850968350368e-06,
+      "loss": 0.6107,
+      "step": 21498
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.354009826887728e-06,
+      "loss": 1.06,
+      "step": 21499
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.353168697394913e-06,
+      "loss": 0.8425,
+      "step": 21500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.352327579878038e-06,
+      "loss": 0.8525,
+      "step": 21501
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35148647434322e-06,
+      "loss": 1.3354,
+      "step": 21502
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.350645380796578e-06,
+      "loss": 1.0522,
+      "step": 21503
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349804299244225e-06,
+      "loss": 0.8654,
+      "step": 21504
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.348963229692274e-06,
+      "loss": 0.7478,
+      "step": 21505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.348122172146849e-06,
+      "loss": 0.8111,
+      "step": 21506
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347281126614062e-06,
+      "loss": 0.9137,
+      "step": 21507
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.346440093100029e-06,
+      "loss": 0.8306,
+      "step": 21508
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34559907161087e-06,
+      "loss": 0.6973,
+      "step": 21509
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.344758062152696e-06,
+      "loss": 0.7767,
+      "step": 21510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.343917064731626e-06,
+      "loss": 0.7852,
+      "step": 21511
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.343076079353774e-06,
+      "loss": 0.5214,
+      "step": 21512
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.342235106025262e-06,
+      "loss": 0.7904,
+      "step": 21513
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.341394144752194e-06,
+      "loss": 0.7587,
+      "step": 21514
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.340553195540696e-06,
+      "loss": 0.656,
+      "step": 21515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.339712258396877e-06,
+      "loss": 0.8082,
+      "step": 21516
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.338871333326856e-06,
+      "loss": 1.2481,
+      "step": 21517
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.338030420336747e-06,
+      "loss": 0.6971,
+      "step": 21518
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.337189519432665e-06,
+      "loss": 0.707,
+      "step": 21519
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.336348630620725e-06,
+      "loss": 0.7586,
+      "step": 21520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.335507753907042e-06,
+      "loss": 1.0308,
+      "step": 21521
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.334666889297733e-06,
+      "loss": 0.9009,
+      "step": 21522
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.333826036798914e-06,
+      "loss": 1.2358,
+      "step": 21523
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.332985196416693e-06,
+      "loss": 0.8526,
+      "step": 21524
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.332144368157192e-06,
+      "loss": 1.113,
+      "step": 21525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.331303552026518e-06,
+      "loss": 1.0838,
+      "step": 21526
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330462748030791e-06,
+      "loss": 1.1086,
+      "step": 21527
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.329621956176126e-06,
+      "loss": 0.7747,
+      "step": 21528
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.328781176468634e-06,
+      "loss": 0.9997,
+      "step": 21529
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32794040891443e-06,
+      "loss": 0.9199,
+      "step": 21530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.327099653519632e-06,
+      "loss": 0.9863,
+      "step": 21531
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32625891029035e-06,
+      "loss": 0.7848,
+      "step": 21532
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.325418179232701e-06,
+      "loss": 0.8495,
+      "step": 21533
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.324577460352795e-06,
+      "loss": 0.7778,
+      "step": 21534
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.323736753656746e-06,
+      "loss": 0.9194,
+      "step": 21535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32289605915067e-06,
+      "loss": 0.873,
+      "step": 21536
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.322055376840684e-06,
+      "loss": 0.9031,
+      "step": 21537
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.321214706732894e-06,
+      "loss": 0.8018,
+      "step": 21538
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.320374048833416e-06,
+      "loss": 0.966,
+      "step": 21539
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.319533403148368e-06,
+      "loss": 1.2916,
+      "step": 21540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.318692769683859e-06,
+      "loss": 1.0315,
+      "step": 21541
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.317852148445998e-06,
+      "loss": 0.808,
+      "step": 21542
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.317011539440912e-06,
+      "loss": 0.7868,
+      "step": 21543
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.316170942674698e-06,
+      "loss": 0.9708,
+      "step": 21544
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.315330358153479e-06,
+      "loss": 0.9846,
+      "step": 21545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.314489785883363e-06,
+      "loss": 0.7642,
+      "step": 21546
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.313649225870467e-06,
+      "loss": 0.8531,
+      "step": 21547
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.312808678120896e-06,
+      "loss": 0.804,
+      "step": 21548
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31196814264077e-06,
+      "loss": 0.8987,
+      "step": 21549
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.311127619436202e-06,
+      "loss": 1.0528,
+      "step": 21550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.310287108513296e-06,
+      "loss": 0.7195,
+      "step": 21551
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.309446609878174e-06,
+      "loss": 0.9451,
+      "step": 21552
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.308606123536945e-06,
+      "loss": 0.8463,
+      "step": 21553
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.307765649495719e-06,
+      "loss": 0.8349,
+      "step": 21554
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.306925187760608e-06,
+      "loss": 0.9113,
+      "step": 21555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.306084738337727e-06,
+      "loss": 0.7897,
+      "step": 21556
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.305244301233182e-06,
+      "loss": 0.9171,
+      "step": 21557
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.304403876453091e-06,
+      "loss": 0.5749,
+      "step": 21558
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.303563464003563e-06,
+      "loss": 0.8517,
+      "step": 21559
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.302723063890708e-06,
+      "loss": 0.8235,
+      "step": 21560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.301882676120642e-06,
+      "loss": 0.8683,
+      "step": 21561
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.301042300699473e-06,
+      "loss": 0.8607,
+      "step": 21562
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.300201937633316e-06,
+      "loss": 0.99,
+      "step": 21563
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.299361586928276e-06,
+      "loss": 1.1431,
+      "step": 21564
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29852124859047e-06,
+      "loss": 0.8397,
+      "step": 21565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.297680922626002e-06,
+      "loss": 0.9421,
+      "step": 21566
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29684060904099e-06,
+      "loss": 1.306,
+      "step": 21567
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.296000307841543e-06,
+      "loss": 0.7544,
+      "step": 21568
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.295160019033766e-06,
+      "loss": 1.1744,
+      "step": 21569
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29431974262378e-06,
+      "loss": 0.7316,
+      "step": 21570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.293479478617688e-06,
+      "loss": 0.7499,
+      "step": 21571
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.292639227021604e-06,
+      "loss": 1.0971,
+      "step": 21572
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29179898784164e-06,
+      "loss": 1.0644,
+      "step": 21573
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2909587610839e-06,
+      "loss": 0.8605,
+      "step": 21574
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290118546754495e-06,
+      "loss": 0.6948,
+      "step": 21575
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.289278344859541e-06,
+      "loss": 1.3383,
+      "step": 21576
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.288438155405146e-06,
+      "loss": 0.9464,
+      "step": 21577
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.287597978397415e-06,
+      "loss": 0.8583,
+      "step": 21578
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.286757813842465e-06,
+      "loss": 1.0173,
+      "step": 21579
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2859176617464e-06,
+      "loss": 0.967,
+      "step": 21580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.285077522115334e-06,
+      "loss": 1.0306,
+      "step": 21581
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.284237394955372e-06,
+      "loss": 0.6936,
+      "step": 21582
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.283397280272634e-06,
+      "loss": 0.9847,
+      "step": 21583
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.282557178073214e-06,
+      "loss": 0.603,
+      "step": 21584
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28171708836323e-06,
+      "loss": 1.0221,
+      "step": 21585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280877011148793e-06,
+      "loss": 0.8534,
+      "step": 21586
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280036946436005e-06,
+      "loss": 0.7216,
+      "step": 21587
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.279196894230981e-06,
+      "loss": 0.6843,
+      "step": 21588
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.278356854539829e-06,
+      "loss": 0.9735,
+      "step": 21589
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.277516827368658e-06,
+      "loss": 0.8664,
+      "step": 21590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.276676812723574e-06,
+      "loss": 1.4618,
+      "step": 21591
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.275836810610688e-06,
+      "loss": 1.2541,
+      "step": 21592
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27499682103611e-06,
+      "loss": 0.9187,
+      "step": 21593
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274156844005945e-06,
+      "loss": 1.0868,
+      "step": 21594
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.273316879526305e-06,
+      "loss": 0.6019,
+      "step": 21595
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.272476927603294e-06,
+      "loss": 0.944,
+      "step": 21596
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.271636988243022e-06,
+      "loss": 0.8584,
+      "step": 21597
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2707970614516e-06,
+      "loss": 1.01,
+      "step": 21598
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.269957147235134e-06,
+      "loss": 1.0192,
+      "step": 21599
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.269117245599729e-06,
+      "loss": 1.0783,
+      "step": 21600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.268277356551498e-06,
+      "loss": 0.9396,
+      "step": 21601
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.267437480096546e-06,
+      "loss": 0.8319,
+      "step": 21602
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.266597616240982e-06,
+      "loss": 1.1058,
+      "step": 21603
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.265757764990912e-06,
+      "loss": 1.1322,
+      "step": 21604
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.264917926352442e-06,
+      "loss": 0.6994,
+      "step": 21605
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.264078100331682e-06,
+      "loss": 0.908,
+      "step": 21606
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26323828693474e-06,
+      "loss": 0.8307,
+      "step": 21607
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.262398486167723e-06,
+      "loss": 0.8183,
+      "step": 21608
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.261558698036734e-06,
+      "loss": 0.795,
+      "step": 21609
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.260718922547886e-06,
+      "loss": 0.7654,
+      "step": 21610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.259879159707282e-06,
+      "loss": 0.8296,
+      "step": 21611
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25903940952103e-06,
+      "loss": 0.7277,
+      "step": 21612
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.258199671995241e-06,
+      "loss": 0.9213,
+      "step": 21613
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.257359947136012e-06,
+      "loss": 0.9483,
+      "step": 21614
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.256520234949456e-06,
+      "loss": 1.0814,
+      "step": 21615
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25568053544168e-06,
+      "loss": 0.9982,
+      "step": 21616
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254840848618788e-06,
+      "loss": 1.0264,
+      "step": 21617
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254001174486885e-06,
+      "loss": 0.9466,
+      "step": 21618
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.253161513052082e-06,
+      "loss": 0.8437,
+      "step": 21619
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.252321864320481e-06,
+      "loss": 0.6276,
+      "step": 21620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25148222829819e-06,
+      "loss": 1.0564,
+      "step": 21621
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.250642604991314e-06,
+      "loss": 0.988,
+      "step": 21622
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.249802994405962e-06,
+      "loss": 1.0934,
+      "step": 21623
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.248963396548234e-06,
+      "loss": 0.8644,
+      "step": 21624
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24812381142424e-06,
+      "loss": 0.627,
+      "step": 21625
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.247284239040085e-06,
+      "loss": 1.1719,
+      "step": 21626
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246444679401871e-06,
+      "loss": 0.7842,
+      "step": 21627
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.245605132515706e-06,
+      "loss": 0.8197,
+      "step": 21628
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.244765598387696e-06,
+      "loss": 0.9959,
+      "step": 21629
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.243926077023945e-06,
+      "loss": 0.9595,
+      "step": 21630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24308656843056e-06,
+      "loss": 1.1552,
+      "step": 21631
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.242247072613645e-06,
+      "loss": 0.8426,
+      "step": 21632
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.241407589579307e-06,
+      "loss": 1.1852,
+      "step": 21633
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.240568119333645e-06,
+      "loss": 1.0232,
+      "step": 21634
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.239728661882766e-06,
+      "loss": 0.9648,
+      "step": 21635
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.238889217232777e-06,
+      "loss": 0.8201,
+      "step": 21636
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23804978538978e-06,
+      "loss": 0.5714,
+      "step": 21637
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.237210366359882e-06,
+      "loss": 0.6421,
+      "step": 21638
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.236370960149184e-06,
+      "loss": 0.8548,
+      "step": 21639
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.235531566763793e-06,
+      "loss": 0.8388,
+      "step": 21640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.234692186209815e-06,
+      "loss": 1.082,
+      "step": 21641
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.233852818493349e-06,
+      "loss": 1.0209,
+      "step": 21642
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.233013463620503e-06,
+      "loss": 0.6962,
+      "step": 21643
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23217412159738e-06,
+      "loss": 0.8325,
+      "step": 21644
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23133479243008e-06,
+      "loss": 0.62,
+      "step": 21645
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23049547612471e-06,
+      "loss": 0.9029,
+      "step": 21646
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.229656172687377e-06,
+      "loss": 0.8124,
+      "step": 21647
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.228816882124176e-06,
+      "loss": 0.6818,
+      "step": 21648
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.227977604441217e-06,
+      "loss": 0.7976,
+      "step": 21649
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.227138339644603e-06,
+      "loss": 1.0342,
+      "step": 21650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.226299087740435e-06,
+      "loss": 0.6738,
+      "step": 21651
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.225459848734817e-06,
+      "loss": 0.8708,
+      "step": 21652
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.224620622633856e-06,
+      "loss": 1.0792,
+      "step": 21653
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.223781409443645e-06,
+      "loss": 1.0664,
+      "step": 21654
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.222942209170296e-06,
+      "loss": 1.1041,
+      "step": 21655
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.222103021819908e-06,
+      "loss": 0.9834,
+      "step": 21656
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.221263847398582e-06,
+      "loss": 1.0074,
+      "step": 21657
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.220424685912424e-06,
+      "loss": 0.5705,
+      "step": 21658
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.219585537367536e-06,
+      "loss": 0.8808,
+      "step": 21659
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.218746401770021e-06,
+      "loss": 0.9511,
+      "step": 21660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217907279125978e-06,
+      "loss": 1.0044,
+      "step": 21661
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217068169441511e-06,
+      "loss": 1.0977,
+      "step": 21662
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.216229072722726e-06,
+      "loss": 0.9848,
+      "step": 21663
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.215389988975719e-06,
+      "loss": 0.7959,
+      "step": 21664
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214550918206594e-06,
+      "loss": 1.0187,
+      "step": 21665
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21371186042145e-06,
+      "loss": 0.9921,
+      "step": 21666
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.212872815626394e-06,
+      "loss": 0.7246,
+      "step": 21667
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.212033783827527e-06,
+      "loss": 1.059,
+      "step": 21668
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.211194765030948e-06,
+      "loss": 0.9353,
+      "step": 21669
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.210355759242757e-06,
+      "loss": 0.8936,
+      "step": 21670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20951676646906e-06,
+      "loss": 0.8389,
+      "step": 21671
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208677786715955e-06,
+      "loss": 1.0085,
+      "step": 21672
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.207838819989547e-06,
+      "loss": 0.9743,
+      "step": 21673
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206999866295933e-06,
+      "loss": 0.9343,
+      "step": 21674
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206160925641211e-06,
+      "loss": 0.9494,
+      "step": 21675
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20532199803149e-06,
+      "loss": 0.8826,
+      "step": 21676
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.204483083472865e-06,
+      "loss": 0.7007,
+      "step": 21677
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203644181971438e-06,
+      "loss": 0.9547,
+      "step": 21678
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.202805293533309e-06,
+      "loss": 0.9558,
+      "step": 21679
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.201966418164581e-06,
+      "loss": 0.5704,
+      "step": 21680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.201127555871354e-06,
+      "loss": 0.853,
+      "step": 21681
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.200288706659726e-06,
+      "loss": 0.9354,
+      "step": 21682
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.199449870535803e-06,
+      "loss": 0.8661,
+      "step": 21683
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.198611047505673e-06,
+      "loss": 0.7486,
+      "step": 21684
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.197772237575448e-06,
+      "loss": 0.7473,
+      "step": 21685
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.196933440751222e-06,
+      "loss": 0.9814,
+      "step": 21686
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.196094657039096e-06,
+      "loss": 1.1609,
+      "step": 21687
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195255886445168e-06,
+      "loss": 0.6173,
+      "step": 21688
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.194417128975542e-06,
+      "loss": 1.0485,
+      "step": 21689
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193578384636317e-06,
+      "loss": 0.9332,
+      "step": 21690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.192739653433586e-06,
+      "loss": 1.2128,
+      "step": 21691
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.191900935373456e-06,
+      "loss": 1.1925,
+      "step": 21692
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.191062230462025e-06,
+      "loss": 0.8762,
+      "step": 21693
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.190223538705388e-06,
+      "loss": 0.7214,
+      "step": 21694
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.189384860109646e-06,
+      "loss": 0.7321,
+      "step": 21695
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1885461946809e-06,
+      "loss": 0.828,
+      "step": 21696
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.187707542425244e-06,
+      "loss": 0.7707,
+      "step": 21697
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186868903348782e-06,
+      "loss": 0.7343,
+      "step": 21698
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186030277457612e-06,
+      "loss": 0.898,
+      "step": 21699
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.185191664757828e-06,
+      "loss": 0.9645,
+      "step": 21700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.184353065255534e-06,
+      "loss": 1.0256,
+      "step": 21701
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.183514478956827e-06,
+      "loss": 0.9744,
+      "step": 21702
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.182675905867805e-06,
+      "loss": 0.8622,
+      "step": 21703
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.181837345994564e-06,
+      "loss": 1.0369,
+      "step": 21704
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.180998799343203e-06,
+      "loss": 1.0836,
+      "step": 21705
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.180160265919821e-06,
+      "loss": 0.9794,
+      "step": 21706
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179321745730514e-06,
+      "loss": 0.7456,
+      "step": 21707
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.178483238781385e-06,
+      "loss": 0.6787,
+      "step": 21708
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.177644745078525e-06,
+      "loss": 0.8308,
+      "step": 21709
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.176806264628037e-06,
+      "loss": 0.7577,
+      "step": 21710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.175967797436016e-06,
+      "loss": 1.1808,
+      "step": 21711
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17512934350856e-06,
+      "loss": 1.4987,
+      "step": 21712
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.174290902851768e-06,
+      "loss": 0.9039,
+      "step": 21713
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.173452475471732e-06,
+      "loss": 1.2162,
+      "step": 21714
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.172614061374552e-06,
+      "loss": 0.6142,
+      "step": 21715
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171775660566327e-06,
+      "loss": 1.1095,
+      "step": 21716
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.170937273053152e-06,
+      "loss": 1.0477,
+      "step": 21717
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.170098898841124e-06,
+      "loss": 1.0533,
+      "step": 21718
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16926053793634e-06,
+      "loss": 0.9207,
+      "step": 21719
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.168422190344896e-06,
+      "loss": 0.9973,
+      "step": 21720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16758385607289e-06,
+      "loss": 0.7832,
+      "step": 21721
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166745535126416e-06,
+      "loss": 0.8178,
+      "step": 21722
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.165907227511581e-06,
+      "loss": 0.7661,
+      "step": 21723
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.165068933234462e-06,
+      "loss": 1.1202,
+      "step": 21724
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16423065230117e-06,
+      "loss": 1.1887,
+      "step": 21725
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163392384717796e-06,
+      "loss": 1.0311,
+      "step": 21726
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.162554130490433e-06,
+      "loss": 0.9647,
+      "step": 21727
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.161715889625184e-06,
+      "loss": 0.7199,
+      "step": 21728
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16087766212814e-06,
+      "loss": 0.9642,
+      "step": 21729
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.160039448005399e-06,
+      "loss": 0.8294,
+      "step": 21730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.159201247263053e-06,
+      "loss": 1.1068,
+      "step": 21731
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.158363059907202e-06,
+      "loss": 0.7938,
+      "step": 21732
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.157524885943942e-06,
+      "loss": 1.2148,
+      "step": 21733
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156686725379363e-06,
+      "loss": 0.7485,
+      "step": 21734
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.155848578219563e-06,
+      "loss": 1.2489,
+      "step": 21735
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.155010444470635e-06,
+      "loss": 0.9695,
+      "step": 21736
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.154172324138678e-06,
+      "loss": 0.8582,
+      "step": 21737
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153334217229783e-06,
+      "loss": 0.8992,
+      "step": 21738
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.15249612375005e-06,
+      "loss": 0.8132,
+      "step": 21739
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.151658043705566e-06,
+      "loss": 1.1184,
+      "step": 21740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.15081997710243e-06,
+      "loss": 0.8993,
+      "step": 21741
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14998192394674e-06,
+      "loss": 0.9117,
+      "step": 21742
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.149143884244588e-06,
+      "loss": 0.6473,
+      "step": 21743
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.148305858002066e-06,
+      "loss": 0.6772,
+      "step": 21744
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.147467845225265e-06,
+      "loss": 1.0055,
+      "step": 21745
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.146629845920287e-06,
+      "loss": 0.7484,
+      "step": 21746
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145791860093224e-06,
+      "loss": 0.8952,
+      "step": 21747
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.144953887750167e-06,
+      "loss": 0.9953,
+      "step": 21748
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.144115928897208e-06,
+      "loss": 0.9541,
+      "step": 21749
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14327798354045e-06,
+      "loss": 0.6532,
+      "step": 21750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.142440051685975e-06,
+      "loss": 0.7186,
+      "step": 21751
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.141602133339885e-06,
+      "loss": 0.7558,
+      "step": 21752
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.140764228508276e-06,
+      "loss": 0.8422,
+      "step": 21753
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.139926337197227e-06,
+      "loss": 0.865,
+      "step": 21754
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.139088459412844e-06,
+      "loss": 0.8683,
+      "step": 21755
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.138250595161218e-06,
+      "loss": 0.9597,
+      "step": 21756
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.137412744448439e-06,
+      "loss": 0.757,
+      "step": 21757
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.136574907280598e-06,
+      "loss": 0.826,
+      "step": 21758
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.135737083663796e-06,
+      "loss": 1.0876,
+      "step": 21759
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.134899273604118e-06,
+      "loss": 0.7667,
+      "step": 21760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13406147710766e-06,
+      "loss": 1.1815,
+      "step": 21761
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.133223694180516e-06,
+      "loss": 1.0586,
+      "step": 21762
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.132385924828777e-06,
+      "loss": 0.8582,
+      "step": 21763
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.131548169058535e-06,
+      "loss": 0.9354,
+      "step": 21764
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.130710426875881e-06,
+      "loss": 0.6163,
+      "step": 21765
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.129872698286908e-06,
+      "loss": 1.0452,
+      "step": 21766
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.129034983297707e-06,
+      "loss": 1.1551,
+      "step": 21767
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.128197281914372e-06,
+      "loss": 0.7476,
+      "step": 21768
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.127359594142996e-06,
+      "loss": 0.7049,
+      "step": 21769
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.126521919989665e-06,
+      "loss": 0.9395,
+      "step": 21770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.125684259460478e-06,
+      "loss": 0.5444,
+      "step": 21771
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.124846612561525e-06,
+      "loss": 0.8843,
+      "step": 21772
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.124008979298896e-06,
+      "loss": 0.8474,
+      "step": 21773
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.123171359678678e-06,
+      "loss": 0.7123,
+      "step": 21774
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.12233375370697e-06,
+      "loss": 0.7161,
+      "step": 21775
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.121496161389855e-06,
+      "loss": 0.7979,
+      "step": 21776
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.120658582733431e-06,
+      "loss": 0.9149,
+      "step": 21777
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.119821017743787e-06,
+      "loss": 0.8376,
+      "step": 21778
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11898346642701e-06,
+      "loss": 0.7979,
+      "step": 21779
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.118145928789198e-06,
+      "loss": 0.9141,
+      "step": 21780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.117308404836437e-06,
+      "loss": 0.8724,
+      "step": 21781
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.116470894574818e-06,
+      "loss": 0.8113,
+      "step": 21782
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.115633398010434e-06,
+      "loss": 0.8382,
+      "step": 21783
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.114795915149372e-06,
+      "loss": 0.8303,
+      "step": 21784
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11395844599772e-06,
+      "loss": 0.7579,
+      "step": 21785
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.113120990561574e-06,
+      "loss": 0.9628,
+      "step": 21786
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.112283548847021e-06,
+      "loss": 1.0529,
+      "step": 21787
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.111446120860152e-06,
+      "loss": 0.7816,
+      "step": 21788
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.110608706607057e-06,
+      "loss": 1.1162,
+      "step": 21789
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.109771306093825e-06,
+      "loss": 0.8227,
+      "step": 21790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.108933919326547e-06,
+      "loss": 0.995,
+      "step": 21791
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10809654631131e-06,
+      "loss": 0.8537,
+      "step": 21792
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10725918705421e-06,
+      "loss": 1.1972,
+      "step": 21793
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.106421841561324e-06,
+      "loss": 0.8607,
+      "step": 21794
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105584509838754e-06,
+      "loss": 0.8009,
+      "step": 21795
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10474719189258e-06,
+      "loss": 1.0426,
+      "step": 21796
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.103909887728894e-06,
+      "loss": 0.6912,
+      "step": 21797
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10307259735379e-06,
+      "loss": 1.3147,
+      "step": 21798
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10223532077335e-06,
+      "loss": 0.9929,
+      "step": 21799
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.101398057993667e-06,
+      "loss": 0.728,
+      "step": 21800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.100560809020826e-06,
+      "loss": 0.897,
+      "step": 21801
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09972357386092e-06,
+      "loss": 1.3332,
+      "step": 21802
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.098886352520037e-06,
+      "loss": 0.7978,
+      "step": 21803
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.098049145004262e-06,
+      "loss": 0.9183,
+      "step": 21804
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.097211951319684e-06,
+      "loss": 0.8661,
+      "step": 21805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.096374771472389e-06,
+      "loss": 0.9074,
+      "step": 21806
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.095537605468472e-06,
+      "loss": 0.827,
+      "step": 21807
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.094700453314016e-06,
+      "loss": 0.9255,
+      "step": 21808
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09386331501511e-06,
+      "loss": 0.9972,
+      "step": 21809
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093026190577839e-06,
+      "loss": 0.8799,
+      "step": 21810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.092189080008296e-06,
+      "loss": 1.0431,
+      "step": 21811
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.091351983312567e-06,
+      "loss": 0.9434,
+      "step": 21812
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09051490049674e-06,
+      "loss": 0.84,
+      "step": 21813
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.089677831566898e-06,
+      "loss": 1.1624,
+      "step": 21814
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088840776529129e-06,
+      "loss": 0.7728,
+      "step": 21815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.088003735389523e-06,
+      "loss": 0.8263,
+      "step": 21816
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.087166708154169e-06,
+      "loss": 0.7456,
+      "step": 21817
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.086329694829148e-06,
+      "loss": 0.6879,
+      "step": 21818
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.085492695420549e-06,
+      "loss": 1.1525,
+      "step": 21819
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.084655709934462e-06,
+      "loss": 1.0692,
+      "step": 21820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.083818738376972e-06,
+      "loss": 1.3013,
+      "step": 21821
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.082981780754162e-06,
+      "loss": 0.7985,
+      "step": 21822
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.082144837072128e-06,
+      "loss": 0.9568,
+      "step": 21823
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.081307907336944e-06,
+      "loss": 1.0628,
+      "step": 21824
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.080470991554703e-06,
+      "loss": 0.9824,
+      "step": 21825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07963408973149e-06,
+      "loss": 1.0404,
+      "step": 21826
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.078797201873393e-06,
+      "loss": 1.1099,
+      "step": 21827
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.077960327986492e-06,
+      "loss": 0.4722,
+      "step": 21828
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07712346807688e-06,
+      "loss": 0.7239,
+      "step": 21829
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.076286622150639e-06,
+      "loss": 0.5482,
+      "step": 21830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.075449790213853e-06,
+      "loss": 0.7636,
+      "step": 21831
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.074612972272613e-06,
+      "loss": 0.9835,
+      "step": 21832
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.073776168333002e-06,
+      "loss": 0.8751,
+      "step": 21833
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072939378401104e-06,
+      "loss": 1.0494,
+      "step": 21834
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072102602483005e-06,
+      "loss": 0.8492,
+      "step": 21835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.071265840584788e-06,
+      "loss": 0.9729,
+      "step": 21836
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07042909271254e-06,
+      "loss": 0.7684,
+      "step": 21837
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.069592358872345e-06,
+      "loss": 1.3042,
+      "step": 21838
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.068755639070292e-06,
+      "loss": 0.6632,
+      "step": 21839
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067918933312459e-06,
+      "loss": 1.005,
+      "step": 21840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.067082241604937e-06,
+      "loss": 0.9305,
+      "step": 21841
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.066245563953807e-06,
+      "loss": 0.6211,
+      "step": 21842
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.065408900365155e-06,
+      "loss": 0.7956,
+      "step": 21843
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.064572250845063e-06,
+      "loss": 0.8504,
+      "step": 21844
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.063735615399617e-06,
+      "loss": 0.7812,
+      "step": 21845
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.062898994034898e-06,
+      "loss": 0.9103,
+      "step": 21846
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.062062386756995e-06,
+      "loss": 0.9289,
+      "step": 21847
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.06122579357199e-06,
+      "loss": 0.7123,
+      "step": 21848
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.060389214485962e-06,
+      "loss": 0.6874,
+      "step": 21849
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059552649505005e-06,
+      "loss": 1.1437,
+      "step": 21850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.058716098635193e-06,
+      "loss": 0.7901,
+      "step": 21851
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.057879561882616e-06,
+      "loss": 0.788,
+      "step": 21852
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.057043039253354e-06,
+      "loss": 0.7927,
+      "step": 21853
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05620653075349e-06,
+      "loss": 0.8908,
+      "step": 21854
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.055370036389105e-06,
+      "loss": 0.8128,
+      "step": 21855
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054533556166288e-06,
+      "loss": 0.7086,
+      "step": 21856
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05369709009112e-06,
+      "loss": 0.7417,
+      "step": 21857
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.052860638169679e-06,
+      "loss": 1.0876,
+      "step": 21858
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.052024200408055e-06,
+      "loss": 1.0654,
+      "step": 21859
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.051187776812326e-06,
+      "loss": 0.743,
+      "step": 21860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.050351367388577e-06,
+      "loss": 0.7902,
+      "step": 21861
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.049514972142886e-06,
+      "loss": 1.1185,
+      "step": 21862
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.048678591081347e-06,
+      "loss": 0.7087,
+      "step": 21863
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.047842224210027e-06,
+      "loss": 1.2328,
+      "step": 21864
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.047005871535017e-06,
+      "loss": 0.745,
+      "step": 21865
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.046169533062395e-06,
+      "loss": 0.8778,
+      "step": 21866
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.045333208798249e-06,
+      "loss": 0.9979,
+      "step": 21867
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044496898748652e-06,
+      "loss": 1.0968,
+      "step": 21868
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.043660602919695e-06,
+      "loss": 0.8401,
+      "step": 21869
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.042824321317453e-06,
+      "loss": 0.9994,
+      "step": 21870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041988053948009e-06,
+      "loss": 0.6905,
+      "step": 21871
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041151800817447e-06,
+      "loss": 0.8287,
+      "step": 21872
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.04031556193185e-06,
+      "loss": 0.7512,
+      "step": 21873
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.039479337297292e-06,
+      "loss": 0.9026,
+      "step": 21874
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.038643126919858e-06,
+      "loss": 1.03,
+      "step": 21875
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03780693080563e-06,
+      "loss": 0.9516,
+      "step": 21876
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036970748960687e-06,
+      "loss": 1.1188,
+      "step": 21877
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036134581391112e-06,
+      "loss": 0.9486,
+      "step": 21878
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.035298428102983e-06,
+      "loss": 0.6866,
+      "step": 21879
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03446228910238e-06,
+      "loss": 1.0106,
+      "step": 21880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.033626164395388e-06,
+      "loss": 0.9176,
+      "step": 21881
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.032790053988086e-06,
+      "loss": 0.8934,
+      "step": 21882
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.031953957886553e-06,
+      "loss": 0.8287,
+      "step": 21883
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03111787609687e-06,
+      "loss": 0.6453,
+      "step": 21884
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.030281808625114e-06,
+      "loss": 0.8594,
+      "step": 21885
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.029445755477367e-06,
+      "loss": 1.1116,
+      "step": 21886
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02860971665971e-06,
+      "loss": 0.8175,
+      "step": 21887
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.027773692178224e-06,
+      "loss": 1.0631,
+      "step": 21888
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.026937682038983e-06,
+      "loss": 0.894,
+      "step": 21889
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.026101686248073e-06,
+      "loss": 0.7378,
+      "step": 21890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02526570481157e-06,
+      "loss": 0.8306,
+      "step": 21891
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.024429737735554e-06,
+      "loss": 0.8684,
+      "step": 21892
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.023593785026108e-06,
+      "loss": 0.9721,
+      "step": 21893
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.022757846689306e-06,
+      "loss": 0.848,
+      "step": 21894
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.021921922731224e-06,
+      "loss": 0.9839,
+      "step": 21895
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.021086013157949e-06,
+      "loss": 1.0217,
+      "step": 21896
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.020250117975555e-06,
+      "loss": 1.0983,
+      "step": 21897
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.019414237190121e-06,
+      "loss": 1.122,
+      "step": 21898
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.018578370807729e-06,
+      "loss": 1.2776,
+      "step": 21899
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.017742518834454e-06,
+      "loss": 1.0237,
+      "step": 21900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016906681276374e-06,
+      "loss": 1.1114,
+      "step": 21901
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016070858139571e-06,
+      "loss": 0.7177,
+      "step": 21902
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.015235049430125e-06,
+      "loss": 0.7988,
+      "step": 21903
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.014399255154104e-06,
+      "loss": 1.0378,
+      "step": 21904
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.013563475317595e-06,
+      "loss": 0.5616,
+      "step": 21905
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.012727709926671e-06,
+      "loss": 0.9798,
+      "step": 21906
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01189195898741e-06,
+      "loss": 1.0897,
+      "step": 21907
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.011056222505894e-06,
+      "loss": 1.0028,
+      "step": 21908
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010220500488199e-06,
+      "loss": 0.789,
+      "step": 21909
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.0093847929404e-06,
+      "loss": 0.8448,
+      "step": 21910
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.008549099868575e-06,
+      "loss": 1.0192,
+      "step": 21911
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.007713421278802e-06,
+      "loss": 0.5175,
+      "step": 21912
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006877757177165e-06,
+      "loss": 0.6665,
+      "step": 21913
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006042107569728e-06,
+      "loss": 1.1016,
+      "step": 21914
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.005206472462576e-06,
+      "loss": 0.8527,
+      "step": 21915
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.004370851861781e-06,
+      "loss": 0.971,
+      "step": 21916
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.003535245773424e-06,
+      "loss": 0.8891,
+      "step": 21917
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.002699654203582e-06,
+      "loss": 0.8204,
+      "step": 21918
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00186407715833e-06,
+      "loss": 0.5762,
+      "step": 21919
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00102851464374e-06,
+      "loss": 0.8434,
+      "step": 21920
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.000192966665899e-06,
+      "loss": 0.8111,
+      "step": 21921
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.999357433230875e-06,
+      "loss": 1.0275,
+      "step": 21922
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.998521914344745e-06,
+      "loss": 1.1727,
+      "step": 21923
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.997686410013587e-06,
+      "loss": 0.8469,
+      "step": 21924
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.996850920243473e-06,
+      "loss": 0.8633,
+      "step": 21925
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.996015445040484e-06,
+      "loss": 0.8132,
+      "step": 21926
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.995179984410692e-06,
+      "loss": 0.8177,
+      "step": 21927
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.994344538360175e-06,
+      "loss": 0.9856,
+      "step": 21928
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.993509106895004e-06,
+      "loss": 0.8155,
+      "step": 21929
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99267369002126e-06,
+      "loss": 0.7205,
+      "step": 21930
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991838287745018e-06,
+      "loss": 0.8667,
+      "step": 21931
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991002900072347e-06,
+      "loss": 0.7937,
+      "step": 21932
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99016752700933e-06,
+      "loss": 0.8903,
+      "step": 21933
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.989332168562036e-06,
+      "loss": 1.0164,
+      "step": 21934
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.988496824736542e-06,
+      "loss": 1.1426,
+      "step": 21935
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.987661495538921e-06,
+      "loss": 0.9332,
+      "step": 21936
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.986826180975251e-06,
+      "loss": 1.234,
+      "step": 21937
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.985990881051602e-06,
+      "loss": 0.9135,
+      "step": 21938
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.985155595774052e-06,
+      "loss": 0.9491,
+      "step": 21939
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.984320325148676e-06,
+      "loss": 0.8322,
+      "step": 21940
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.983485069181543e-06,
+      "loss": 1.0384,
+      "step": 21941
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.982649827878731e-06,
+      "loss": 0.8996,
+      "step": 21942
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98181460124632e-06,
+      "loss": 0.8725,
+      "step": 21943
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980979389290374e-06,
+      "loss": 0.8936,
+      "step": 21944
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980144192016967e-06,
+      "loss": 0.8828,
+      "step": 21945
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.979309009432177e-06,
+      "loss": 0.9084,
+      "step": 21946
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.978473841542076e-06,
+      "loss": 1.1437,
+      "step": 21947
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97763868835274e-06,
+      "loss": 1.1747,
+      "step": 21948
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.976803549870238e-06,
+      "loss": 0.9797,
+      "step": 21949
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.975968426100643e-06,
+      "loss": 0.8658,
+      "step": 21950
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.975133317050034e-06,
+      "loss": 1.3691,
+      "step": 21951
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97429822272448e-06,
+      "loss": 1.0497,
+      "step": 21952
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.973463143130055e-06,
+      "loss": 0.9661,
+      "step": 21953
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.972628078272831e-06,
+      "loss": 0.8213,
+      "step": 21954
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97179302815888e-06,
+      "loss": 0.953,
+      "step": 21955
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.970957992794273e-06,
+      "loss": 0.7174,
+      "step": 21956
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.970122972185086e-06,
+      "loss": 0.7602,
+      "step": 21957
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.969287966337392e-06,
+      "loss": 0.9885,
+      "step": 21958
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.968452975257258e-06,
+      "loss": 0.6658,
+      "step": 21959
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.967617998950762e-06,
+      "loss": 0.6872,
+      "step": 21960
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.966783037423974e-06,
+      "loss": 1.0101,
+      "step": 21961
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965948090682965e-06,
+      "loss": 1.0201,
+      "step": 21962
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96511315873381e-06,
+      "loss": 0.9748,
+      "step": 21963
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.964278241582574e-06,
+      "loss": 0.8893,
+      "step": 21964
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.963443339235333e-06,
+      "loss": 0.9672,
+      "step": 21965
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96260845169816e-06,
+      "loss": 0.9916,
+      "step": 21966
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.961773578977123e-06,
+      "loss": 0.6609,
+      "step": 21967
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.960938721078293e-06,
+      "loss": 0.7634,
+      "step": 21968
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.960103878007748e-06,
+      "loss": 0.8492,
+      "step": 21969
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.959269049771551e-06,
+      "loss": 0.8137,
+      "step": 21970
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.958434236375779e-06,
+      "loss": 0.97,
+      "step": 21971
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957599437826496e-06,
+      "loss": 0.8887,
+      "step": 21972
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.956764654129782e-06,
+      "loss": 0.881,
+      "step": 21973
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.955929885291698e-06,
+      "loss": 0.9062,
+      "step": 21974
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.955095131318319e-06,
+      "loss": 0.8469,
+      "step": 21975
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.954260392215716e-06,
+      "loss": 0.9192,
+      "step": 21976
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.953425667989958e-06,
+      "loss": 1.0931,
+      "step": 21977
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952590958647116e-06,
+      "loss": 0.9407,
+      "step": 21978
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.951756264193261e-06,
+      "loss": 0.8658,
+      "step": 21979
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95092158463446e-06,
+      "loss": 0.693,
+      "step": 21980
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.950086919976786e-06,
+      "loss": 1.0617,
+      "step": 21981
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.949252270226308e-06,
+      "loss": 0.8132,
+      "step": 21982
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.948417635389097e-06,
+      "loss": 0.8488,
+      "step": 21983
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.947583015471218e-06,
+      "loss": 0.6489,
+      "step": 21984
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.946748410478745e-06,
+      "loss": 0.9224,
+      "step": 21985
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.945913820417742e-06,
+      "loss": 1.0636,
+      "step": 21986
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.945079245294284e-06,
+      "loss": 1.0631,
+      "step": 21987
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944244685114439e-06,
+      "loss": 0.7418,
+      "step": 21988
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.943410139884274e-06,
+      "loss": 1.0697,
+      "step": 21989
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.942575609609857e-06,
+      "loss": 0.8753,
+      "step": 21990
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.941741094297262e-06,
+      "loss": 0.8728,
+      "step": 21991
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940906593952554e-06,
+      "loss": 0.8788,
+      "step": 21992
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940072108581803e-06,
+      "loss": 0.6272,
+      "step": 21993
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.939237638191073e-06,
+      "loss": 0.895,
+      "step": 21994
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.938403182786435e-06,
+      "loss": 0.9028,
+      "step": 21995
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93756874237396e-06,
+      "loss": 0.9698,
+      "step": 21996
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.936734316959715e-06,
+      "loss": 0.8714,
+      "step": 21997
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.935899906549767e-06,
+      "loss": 0.7305,
+      "step": 21998
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.935065511150181e-06,
+      "loss": 0.8864,
+      "step": 21999
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.934231130767031e-06,
+      "loss": 0.7569,
+      "step": 22000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.933396765406381e-06,
+      "loss": 1.1855,
+      "step": 22001
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.932562415074296e-06,
+      "loss": 1.061,
+      "step": 22002
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.931728079776855e-06,
+      "loss": 0.7412,
+      "step": 22003
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93089375952011e-06,
+      "loss": 0.5449,
+      "step": 22004
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.930059454310138e-06,
+      "loss": 0.738,
+      "step": 22005
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.929225164153003e-06,
+      "loss": 0.9697,
+      "step": 22006
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.928390889054771e-06,
+      "loss": 0.971,
+      "step": 22007
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.927556629021511e-06,
+      "loss": 0.8503,
+      "step": 22008
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92672238405929e-06,
+      "loss": 0.9644,
+      "step": 22009
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925888154174174e-06,
+      "loss": 0.757,
+      "step": 22010
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925053939372228e-06,
+      "loss": 0.5051,
+      "step": 22011
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.924219739659522e-06,
+      "loss": 0.7916,
+      "step": 22012
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.923385555042123e-06,
+      "loss": 1.1305,
+      "step": 22013
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.922551385526092e-06,
+      "loss": 1.0238,
+      "step": 22014
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.921717231117499e-06,
+      "loss": 1.0295,
+      "step": 22015
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92088309182241e-06,
+      "loss": 0.8035,
+      "step": 22016
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.920048967646887e-06,
+      "loss": 1.2046,
+      "step": 22017
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.919214858597e-06,
+      "loss": 0.8283,
+      "step": 22018
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.918380764678814e-06,
+      "loss": 0.823,
+      "step": 22019
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.917546685898393e-06,
+      "loss": 0.9851,
+      "step": 22020
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.916712622261804e-06,
+      "loss": 0.857,
+      "step": 22021
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.915878573775114e-06,
+      "loss": 0.6919,
+      "step": 22022
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.915044540444388e-06,
+      "loss": 0.8294,
+      "step": 22023
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.914210522275687e-06,
+      "loss": 0.6669,
+      "step": 22024
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91337651927508e-06,
+      "loss": 0.9657,
+      "step": 22025
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.912542531448628e-06,
+      "loss": 0.9914,
+      "step": 22026
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.9117085588024e-06,
+      "loss": 0.7679,
+      "step": 22027
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91087460134246e-06,
+      "loss": 0.9335,
+      "step": 22028
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.910040659074869e-06,
+      "loss": 0.9314,
+      "step": 22029
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.909206732005697e-06,
+      "loss": 0.981,
+      "step": 22030
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.908372820141005e-06,
+      "loss": 1.1734,
+      "step": 22031
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90753892348686e-06,
+      "loss": 0.8348,
+      "step": 22032
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.906705042049325e-06,
+      "loss": 0.9744,
+      "step": 22033
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90587117583446e-06,
+      "loss": 0.9575,
+      "step": 22034
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.905037324848334e-06,
+      "loss": 1.1263,
+      "step": 22035
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.904203489097008e-06,
+      "loss": 0.8513,
+      "step": 22036
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90336966858655e-06,
+      "loss": 0.6258,
+      "step": 22037
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.902535863323015e-06,
+      "loss": 0.7824,
+      "step": 22038
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.901702073312477e-06,
+      "loss": 0.9762,
+      "step": 22039
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.900868298560992e-06,
+      "loss": 1.0563,
+      "step": 22040
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.900034539074628e-06,
+      "loss": 0.9874,
+      "step": 22041
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.899200794859443e-06,
+      "loss": 0.8663,
+      "step": 22042
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89836706592151e-06,
+      "loss": 0.9657,
+      "step": 22043
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.897533352266878e-06,
+      "loss": 0.8278,
+      "step": 22044
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.896699653901618e-06,
+      "loss": 1.1345,
+      "step": 22045
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.895865970831794e-06,
+      "loss": 0.8975,
+      "step": 22046
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.895032303063463e-06,
+      "loss": 0.7367,
+      "step": 22047
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.894198650602694e-06,
+      "loss": 0.9045,
+      "step": 22048
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.893365013455546e-06,
+      "loss": 0.7214,
+      "step": 22049
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89253139162808e-06,
+      "loss": 1.098,
+      "step": 22050
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.891697785126358e-06,
+      "loss": 0.9922,
+      "step": 22051
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890864193956446e-06,
+      "loss": 0.6317,
+      "step": 22052
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890030618124407e-06,
+      "loss": 0.8598,
+      "step": 22053
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.889197057636297e-06,
+      "loss": 1.1262,
+      "step": 22054
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88836351249818e-06,
+      "loss": 0.7995,
+      "step": 22055
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.887529982716117e-06,
+      "loss": 0.9717,
+      "step": 22056
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.886696468296172e-06,
+      "loss": 0.7286,
+      "step": 22057
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.885862969244405e-06,
+      "loss": 0.8519,
+      "step": 22058
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.885029485566878e-06,
+      "loss": 0.6521,
+      "step": 22059
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.884196017269647e-06,
+      "loss": 0.5584,
+      "step": 22060
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.883362564358782e-06,
+      "loss": 1.0838,
+      "step": 22061
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88252912684034e-06,
+      "loss": 0.9116,
+      "step": 22062
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.881695704720381e-06,
+      "loss": 1.0649,
+      "step": 22063
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880862298004965e-06,
+      "loss": 1.0149,
+      "step": 22064
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880028906700153e-06,
+      "loss": 1.035,
+      "step": 22065
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.879195530812006e-06,
+      "loss": 0.8523,
+      "step": 22066
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878362170346584e-06,
+      "loss": 0.5844,
+      "step": 22067
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87752882530995e-06,
+      "loss": 0.8266,
+      "step": 22068
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87669549570816e-06,
+      "loss": 0.9661,
+      "step": 22069
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.875862181547278e-06,
+      "loss": 0.7702,
+      "step": 22070
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.875028882833362e-06,
+      "loss": 1.0462,
+      "step": 22071
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.87419559957247e-06,
+      "loss": 0.8621,
+      "step": 22072
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.873362331770668e-06,
+      "loss": 0.8627,
+      "step": 22073
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872529079434008e-06,
+      "loss": 1.1884,
+      "step": 22074
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.871695842568553e-06,
+      "loss": 1.018,
+      "step": 22075
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870862621180362e-06,
+      "loss": 1.0672,
+      "step": 22076
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870029415275495e-06,
+      "loss": 0.9668,
+      "step": 22077
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86919622486001e-06,
+      "loss": 1.1605,
+      "step": 22078
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.868363049939966e-06,
+      "loss": 1.0031,
+      "step": 22079
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.867529890521424e-06,
+      "loss": 1.0876,
+      "step": 22080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86669674661044e-06,
+      "loss": 0.8735,
+      "step": 22081
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.865863618213076e-06,
+      "loss": 0.7346,
+      "step": 22082
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86503050533539e-06,
+      "loss": 0.9593,
+      "step": 22083
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.864197407983438e-06,
+      "loss": 0.9519,
+      "step": 22084
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.863364326163279e-06,
+      "loss": 0.8466,
+      "step": 22085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.862531259880972e-06,
+      "loss": 1.0062,
+      "step": 22086
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.861698209142573e-06,
+      "loss": 0.7002,
+      "step": 22087
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.860865173954144e-06,
+      "loss": 0.9884,
+      "step": 22088
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.860032154321741e-06,
+      "loss": 0.8488,
+      "step": 22089
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85919915025142e-06,
+      "loss": 0.8265,
+      "step": 22090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.858366161749242e-06,
+      "loss": 0.8755,
+      "step": 22091
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.857533188821265e-06,
+      "loss": 0.7072,
+      "step": 22092
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.856700231473546e-06,
+      "loss": 0.8859,
+      "step": 22093
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.855867289712136e-06,
+      "loss": 1.186,
+      "step": 22094
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8550343635431e-06,
+      "loss": 0.9166,
+      "step": 22095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85420145297249e-06,
+      "loss": 0.9293,
+      "step": 22096
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853368558006367e-06,
+      "loss": 0.8367,
+      "step": 22097
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.852535678650785e-06,
+      "loss": 0.6731,
+      "step": 22098
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.851702814911801e-06,
+      "loss": 0.9388,
+      "step": 22099
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850869966795476e-06,
+      "loss": 0.9169,
+      "step": 22100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850037134307862e-06,
+      "loss": 1.0563,
+      "step": 22101
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.849204317455016e-06,
+      "loss": 1.1823,
+      "step": 22102
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.848371516242998e-06,
+      "loss": 0.9851,
+      "step": 22103
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.847538730677859e-06,
+      "loss": 1.0052,
+      "step": 22104
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.846705960765656e-06,
+      "loss": 0.6871,
+      "step": 22105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.845873206512448e-06,
+      "loss": 0.8204,
+      "step": 22106
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84504046792429e-06,
+      "loss": 0.8027,
+      "step": 22107
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.844207745007233e-06,
+      "loss": 0.8258,
+      "step": 22108
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84337503776734e-06,
+      "loss": 0.8548,
+      "step": 22109
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.842542346210663e-06,
+      "loss": 0.9911,
+      "step": 22110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.841709670343258e-06,
+      "loss": 0.5329,
+      "step": 22111
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840877010171177e-06,
+      "loss": 1.1055,
+      "step": 22112
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840044365700485e-06,
+      "loss": 0.9985,
+      "step": 22113
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.839211736937224e-06,
+      "loss": 1.2598,
+      "step": 22114
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.838379123887458e-06,
+      "loss": 0.9117,
+      "step": 22115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.837546526557237e-06,
+      "loss": 0.9538,
+      "step": 22116
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.836713944952617e-06,
+      "loss": 0.8372,
+      "step": 22117
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.835881379079657e-06,
+      "loss": 0.7726,
+      "step": 22118
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.835048828944405e-06,
+      "loss": 0.7885,
+      "step": 22119
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83421629455292e-06,
+      "loss": 0.7676,
+      "step": 22120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.833383775911252e-06,
+      "loss": 0.7563,
+      "step": 22121
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.832551273025461e-06,
+      "loss": 0.9394,
+      "step": 22122
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8317187859016e-06,
+      "loss": 0.9244,
+      "step": 22123
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.830886314545715e-06,
+      "loss": 0.7811,
+      "step": 22124
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83005385896387e-06,
+      "loss": 1.0908,
+      "step": 22125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.82922141916211e-06,
+      "loss": 0.9553,
+      "step": 22126
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.828388995146495e-06,
+      "loss": 0.796,
+      "step": 22127
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.827556586923077e-06,
+      "loss": 0.8007,
+      "step": 22128
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.826724194497909e-06,
+      "loss": 0.9438,
+      "step": 22129
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825891817877041e-06,
+      "loss": 0.7815,
+      "step": 22130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825059457066532e-06,
+      "loss": 0.8345,
+      "step": 22131
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.824227112072433e-06,
+      "loss": 0.7905,
+      "step": 22132
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.823394782900797e-06,
+      "loss": 0.8367,
+      "step": 22133
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.822562469557673e-06,
+      "loss": 1.0973,
+      "step": 22134
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.821730172049116e-06,
+      "loss": 0.7517,
+      "step": 22135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.82089789038118e-06,
+      "loss": 0.7495,
+      "step": 22136
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.820065624559916e-06,
+      "loss": 0.9507,
+      "step": 22137
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.819233374591378e-06,
+      "loss": 0.6709,
+      "step": 22138
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.818401140481614e-06,
+      "loss": 0.7795,
+      "step": 22139
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817568922236683e-06,
+      "loss": 0.937,
+      "step": 22140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.816736719862632e-06,
+      "loss": 0.7349,
+      "step": 22141
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.815904533365512e-06,
+      "loss": 1.016,
+      "step": 22142
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.815072362751383e-06,
+      "loss": 1.0642,
+      "step": 22143
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814240208026284e-06,
+      "loss": 0.9047,
+      "step": 22144
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.813408069196275e-06,
+      "loss": 0.5429,
+      "step": 22145
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.812575946267406e-06,
+      "loss": 1.0528,
+      "step": 22146
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.811743839245727e-06,
+      "loss": 0.9241,
+      "step": 22147
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810911748137289e-06,
+      "loss": 0.9343,
+      "step": 22148
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810079672948145e-06,
+      "loss": 0.8511,
+      "step": 22149
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.809247613684344e-06,
+      "loss": 1.073,
+      "step": 22150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.808415570351936e-06,
+      "loss": 0.9514,
+      "step": 22151
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.807583542956976e-06,
+      "loss": 0.6906,
+      "step": 22152
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.806751531505514e-06,
+      "loss": 0.9374,
+      "step": 22153
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.805919536003597e-06,
+      "loss": 1.0374,
+      "step": 22154
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.805087556457275e-06,
+      "loss": 0.9296,
+      "step": 22155
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.804255592872601e-06,
+      "loss": 1.1437,
+      "step": 22156
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.803423645255622e-06,
+      "loss": 0.8424,
+      "step": 22157
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.802591713612392e-06,
+      "loss": 0.9735,
+      "step": 22158
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.801759797948958e-06,
+      "loss": 0.8635,
+      "step": 22159
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80092789827137e-06,
+      "loss": 0.8126,
+      "step": 22160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.800096014585681e-06,
+      "loss": 0.9402,
+      "step": 22161
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.799264146897939e-06,
+      "loss": 0.9625,
+      "step": 22162
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.798432295214193e-06,
+      "loss": 0.9215,
+      "step": 22163
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.797600459540491e-06,
+      "loss": 0.6542,
+      "step": 22164
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.796768639882882e-06,
+      "loss": 0.7182,
+      "step": 22165
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.795936836247414e-06,
+      "loss": 0.8957,
+      "step": 22166
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79510504864014e-06,
+      "loss": 0.874,
+      "step": 22167
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.794273277067107e-06,
+      "loss": 1.1168,
+      "step": 22168
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.793441521534362e-06,
+      "loss": 0.8027,
+      "step": 22169
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.792609782047958e-06,
+      "loss": 0.823,
+      "step": 22170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791778058613938e-06,
+      "loss": 0.9334,
+      "step": 22171
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.790946351238355e-06,
+      "loss": 0.8548,
+      "step": 22172
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.790114659927259e-06,
+      "loss": 0.6536,
+      "step": 22173
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.789282984686692e-06,
+      "loss": 0.9832,
+      "step": 22174
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.788451325522701e-06,
+      "loss": 0.92,
+      "step": 22175
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78761968244134e-06,
+      "loss": 0.8847,
+      "step": 22176
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.786788055448656e-06,
+      "loss": 0.9398,
+      "step": 22177
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785956444550691e-06,
+      "loss": 1.1462,
+      "step": 22178
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785124849753499e-06,
+      "loss": 1.0229,
+      "step": 22179
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.784293271063125e-06,
+      "loss": 0.8936,
+      "step": 22180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.783461708485617e-06,
+      "loss": 1.0441,
+      "step": 22181
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78263016202702e-06,
+      "loss": 0.8558,
+      "step": 22182
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.781798631693388e-06,
+      "loss": 0.8028,
+      "step": 22183
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.780967117490756e-06,
+      "loss": 0.7199,
+      "step": 22184
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78013561942518e-06,
+      "loss": 0.6777,
+      "step": 22185
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.779304137502705e-06,
+      "loss": 0.6697,
+      "step": 22186
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.778472671729374e-06,
+      "loss": 1.1248,
+      "step": 22187
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77764122211124e-06,
+      "loss": 1.0636,
+      "step": 22188
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.776809788654343e-06,
+      "loss": 0.9812,
+      "step": 22189
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.775978371364734e-06,
+      "loss": 1.0965,
+      "step": 22190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.775146970248455e-06,
+      "loss": 0.9256,
+      "step": 22191
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.774315585311558e-06,
+      "loss": 0.9451,
+      "step": 22192
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.773484216560086e-06,
+      "loss": 0.8868,
+      "step": 22193
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77265286400008e-06,
+      "loss": 0.8573,
+      "step": 22194
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.771821527637592e-06,
+      "loss": 0.9573,
+      "step": 22195
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770990207478662e-06,
+      "loss": 0.9982,
+      "step": 22196
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77015890352934e-06,
+      "loss": 0.7019,
+      "step": 22197
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.769327615795672e-06,
+      "loss": 1.0966,
+      "step": 22198
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.768496344283703e-06,
+      "loss": 0.8063,
+      "step": 22199
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76766508899947e-06,
+      "loss": 0.9793,
+      "step": 22200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76683384994903e-06,
+      "loss": 0.8362,
+      "step": 22201
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76600262713842e-06,
+      "loss": 0.906,
+      "step": 22202
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76517142057369e-06,
+      "loss": 1.0709,
+      "step": 22203
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.76434023026088e-06,
+      "loss": 0.7347,
+      "step": 22204
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.763509056206032e-06,
+      "loss": 1.0599,
+      "step": 22205
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.762677898415199e-06,
+      "loss": 1.0126,
+      "step": 22206
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.761846756894418e-06,
+      "loss": 0.7992,
+      "step": 22207
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.761015631649738e-06,
+      "loss": 0.7658,
+      "step": 22208
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.760184522687198e-06,
+      "loss": 0.9839,
+      "step": 22209
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.759353430012847e-06,
+      "loss": 0.8514,
+      "step": 22210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.758522353632727e-06,
+      "loss": 0.567,
+      "step": 22211
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75769129355288e-06,
+      "loss": 0.8143,
+      "step": 22212
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.756860249779358e-06,
+      "loss": 0.819,
+      "step": 22213
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.756029222318189e-06,
+      "loss": 1.0472,
+      "step": 22214
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.755198211175428e-06,
+      "loss": 0.9159,
+      "step": 22215
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.754367216357114e-06,
+      "loss": 0.9111,
+      "step": 22216
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.753536237869291e-06,
+      "loss": 0.7775,
+      "step": 22217
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.752705275718e-06,
+      "loss": 1.1014,
+      "step": 22218
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.751874329909289e-06,
+      "loss": 0.8745,
+      "step": 22219
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.751043400449197e-06,
+      "loss": 0.866,
+      "step": 22220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750212487343764e-06,
+      "loss": 0.5996,
+      "step": 22221
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.749381590599037e-06,
+      "loss": 0.7241,
+      "step": 22222
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.74855071022106e-06,
+      "loss": 0.6247,
+      "step": 22223
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747719846215871e-06,
+      "loss": 0.9384,
+      "step": 22224
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.746888998589514e-06,
+      "loss": 1.1928,
+      "step": 22225
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.746058167348029e-06,
+      "loss": 0.9714,
+      "step": 22226
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.745227352497457e-06,
+      "loss": 1.1581,
+      "step": 22227
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.744396554043845e-06,
+      "loss": 1.0228,
+      "step": 22228
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.74356577199323e-06,
+      "loss": 0.8859,
+      "step": 22229
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742735006351656e-06,
+      "loss": 1.161,
+      "step": 22230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.741904257125164e-06,
+      "loss": 1.0832,
+      "step": 22231
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.741073524319795e-06,
+      "loss": 0.6802,
+      "step": 22232
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.740242807941593e-06,
+      "loss": 0.9726,
+      "step": 22233
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.739412107996594e-06,
+      "loss": 0.9613,
+      "step": 22234
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.738581424490841e-06,
+      "loss": 0.8458,
+      "step": 22235
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.737750757430374e-06,
+      "loss": 0.4803,
+      "step": 22236
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.736920106821236e-06,
+      "loss": 0.975,
+      "step": 22237
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.736089472669465e-06,
+      "loss": 1.1083,
+      "step": 22238
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.735258854981102e-06,
+      "loss": 1.085,
+      "step": 22239
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.734428253762189e-06,
+      "loss": 0.8528,
+      "step": 22240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.733597669018767e-06,
+      "loss": 0.9612,
+      "step": 22241
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.732767100756874e-06,
+      "loss": 0.9754,
+      "step": 22242
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731936548982551e-06,
+      "loss": 0.8964,
+      "step": 22243
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731106013701836e-06,
+      "loss": 1.1251,
+      "step": 22244
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73027549492077e-06,
+      "loss": 0.9812,
+      "step": 22245
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.729444992645391e-06,
+      "loss": 1.0976,
+      "step": 22246
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.728614506881743e-06,
+      "loss": 0.907,
+      "step": 22247
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72778403763586e-06,
+      "loss": 0.8072,
+      "step": 22248
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.726953584913786e-06,
+      "loss": 1.0125,
+      "step": 22249
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.726123148721557e-06,
+      "loss": 0.9848,
+      "step": 22250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.725292729065214e-06,
+      "loss": 1.0497,
+      "step": 22251
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724462325950792e-06,
+      "loss": 0.8198,
+      "step": 22252
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.723631939384338e-06,
+      "loss": 0.9447,
+      "step": 22253
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72280156937188e-06,
+      "loss": 0.8225,
+      "step": 22254
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721971215919465e-06,
+      "loss": 0.8835,
+      "step": 22255
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721140879033128e-06,
+      "loss": 0.8973,
+      "step": 22256
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.720310558718904e-06,
+      "loss": 0.9901,
+      "step": 22257
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.719480254982838e-06,
+      "loss": 1.3023,
+      "step": 22258
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718649967830965e-06,
+      "loss": 1.1804,
+      "step": 22259
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.717819697269322e-06,
+      "loss": 0.6289,
+      "step": 22260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.716989443303947e-06,
+      "loss": 1.0771,
+      "step": 22261
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71615920594088e-06,
+      "loss": 0.7411,
+      "step": 22262
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.715328985186158e-06,
+      "loss": 0.5686,
+      "step": 22263
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.714498781045814e-06,
+      "loss": 0.9703,
+      "step": 22264
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71366859352589e-06,
+      "loss": 0.7676,
+      "step": 22265
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.712838422632421e-06,
+      "loss": 0.7094,
+      "step": 22266
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.712008268371444e-06,
+      "loss": 0.8221,
+      "step": 22267
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.711178130749e-06,
+      "loss": 0.8553,
+      "step": 22268
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.71034800977112e-06,
+      "loss": 0.7541,
+      "step": 22269
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.709517905443844e-06,
+      "loss": 1.0197,
+      "step": 22270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.708687817773211e-06,
+      "loss": 1.0043,
+      "step": 22271
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707857746765253e-06,
+      "loss": 0.7743,
+      "step": 22272
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707027692426009e-06,
+      "loss": 1.0388,
+      "step": 22273
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.706197654761514e-06,
+      "loss": 0.9491,
+      "step": 22274
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7053676337778e-06,
+      "loss": 1.0134,
+      "step": 22275
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.704537629480912e-06,
+      "loss": 0.6123,
+      "step": 22276
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.703707641876881e-06,
+      "loss": 0.9034,
+      "step": 22277
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.702877670971743e-06,
+      "loss": 0.7717,
+      "step": 22278
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.70204771677153e-06,
+      "loss": 0.8563,
+      "step": 22279
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.701217779282283e-06,
+      "loss": 0.6789,
+      "step": 22280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.700387858510038e-06,
+      "loss": 0.7312,
+      "step": 22281
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.699557954460824e-06,
+      "loss": 0.6587,
+      "step": 22282
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.698728067140688e-06,
+      "loss": 0.6708,
+      "step": 22283
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.697898196555648e-06,
+      "loss": 0.8042,
+      "step": 22284
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.697068342711752e-06,
+      "loss": 0.9329,
+      "step": 22285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69623850561503e-06,
+      "loss": 0.6022,
+      "step": 22286
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.695408685271517e-06,
+      "loss": 0.8621,
+      "step": 22287
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694578881687248e-06,
+      "loss": 0.7766,
+      "step": 22288
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.693749094868259e-06,
+      "loss": 0.8434,
+      "step": 22289
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69291932482058e-06,
+      "loss": 0.6298,
+      "step": 22290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.692089571550249e-06,
+      "loss": 0.802,
+      "step": 22291
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.691259835063298e-06,
+      "loss": 1.0844,
+      "step": 22292
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.690430115365766e-06,
+      "loss": 1.0969,
+      "step": 22293
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68960041246368e-06,
+      "loss": 0.7497,
+      "step": 22294
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.688770726363078e-06,
+      "loss": 0.8618,
+      "step": 22295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68794105706999e-06,
+      "loss": 0.7672,
+      "step": 22296
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.687111404590449e-06,
+      "loss": 0.8563,
+      "step": 22297
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686281768930493e-06,
+      "loss": 0.6447,
+      "step": 22298
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.685452150096154e-06,
+      "loss": 0.7908,
+      "step": 22299
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.684622548093461e-06,
+      "loss": 0.902,
+      "step": 22300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.683792962928453e-06,
+      "loss": 0.8965,
+      "step": 22301
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.682963394607159e-06,
+      "loss": 0.8122,
+      "step": 22302
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.682133843135616e-06,
+      "loss": 0.5269,
+      "step": 22303
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.681304308519848e-06,
+      "loss": 1.1253,
+      "step": 22304
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.680474790765895e-06,
+      "loss": 0.6817,
+      "step": 22305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679645289879784e-06,
+      "loss": 0.991,
+      "step": 22306
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.678815805867552e-06,
+      "loss": 0.9666,
+      "step": 22307
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.67798633873523e-06,
+      "loss": 0.8865,
+      "step": 22308
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.677156888488847e-06,
+      "loss": 0.8132,
+      "step": 22309
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.676327455134438e-06,
+      "loss": 0.8667,
+      "step": 22310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.675498038678036e-06,
+      "loss": 0.7516,
+      "step": 22311
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.674668639125668e-06,
+      "loss": 0.9669,
+      "step": 22312
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.673839256483371e-06,
+      "loss": 1.3869,
+      "step": 22313
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.67300989075717e-06,
+      "loss": 0.6109,
+      "step": 22314
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.672180541953099e-06,
+      "loss": 0.9198,
+      "step": 22315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.671351210077192e-06,
+      "loss": 0.9034,
+      "step": 22316
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.670521895135477e-06,
+      "loss": 1.0278,
+      "step": 22317
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.669692597133983e-06,
+      "loss": 1.0363,
+      "step": 22318
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.668863316078744e-06,
+      "loss": 0.8935,
+      "step": 22319
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.668034051975793e-06,
+      "loss": 1.0506,
+      "step": 22320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.667204804831155e-06,
+      "loss": 0.9138,
+      "step": 22321
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66637557465086e-06,
+      "loss": 0.7234,
+      "step": 22322
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66554636144095e-06,
+      "loss": 1.0056,
+      "step": 22323
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664717165207438e-06,
+      "loss": 0.9414,
+      "step": 22324
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.663887985956365e-06,
+      "loss": 0.7779,
+      "step": 22325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.663058823693757e-06,
+      "loss": 1.0475,
+      "step": 22326
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.662229678425642e-06,
+      "loss": 0.8022,
+      "step": 22327
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.661400550158056e-06,
+      "loss": 0.8464,
+      "step": 22328
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.660571438897025e-06,
+      "loss": 1.0227,
+      "step": 22329
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.659742344648577e-06,
+      "loss": 0.7771,
+      "step": 22330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.658913267418741e-06,
+      "loss": 1.2058,
+      "step": 22331
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.658084207213554e-06,
+      "loss": 1.1326,
+      "step": 22332
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.657255164039034e-06,
+      "loss": 0.45,
+      "step": 22333
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.656426137901214e-06,
+      "loss": 1.0764,
+      "step": 22334
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.655597128806125e-06,
+      "loss": 1.1117,
+      "step": 22335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65476813675979e-06,
+      "loss": 1.1646,
+      "step": 22336
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653939161768248e-06,
+      "loss": 0.6933,
+      "step": 22337
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653110203837517e-06,
+      "loss": 0.705,
+      "step": 22338
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.652281262973631e-06,
+      "loss": 0.8848,
+      "step": 22339
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.651452339182614e-06,
+      "loss": 0.9123,
+      "step": 22340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.650623432470498e-06,
+      "loss": 0.7969,
+      "step": 22341
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.64979454284331e-06,
+      "loss": 0.7875,
+      "step": 22342
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648965670307078e-06,
+      "loss": 0.402,
+      "step": 22343
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648136814867827e-06,
+      "loss": 1.1401,
+      "step": 22344
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.647307976531584e-06,
+      "loss": 0.894,
+      "step": 22345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.64647915530438e-06,
+      "loss": 0.7708,
+      "step": 22346
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645650351192244e-06,
+      "loss": 0.7101,
+      "step": 22347
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.644821564201197e-06,
+      "loss": 0.9133,
+      "step": 22348
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643992794337269e-06,
+      "loss": 0.9154,
+      "step": 22349
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643164041606489e-06,
+      "loss": 0.7978,
+      "step": 22350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.64233530601488e-06,
+      "loss": 0.5646,
+      "step": 22351
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.641506587568474e-06,
+      "loss": 1.2351,
+      "step": 22352
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.640677886273292e-06,
+      "loss": 1.0323,
+      "step": 22353
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639849202135358e-06,
+      "loss": 0.8806,
+      "step": 22354
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639020535160709e-06,
+      "loss": 0.8297,
+      "step": 22355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.638191885355362e-06,
+      "loss": 0.906,
+      "step": 22356
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.637363252725346e-06,
+      "loss": 0.6176,
+      "step": 22357
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.636534637276685e-06,
+      "loss": 0.5755,
+      "step": 22358
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.635706039015408e-06,
+      "loss": 0.9363,
+      "step": 22359
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63487745794754e-06,
+      "loss": 1.0506,
+      "step": 22360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.634048894079105e-06,
+      "loss": 0.8929,
+      "step": 22361
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.633220347416133e-06,
+      "loss": 1.0029,
+      "step": 22362
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63239181796464e-06,
+      "loss": 0.815,
+      "step": 22363
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63156330573066e-06,
+      "loss": 0.7775,
+      "step": 22364
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.630734810720212e-06,
+      "loss": 1.0901,
+      "step": 22365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.629906332939327e-06,
+      "loss": 0.9016,
+      "step": 22366
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.629077872394022e-06,
+      "loss": 0.9673,
+      "step": 22367
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.628249429090329e-06,
+      "loss": 0.9986,
+      "step": 22368
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6274210030342685e-06,
+      "loss": 1.109,
+      "step": 22369
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.626592594231864e-06,
+      "loss": 0.8742,
+      "step": 22370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.625764202689145e-06,
+      "loss": 0.8371,
+      "step": 22371
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.624935828412134e-06,
+      "loss": 0.6443,
+      "step": 22372
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6241074714068495e-06,
+      "loss": 1.0136,
+      "step": 22373
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.623279131679322e-06,
+      "loss": 0.7428,
+      "step": 22374
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.622450809235571e-06,
+      "loss": 0.7652,
+      "step": 22375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.62162250408162e-06,
+      "loss": 0.6589,
+      "step": 22376
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.620794216223497e-06,
+      "loss": 0.952,
+      "step": 22377
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.619965945667222e-06,
+      "loss": 0.8405,
+      "step": 22378
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.619137692418818e-06,
+      "loss": 0.6901,
+      "step": 22379
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.618309456484309e-06,
+      "loss": 1.0926,
+      "step": 22380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.617481237869719e-06,
+      "loss": 1.0239,
+      "step": 22381
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.616653036581072e-06,
+      "loss": 0.8956,
+      "step": 22382
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.615824852624387e-06,
+      "loss": 0.8434,
+      "step": 22383
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.614996686005689e-06,
+      "loss": 1.0633,
+      "step": 22384
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.614168536730997e-06,
+      "loss": 0.7953,
+      "step": 22385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.613340404806339e-06,
+      "loss": 0.832,
+      "step": 22386
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.612512290237733e-06,
+      "loss": 0.9649,
+      "step": 22387
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.611684193031205e-06,
+      "loss": 1.0041,
+      "step": 22388
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.610856113192772e-06,
+      "loss": 1.1057,
+      "step": 22389
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.610028050728459e-06,
+      "loss": 1.0752,
+      "step": 22390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.60920000564429e-06,
+      "loss": 0.9868,
+      "step": 22391
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.608371977946285e-06,
+      "loss": 0.5805,
+      "step": 22392
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.607543967640463e-06,
+      "loss": 0.7458,
+      "step": 22393
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.606715974732844e-06,
+      "loss": 0.7325,
+      "step": 22394
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605887999229454e-06,
+      "loss": 0.8701,
+      "step": 22395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605060041136314e-06,
+      "loss": 1.1531,
+      "step": 22396
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6042321004594425e-06,
+      "loss": 1.0904,
+      "step": 22397
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6034041772048584e-06,
+      "loss": 0.812,
+      "step": 22398
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.602576271378588e-06,
+      "loss": 0.7631,
+      "step": 22399
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.601748382986649e-06,
+      "loss": 0.9666,
+      "step": 22400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600920512035059e-06,
+      "loss": 0.9162,
+      "step": 22401
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600092658529849e-06,
+      "loss": 0.9671,
+      "step": 22402
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.599264822477026e-06,
+      "loss": 0.8917,
+      "step": 22403
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.598437003882617e-06,
+      "loss": 1.2177,
+      "step": 22404
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5976092027526405e-06,
+      "loss": 0.7751,
+      "step": 22405
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.596781419093118e-06,
+      "loss": 0.7588,
+      "step": 22406
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.595953652910064e-06,
+      "loss": 0.6187,
+      "step": 22407
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5951259042095046e-06,
+      "loss": 0.6744,
+      "step": 22408
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5942981729974575e-06,
+      "loss": 1.0615,
+      "step": 22409
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.593470459279939e-06,
+      "loss": 0.8802,
+      "step": 22410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.592642763062972e-06,
+      "loss": 0.8676,
+      "step": 22411
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591815084352577e-06,
+      "loss": 0.9195,
+      "step": 22412
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.590987423154768e-06,
+      "loss": 0.6558,
+      "step": 22413
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.590159779475565e-06,
+      "loss": 0.8527,
+      "step": 22414
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589332153320988e-06,
+      "loss": 1.2331,
+      "step": 22415
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.588504544697053e-06,
+      "loss": 0.8502,
+      "step": 22416
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.587676953609784e-06,
+      "loss": 0.9504,
+      "step": 22417
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.586849380065195e-06,
+      "loss": 0.6216,
+      "step": 22418
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.586021824069303e-06,
+      "loss": 1.1196,
+      "step": 22419
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.585194285628131e-06,
+      "loss": 1.1607,
+      "step": 22420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5843667647476945e-06,
+      "loss": 1.0076,
+      "step": 22421
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.583539261434013e-06,
+      "loss": 0.9939,
+      "step": 22422
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582711775693101e-06,
+      "loss": 0.6326,
+      "step": 22423
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.581884307530976e-06,
+      "loss": 0.7621,
+      "step": 22424
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.581056856953656e-06,
+      "loss": 1.0209,
+      "step": 22425
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.58022942396716e-06,
+      "loss": 0.7837,
+      "step": 22426
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.579402008577506e-06,
+      "loss": 0.9893,
+      "step": 22427
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.578574610790706e-06,
+      "loss": 1.0005,
+      "step": 22428
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.577747230612782e-06,
+      "loss": 1.0177,
+      "step": 22429
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.57691986804975e-06,
+      "loss": 0.8976,
+      "step": 22430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.576092523107625e-06,
+      "loss": 0.7715,
+      "step": 22431
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.575265195792428e-06,
+      "loss": 0.8626,
+      "step": 22432
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.574437886110169e-06,
+      "loss": 0.8751,
+      "step": 22433
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.573610594066865e-06,
+      "loss": 1.1663,
+      "step": 22434
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.572783319668535e-06,
+      "loss": 0.7607,
+      "step": 22435
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5719560629211955e-06,
+      "loss": 0.8743,
+      "step": 22436
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.571128823830859e-06,
+      "loss": 0.8249,
+      "step": 22437
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.570301602403545e-06,
+      "loss": 0.7877,
+      "step": 22438
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.569474398645268e-06,
+      "loss": 0.8417,
+      "step": 22439
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568647212562043e-06,
+      "loss": 0.9419,
+      "step": 22440
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5678200441598835e-06,
+      "loss": 0.4727,
+      "step": 22441
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5669928934448135e-06,
+      "loss": 0.8337,
+      "step": 22442
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.566165760422834e-06,
+      "loss": 0.9428,
+      "step": 22443
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.565338645099969e-06,
+      "loss": 0.9753,
+      "step": 22444
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.564511547482233e-06,
+      "loss": 0.6222,
+      "step": 22445
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5636844675756374e-06,
+      "loss": 0.953,
+      "step": 22446
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5628574053862e-06,
+      "loss": 0.8488,
+      "step": 22447
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.562030360919934e-06,
+      "loss": 0.8237,
+      "step": 22448
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.561203334182856e-06,
+      "loss": 0.7629,
+      "step": 22449
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.560376325180974e-06,
+      "loss": 0.9928,
+      "step": 22450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559549333920309e-06,
+      "loss": 0.6859,
+      "step": 22451
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.558722360406875e-06,
+      "loss": 0.6215,
+      "step": 22452
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.55789540464668e-06,
+      "loss": 0.5112,
+      "step": 22453
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.557068466645741e-06,
+      "loss": 0.9669,
+      "step": 22454
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.55624154641007e-06,
+      "loss": 0.9339,
+      "step": 22455
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.555414643945683e-06,
+      "loss": 0.8503,
+      "step": 22456
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.554587759258594e-06,
+      "loss": 0.9118,
+      "step": 22457
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.553760892354812e-06,
+      "loss": 0.886,
+      "step": 22458
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.552934043240352e-06,
+      "loss": 1.1421,
+      "step": 22459
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.552107211921229e-06,
+      "loss": 0.9223,
+      "step": 22460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.551280398403453e-06,
+      "loss": 0.7523,
+      "step": 22461
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.550453602693042e-06,
+      "loss": 0.7329,
+      "step": 22462
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.549626824796001e-06,
+      "loss": 0.7621,
+      "step": 22463
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.548800064718345e-06,
+      "loss": 0.6738,
+      "step": 22464
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5479733224660875e-06,
+      "loss": 0.956,
+      "step": 22465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.547146598045241e-06,
+      "loss": 0.7678,
+      "step": 22466
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.546319891461817e-06,
+      "loss": 1.1297,
+      "step": 22467
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.545493202721825e-06,
+      "loss": 0.811,
+      "step": 22468
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.544666531831282e-06,
+      "loss": 0.8453,
+      "step": 22469
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.543839878796195e-06,
+      "loss": 0.756,
+      "step": 22470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5430132436225755e-06,
+      "loss": 0.6608,
+      "step": 22471
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.542186626316443e-06,
+      "loss": 0.7235,
+      "step": 22472
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5413600268837975e-06,
+      "loss": 0.8534,
+      "step": 22473
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540533445330655e-06,
+      "loss": 0.908,
+      "step": 22474
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.539706881663027e-06,
+      "loss": 0.7921,
+      "step": 22475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.538880335886924e-06,
+      "loss": 0.862,
+      "step": 22476
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.538053808008354e-06,
+      "loss": 0.6686,
+      "step": 22477
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.537227298033333e-06,
+      "loss": 0.9187,
+      "step": 22478
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.536400805967867e-06,
+      "loss": 0.7461,
+      "step": 22479
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5355743318179675e-06,
+      "loss": 1.0396,
+      "step": 22480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.534747875589646e-06,
+      "loss": 0.9356,
+      "step": 22481
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.533921437288915e-06,
+      "loss": 1.0086,
+      "step": 22482
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.533095016921779e-06,
+      "loss": 0.7141,
+      "step": 22483
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532268614494249e-06,
+      "loss": 0.9253,
+      "step": 22484
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.531442230012336e-06,
+      "loss": 1.0647,
+      "step": 22485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.530615863482047e-06,
+      "loss": 0.9171,
+      "step": 22486
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.529789514909396e-06,
+      "loss": 0.9728,
+      "step": 22487
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.52896318430039e-06,
+      "loss": 0.9232,
+      "step": 22488
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.528136871661036e-06,
+      "loss": 0.974,
+      "step": 22489
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.527310576997347e-06,
+      "loss": 1.0095,
+      "step": 22490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.52648430031533e-06,
+      "loss": 0.5946,
+      "step": 22491
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.525658041620996e-06,
+      "loss": 0.7095,
+      "step": 22492
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.52483180092035e-06,
+      "loss": 0.9863,
+      "step": 22493
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.524005578219402e-06,
+      "loss": 1.1353,
+      "step": 22494
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.523179373524157e-06,
+      "loss": 0.9128,
+      "step": 22495
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.52235318684063e-06,
+      "loss": 0.6598,
+      "step": 22496
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521527018174825e-06,
+      "loss": 0.9538,
+      "step": 22497
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5207008675327495e-06,
+      "loss": 0.9648,
+      "step": 22498
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.519874734920414e-06,
+      "loss": 0.8547,
+      "step": 22499
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.519048620343825e-06,
+      "loss": 1.2675,
+      "step": 22500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51822252380899e-06,
+      "loss": 0.8643,
+      "step": 22501
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51739644532192e-06,
+      "loss": 0.9963,
+      "step": 22502
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.516570384888614e-06,
+      "loss": 1.2009,
+      "step": 22503
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.515744342515083e-06,
+      "loss": 0.8604,
+      "step": 22504
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514918318207337e-06,
+      "loss": 1.2627,
+      "step": 22505
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514092311971382e-06,
+      "loss": 0.8839,
+      "step": 22506
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51326632381322e-06,
+      "loss": 0.7563,
+      "step": 22507
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.512440353738865e-06,
+      "loss": 0.572,
+      "step": 22508
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5116144017543195e-06,
+      "loss": 1.0435,
+      "step": 22509
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51078846786559e-06,
+      "loss": 0.7633,
+      "step": 22510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50996255207868e-06,
+      "loss": 0.9576,
+      "step": 22511
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.509136654399607e-06,
+      "loss": 0.7592,
+      "step": 22512
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.508310774834362e-06,
+      "loss": 0.7941,
+      "step": 22513
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50748491338896e-06,
+      "loss": 0.5073,
+      "step": 22514
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.506659070069404e-06,
+      "loss": 0.9804,
+      "step": 22515
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.505833244881699e-06,
+      "loss": 0.7943,
+      "step": 22516
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.505007437831852e-06,
+      "loss": 0.7547,
+      "step": 22517
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.504181648925869e-06,
+      "loss": 0.7186,
+      "step": 22518
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.503355878169755e-06,
+      "loss": 0.8376,
+      "step": 22519
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5025301255695116e-06,
+      "loss": 0.845,
+      "step": 22520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.501704391131148e-06,
+      "loss": 0.9299,
+      "step": 22521
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.50087867486067e-06,
+      "loss": 0.8033,
+      "step": 22522
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.500052976764077e-06,
+      "loss": 1.3069,
+      "step": 22523
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.499227296847378e-06,
+      "loss": 1.0023,
+      "step": 22524
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.498401635116572e-06,
+      "loss": 0.9466,
+      "step": 22525
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.49757599157767e-06,
+      "loss": 0.7072,
+      "step": 22526
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.496750366236673e-06,
+      "loss": 0.6486,
+      "step": 22527
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.495924759099585e-06,
+      "loss": 0.6735,
+      "step": 22528
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.495099170172408e-06,
+      "loss": 0.7633,
+      "step": 22529
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.494273599461153e-06,
+      "loss": 0.9593,
+      "step": 22530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4934480469718154e-06,
+      "loss": 0.7872,
+      "step": 22531
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.492622512710406e-06,
+      "loss": 0.9822,
+      "step": 22532
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.491796996682921e-06,
+      "loss": 0.9047,
+      "step": 22533
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.490971498895363e-06,
+      "loss": 1.0812,
+      "step": 22534
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.490146019353744e-06,
+      "loss": 0.9888,
+      "step": 22535
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.489320558064059e-06,
+      "loss": 0.8597,
+      "step": 22536
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488495115032317e-06,
+      "loss": 0.9555,
+      "step": 22537
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4876696902645126e-06,
+      "loss": 0.9696,
+      "step": 22538
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.486844283766657e-06,
+      "loss": 1.0559,
+      "step": 22539
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.486018895544748e-06,
+      "loss": 0.9271,
+      "step": 22540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.485193525604786e-06,
+      "loss": 0.9248,
+      "step": 22541
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.484368173952783e-06,
+      "loss": 0.7305,
+      "step": 22542
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.483542840594729e-06,
+      "loss": 0.763,
+      "step": 22543
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.482717525536631e-06,
+      "loss": 0.5621,
+      "step": 22544
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.481892228784491e-06,
+      "loss": 0.8048,
+      "step": 22545
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.481066950344311e-06,
+      "loss": 1.1672,
+      "step": 22546
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48024169022209e-06,
+      "loss": 0.834,
+      "step": 22547
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479416448423833e-06,
+      "loss": 0.804,
+      "step": 22548
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.478591224955538e-06,
+      "loss": 0.7273,
+      "step": 22549
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.477766019823207e-06,
+      "loss": 1.1712,
+      "step": 22550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476940833032844e-06,
+      "loss": 0.8935,
+      "step": 22551
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476115664590448e-06,
+      "loss": 0.802,
+      "step": 22552
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.475290514502021e-06,
+      "loss": 0.5989,
+      "step": 22553
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.474465382773557e-06,
+      "loss": 0.9433,
+      "step": 22554
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473640269411064e-06,
+      "loss": 1.1032,
+      "step": 22555
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.472815174420536e-06,
+      "loss": 0.7399,
+      "step": 22556
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471990097807979e-06,
+      "loss": 1.1032,
+      "step": 22557
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471165039579391e-06,
+      "loss": 1.1983,
+      "step": 22558
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.47033999974077e-06,
+      "loss": 1.0205,
+      "step": 22559
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.469514978298119e-06,
+      "loss": 0.7468,
+      "step": 22560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.468689975257437e-06,
+      "loss": 1.0969,
+      "step": 22561
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.467864990624723e-06,
+      "loss": 1.1449,
+      "step": 22562
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.467040024405975e-06,
+      "loss": 0.657,
+      "step": 22563
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.466215076607193e-06,
+      "loss": 0.8096,
+      "step": 22564
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4653901472343745e-06,
+      "loss": 0.8792,
+      "step": 22565
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.464565236293522e-06,
+      "loss": 1.0097,
+      "step": 22566
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.463740343790633e-06,
+      "loss": 0.7863,
+      "step": 22567
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.462915469731703e-06,
+      "loss": 0.9226,
+      "step": 22568
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.462090614122737e-06,
+      "loss": 0.8912,
+      "step": 22569
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.461265776969729e-06,
+      "loss": 0.931,
+      "step": 22570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4604409582786775e-06,
+      "loss": 1.0753,
+      "step": 22571
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.459616158055585e-06,
+      "loss": 0.871,
+      "step": 22572
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.458791376306443e-06,
+      "loss": 0.8833,
+      "step": 22573
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457966613037249e-06,
+      "loss": 0.8964,
+      "step": 22574
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457141868254007e-06,
+      "loss": 1.0037,
+      "step": 22575
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4563171419627126e-06,
+      "loss": 0.748,
+      "step": 22576
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.45549243416936e-06,
+      "loss": 1.0108,
+      "step": 22577
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.454667744879951e-06,
+      "loss": 0.7109,
+      "step": 22578
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.45384307410048e-06,
+      "loss": 1.1172,
+      "step": 22579
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.453018421836946e-06,
+      "loss": 1.2607,
+      "step": 22580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452193788095341e-06,
+      "loss": 1.4304,
+      "step": 22581
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.451369172881674e-06,
+      "loss": 0.9824,
+      "step": 22582
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.450544576201927e-06,
+      "loss": 0.799,
+      "step": 22583
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4497199980621035e-06,
+      "loss": 1.0829,
+      "step": 22584
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.448895438468201e-06,
+      "loss": 0.7957,
+      "step": 22585
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.448070897426211e-06,
+      "loss": 0.7893,
+      "step": 22586
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.447246374942136e-06,
+      "loss": 0.7675,
+      "step": 22587
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.446421871021967e-06,
+      "loss": 1.0161,
+      "step": 22588
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4455973856717035e-06,
+      "loss": 0.657,
+      "step": 22589
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.444772918897336e-06,
+      "loss": 0.6994,
+      "step": 22590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443948470704867e-06,
+      "loss": 1.0385,
+      "step": 22591
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443124041100291e-06,
+      "loss": 0.6396,
+      "step": 22592
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.442299630089598e-06,
+      "loss": 1.0822,
+      "step": 22593
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.441475237678786e-06,
+      "loss": 0.6413,
+      "step": 22594
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.440650863873847e-06,
+      "loss": 0.6412,
+      "step": 22595
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439826508680783e-06,
+      "loss": 0.9716,
+      "step": 22596
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439002172105584e-06,
+      "loss": 0.8166,
+      "step": 22597
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.438177854154246e-06,
+      "loss": 0.7367,
+      "step": 22598
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4373535548327605e-06,
+      "loss": 0.6001,
+      "step": 22599
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.436529274147127e-06,
+      "loss": 0.7682,
+      "step": 22600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.435705012103338e-06,
+      "loss": 0.9401,
+      "step": 22601
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.434880768707388e-06,
+      "loss": 1.0013,
+      "step": 22602
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.434056543965268e-06,
+      "loss": 0.998,
+      "step": 22603
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.433232337882972e-06,
+      "loss": 0.6788,
+      "step": 22604
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.432408150466497e-06,
+      "loss": 1.1207,
+      "step": 22605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.431583981721835e-06,
+      "loss": 0.7108,
+      "step": 22606
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.43075983165498e-06,
+      "loss": 0.7919,
+      "step": 22607
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.429935700271923e-06,
+      "loss": 0.824,
+      "step": 22608
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.429111587578661e-06,
+      "loss": 1.0324,
+      "step": 22609
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.428287493581184e-06,
+      "loss": 0.7193,
+      "step": 22610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.427463418285485e-06,
+      "loss": 0.6567,
+      "step": 22611
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.426639361697563e-06,
+      "loss": 1.0065,
+      "step": 22612
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4258153238234e-06,
+      "loss": 0.8634,
+      "step": 22613
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.424991304668996e-06,
+      "loss": 0.5298,
+      "step": 22614
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.42416730424034e-06,
+      "loss": 0.8422,
+      "step": 22615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.423343322543426e-06,
+      "loss": 0.8728,
+      "step": 22616
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.422519359584244e-06,
+      "loss": 1.0284,
+      "step": 22617
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.421695415368788e-06,
+      "loss": 1.0285,
+      "step": 22618
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420871489903051e-06,
+      "loss": 0.8125,
+      "step": 22619
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.42004758319302e-06,
+      "loss": 0.8247,
+      "step": 22620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.419223695244692e-06,
+      "loss": 0.7899,
+      "step": 22621
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.418399826064058e-06,
+      "loss": 0.9653,
+      "step": 22622
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.417575975657103e-06,
+      "loss": 1.0068,
+      "step": 22623
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.416752144029824e-06,
+      "loss": 0.7185,
+      "step": 22624
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4159283311882115e-06,
+      "loss": 0.8357,
+      "step": 22625
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.415104537138251e-06,
+      "loss": 0.6551,
+      "step": 22626
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41428076188594e-06,
+      "loss": 1.0885,
+      "step": 22627
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.413457005437268e-06,
+      "loss": 0.618,
+      "step": 22628
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41263326779822e-06,
+      "loss": 0.6649,
+      "step": 22629
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.411809548974792e-06,
+      "loss": 0.9246,
+      "step": 22630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410985848972973e-06,
+      "loss": 0.75,
+      "step": 22631
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410162167798756e-06,
+      "loss": 1.0141,
+      "step": 22632
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4093385054581234e-06,
+      "loss": 0.9999,
+      "step": 22633
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4085148619570675e-06,
+      "loss": 0.7445,
+      "step": 22634
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.40769123730158e-06,
+      "loss": 0.7709,
+      "step": 22635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4068676314976515e-06,
+      "loss": 0.943,
+      "step": 22636
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4060440445512685e-06,
+      "loss": 0.9568,
+      "step": 22637
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.40522047646842e-06,
+      "loss": 0.8521,
+      "step": 22638
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.404396927255098e-06,
+      "loss": 0.8336,
+      "step": 22639
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4035733969172895e-06,
+      "loss": 0.5407,
+      "step": 22640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.402749885460982e-06,
+      "loss": 0.8009,
+      "step": 22641
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.401926392892172e-06,
+      "loss": 0.8052,
+      "step": 22642
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.401102919216836e-06,
+      "loss": 0.9514,
+      "step": 22643
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.400279464440967e-06,
+      "loss": 0.8845,
+      "step": 22644
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.399456028570557e-06,
+      "loss": 1.151,
+      "step": 22645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.398632611611591e-06,
+      "loss": 0.8072,
+      "step": 22646
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.397809213570057e-06,
+      "loss": 0.9791,
+      "step": 22647
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.396985834451944e-06,
+      "loss": 0.5236,
+      "step": 22648
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.39616247426324e-06,
+      "loss": 1.0213,
+      "step": 22649
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.395339133009931e-06,
+      "loss": 0.7925,
+      "step": 22650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.394515810698002e-06,
+      "loss": 0.7921,
+      "step": 22651
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.393692507333452e-06,
+      "loss": 0.7678,
+      "step": 22652
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.392869222922251e-06,
+      "loss": 0.9885,
+      "step": 22653
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3920459574703975e-06,
+      "loss": 0.9362,
+      "step": 22654
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.391222710983876e-06,
+      "loss": 0.9357,
+      "step": 22655
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.390399483468669e-06,
+      "loss": 0.7923,
+      "step": 22656
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.389576274930769e-06,
+      "loss": 0.7778,
+      "step": 22657
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3887530853761625e-06,
+      "loss": 0.8122,
+      "step": 22658
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3879299148108315e-06,
+      "loss": 0.9936,
+      "step": 22659
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.387106763240763e-06,
+      "loss": 0.9407,
+      "step": 22660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.386283630671946e-06,
+      "loss": 1.0323,
+      "step": 22661
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.385460517110368e-06,
+      "loss": 1.0395,
+      "step": 22662
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.384637422562006e-06,
+      "loss": 0.5375,
+      "step": 22663
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.383814347032854e-06,
+      "loss": 0.8126,
+      "step": 22664
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.382991290528892e-06,
+      "loss": 1.1259,
+      "step": 22665
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.38216825305611e-06,
+      "loss": 1.078,
+      "step": 22666
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381345234620492e-06,
+      "loss": 0.9614,
+      "step": 22667
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.380522235228021e-06,
+      "loss": 0.635,
+      "step": 22668
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.37969925488468e-06,
+      "loss": 0.6906,
+      "step": 22669
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.378876293596461e-06,
+      "loss": 0.8383,
+      "step": 22670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.378053351369344e-06,
+      "loss": 0.8456,
+      "step": 22671
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.377230428209316e-06,
+      "loss": 1.0254,
+      "step": 22672
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.376407524122357e-06,
+      "loss": 0.7456,
+      "step": 22673
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3755846391144545e-06,
+      "loss": 0.9535,
+      "step": 22674
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.374761773191592e-06,
+      "loss": 1.0143,
+      "step": 22675
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.373938926359754e-06,
+      "loss": 1.198,
+      "step": 22676
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.373116098624922e-06,
+      "loss": 1.0448,
+      "step": 22677
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.372293289993081e-06,
+      "loss": 0.9391,
+      "step": 22678
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3714705004702166e-06,
+      "loss": 0.7983,
+      "step": 22679
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.370647730062311e-06,
+      "loss": 0.7072,
+      "step": 22680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.369824978775344e-06,
+      "loss": 0.9492,
+      "step": 22681
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3690022466153084e-06,
+      "loss": 0.6501,
+      "step": 22682
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.368179533588176e-06,
+      "loss": 0.6524,
+      "step": 22683
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.367356839699935e-06,
+      "loss": 1.033,
+      "step": 22684
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3665341649565665e-06,
+      "loss": 0.7477,
+      "step": 22685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.365711509364055e-06,
+      "loss": 1.0133,
+      "step": 22686
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.364888872928381e-06,
+      "loss": 1.1581,
+      "step": 22687
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3640662556555286e-06,
+      "loss": 1.0143,
+      "step": 22688
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.36324365755148e-06,
+      "loss": 0.5409,
+      "step": 22689
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.362421078622213e-06,
+      "loss": 1.0688,
+      "step": 22690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.361598518873715e-06,
+      "loss": 1.226,
+      "step": 22691
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.360775978311968e-06,
+      "loss": 0.7847,
+      "step": 22692
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.35995345694295e-06,
+      "loss": 0.9025,
+      "step": 22693
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.359130954772642e-06,
+      "loss": 0.8735,
+      "step": 22694
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.358308471807028e-06,
+      "loss": 0.7168,
+      "step": 22695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.357486008052086e-06,
+      "loss": 0.8326,
+      "step": 22696
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3566635635138e-06,
+      "loss": 0.956,
+      "step": 22697
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.355841138198152e-06,
+      "loss": 0.9569,
+      "step": 22698
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.355018732111117e-06,
+      "loss": 0.7367,
+      "step": 22699
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.354196345258683e-06,
+      "loss": 0.4849,
+      "step": 22700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.353373977646827e-06,
+      "loss": 1.1262,
+      "step": 22701
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.35255162928153e-06,
+      "loss": 0.8737,
+      "step": 22702
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.351729300168771e-06,
+      "loss": 0.9393,
+      "step": 22703
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.35090699031453e-06,
+      "loss": 0.6977,
+      "step": 22704
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.350084699724785e-06,
+      "loss": 1.0517,
+      "step": 22705
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.349262428405521e-06,
+      "loss": 0.8121,
+      "step": 22706
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.348440176362715e-06,
+      "loss": 0.9982,
+      "step": 22707
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.347617943602343e-06,
+      "loss": 1.2249,
+      "step": 22708
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3467957301303915e-06,
+      "loss": 0.7729,
+      "step": 22709
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3459735359528366e-06,
+      "loss": 0.7321,
+      "step": 22710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.345151361075655e-06,
+      "loss": 0.9822,
+      "step": 22711
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.344329205504832e-06,
+      "loss": 0.83,
+      "step": 22712
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3435070692463386e-06,
+      "loss": 0.943,
+      "step": 22713
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.342684952306154e-06,
+      "loss": 0.8729,
+      "step": 22714
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3418628546902625e-06,
+      "loss": 0.5381,
+      "step": 22715
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.341040776404639e-06,
+      "loss": 0.8155,
+      "step": 22716
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3402187174552595e-06,
+      "loss": 0.6841,
+      "step": 22717
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.339396677848107e-06,
+      "loss": 1.0596,
+      "step": 22718
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.338574657589159e-06,
+      "loss": 0.892,
+      "step": 22719
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.337752656684391e-06,
+      "loss": 1.1541,
+      "step": 22720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3369306751397795e-06,
+      "loss": 0.8232,
+      "step": 22721
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.33610871296131e-06,
+      "loss": 0.6066,
+      "step": 22722
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.335286770154946e-06,
+      "loss": 0.5575,
+      "step": 22723
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.334464846726675e-06,
+      "loss": 0.8962,
+      "step": 22724
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.333642942682473e-06,
+      "loss": 1.0784,
+      "step": 22725
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.332821058028312e-06,
+      "loss": 0.6554,
+      "step": 22726
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3319991927701764e-06,
+      "loss": 0.8581,
+      "step": 22727
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.331177346914038e-06,
+      "loss": 0.7766,
+      "step": 22728
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.330355520465874e-06,
+      "loss": 0.9475,
+      "step": 22729
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.329533713431658e-06,
+      "loss": 0.7189,
+      "step": 22730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.328711925817373e-06,
+      "loss": 0.9564,
+      "step": 22731
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.327890157628994e-06,
+      "loss": 0.8979,
+      "step": 22732
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.327068408872491e-06,
+      "loss": 0.7403,
+      "step": 22733
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326246679553844e-06,
+      "loss": 0.9402,
+      "step": 22734
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.325424969679027e-06,
+      "loss": 0.8247,
+      "step": 22735
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3246032792540166e-06,
+      "loss": 1.0069,
+      "step": 22736
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3237816082847905e-06,
+      "loss": 0.8747,
+      "step": 22737
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3229599567773205e-06,
+      "loss": 0.7165,
+      "step": 22738
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.32213832473758e-06,
+      "loss": 0.9369,
+      "step": 22739
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3213167121715514e-06,
+      "loss": 0.7599,
+      "step": 22740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3204951190852024e-06,
+      "loss": 0.8372,
+      "step": 22741
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.319673545484514e-06,
+      "loss": 1.1524,
+      "step": 22742
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.318851991375455e-06,
+      "loss": 0.8183,
+      "step": 22743
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3180304567640004e-06,
+      "loss": 0.8083,
+      "step": 22744
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317208941656129e-06,
+      "loss": 1.0795,
+      "step": 22745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3163874460578095e-06,
+      "loss": 0.8913,
+      "step": 22746
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31556596997502e-06,
+      "loss": 0.9469,
+      "step": 22747
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.314744513413732e-06,
+      "loss": 0.512,
+      "step": 22748
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31392307637992e-06,
+      "loss": 0.7292,
+      "step": 22749
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.313101658879559e-06,
+      "loss": 0.7086,
+      "step": 22750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3122802609186185e-06,
+      "loss": 0.9388,
+      "step": 22751
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.311458882503081e-06,
+      "loss": 0.8138,
+      "step": 22752
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.310637523638906e-06,
+      "loss": 0.9759,
+      "step": 22753
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.309816184332077e-06,
+      "loss": 0.8254,
+      "step": 22754
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.308994864588562e-06,
+      "loss": 0.8273,
+      "step": 22755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3081735644143355e-06,
+      "loss": 0.7489,
+      "step": 22756
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.307352283815368e-06,
+      "loss": 0.6417,
+      "step": 22757
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.306531022797636e-06,
+      "loss": 0.6807,
+      "step": 22758
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.30570978136711e-06,
+      "loss": 1.0482,
+      "step": 22759
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.304888559529759e-06,
+      "loss": 0.9489,
+      "step": 22760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.304067357291559e-06,
+      "loss": 0.9297,
+      "step": 22761
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.303246174658482e-06,
+      "loss": 0.8304,
+      "step": 22762
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3024250116364995e-06,
+      "loss": 0.8112,
+      "step": 22763
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.301603868231579e-06,
+      "loss": 0.8223,
+      "step": 22764
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.300782744449696e-06,
+      "loss": 0.6311,
+      "step": 22765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2999616402968175e-06,
+      "loss": 0.7044,
+      "step": 22766
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2991405557789205e-06,
+      "loss": 0.716,
+      "step": 22767
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.298319490901975e-06,
+      "loss": 0.8324,
+      "step": 22768
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2974984456719455e-06,
+      "loss": 0.8832,
+      "step": 22769
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.296677420094811e-06,
+      "loss": 0.8504,
+      "step": 22770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.295856414176537e-06,
+      "loss": 0.9006,
+      "step": 22771
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2950354279230985e-06,
+      "loss": 0.9599,
+      "step": 22772
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.294214461340461e-06,
+      "loss": 0.7501,
+      "step": 22773
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293393514434597e-06,
+      "loss": 0.9392,
+      "step": 22774
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.292572587211473e-06,
+      "loss": 0.4992,
+      "step": 22775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.291751679677064e-06,
+      "loss": 0.7673,
+      "step": 22776
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2909307918373365e-06,
+      "loss": 0.7969,
+      "step": 22777
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29010992369826e-06,
+      "loss": 0.9238,
+      "step": 22778
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.289289075265807e-06,
+      "loss": 0.9419,
+      "step": 22779
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.288468246545946e-06,
+      "loss": 0.8294,
+      "step": 22780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2876474375446425e-06,
+      "loss": 0.6258,
+      "step": 22781
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.286826648267872e-06,
+      "loss": 1.0666,
+      "step": 22782
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2860058787215956e-06,
+      "loss": 0.7559,
+      "step": 22783
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285185128911786e-06,
+      "loss": 0.828,
+      "step": 22784
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.284364398844412e-06,
+      "loss": 0.7328,
+      "step": 22785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.283543688525443e-06,
+      "loss": 1.2294,
+      "step": 22786
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.282722997960843e-06,
+      "loss": 1.1161,
+      "step": 22787
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.281902327156584e-06,
+      "loss": 0.9601,
+      "step": 22788
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.281081676118635e-06,
+      "loss": 1.0752,
+      "step": 22789
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.280261044852963e-06,
+      "loss": 1.0069,
+      "step": 22790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27944043336553e-06,
+      "loss": 0.9521,
+      "step": 22791
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.278619841662316e-06,
+      "loss": 1.1204,
+      "step": 22792
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.277799269749275e-06,
+      "loss": 0.7753,
+      "step": 22793
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27697871763238e-06,
+      "loss": 1.0099,
+      "step": 22794
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2761581853176e-06,
+      "loss": 1.0192,
+      "step": 22795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.275337672810897e-06,
+      "loss": 0.9214,
+      "step": 22796
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.274517180118244e-06,
+      "loss": 0.8784,
+      "step": 22797
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2736967072456035e-06,
+      "loss": 1.1484,
+      "step": 22798
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.272876254198944e-06,
+      "loss": 1.0906,
+      "step": 22799
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27205582098423e-06,
+      "loss": 0.8373,
+      "step": 22800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2712354076074295e-06,
+      "loss": 0.8522,
+      "step": 22801
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.270415014074511e-06,
+      "loss": 0.757,
+      "step": 22802
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.269594640391435e-06,
+      "loss": 0.9555,
+      "step": 22803
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.268774286564171e-06,
+      "loss": 0.6286,
+      "step": 22804
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.267953952598681e-06,
+      "loss": 0.8052,
+      "step": 22805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.267133638500936e-06,
+      "loss": 0.8555,
+      "step": 22806
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266313344276899e-06,
+      "loss": 0.9988,
+      "step": 22807
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.265493069932532e-06,
+      "loss": 0.5466,
+      "step": 22808
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.264672815473803e-06,
+      "loss": 0.7184,
+      "step": 22809
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2638525809066805e-06,
+      "loss": 0.7691,
+      "step": 22810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.263032366237124e-06,
+      "loss": 1.0654,
+      "step": 22811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.262212171471103e-06,
+      "loss": 1.1906,
+      "step": 22812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2613919966145764e-06,
+      "loss": 1.0034,
+      "step": 22813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.260571841673509e-06,
+      "loss": 0.9188,
+      "step": 22814
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.25975170665387e-06,
+      "loss": 0.4791,
+      "step": 22815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.258931591561621e-06,
+      "loss": 0.817,
+      "step": 22816
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.258111496402725e-06,
+      "loss": 0.9012,
+      "step": 22817
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2572914211831455e-06,
+      "loss": 1.0289,
+      "step": 22818
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.25647136590885e-06,
+      "loss": 0.7553,
+      "step": 22819
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.255651330585798e-06,
+      "loss": 1.202,
+      "step": 22820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.254831315219952e-06,
+      "loss": 0.8307,
+      "step": 22821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.254011319817286e-06,
+      "loss": 0.7261,
+      "step": 22822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.253191344383747e-06,
+      "loss": 0.9162,
+      "step": 22823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2523713889253075e-06,
+      "loss": 0.8494,
+      "step": 22824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251551453447928e-06,
+      "loss": 0.7337,
+      "step": 22825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.250731537957572e-06,
+      "loss": 0.9756,
+      "step": 22826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249911642460199e-06,
+      "loss": 0.7912,
+      "step": 22827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249091766961778e-06,
+      "loss": 0.6788,
+      "step": 22828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.248271911468265e-06,
+      "loss": 0.8744,
+      "step": 22829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.247452075985622e-06,
+      "loss": 0.8747,
+      "step": 22830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.246632260519817e-06,
+      "loss": 0.9613,
+      "step": 22831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.245812465076809e-06,
+      "loss": 0.9048,
+      "step": 22832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.244992689662555e-06,
+      "loss": 1.4672,
+      "step": 22833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.244172934283023e-06,
+      "loss": 0.8719,
+      "step": 22834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.24335319894417e-06,
+      "loss": 0.8915,
+      "step": 22835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.242533483651957e-06,
+      "loss": 0.6887,
+      "step": 22836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.241713788412348e-06,
+      "loss": 1.0094,
+      "step": 22837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.240894113231302e-06,
+      "loss": 0.7602,
+      "step": 22838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.24007445811478e-06,
+      "loss": 0.9154,
+      "step": 22839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.239254823068746e-06,
+      "loss": 0.6979,
+      "step": 22840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.238435208099155e-06,
+      "loss": 1.0287,
+      "step": 22841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.237615613211973e-06,
+      "loss": 1.1287,
+      "step": 22842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.236796038413155e-06,
+      "loss": 1.1886,
+      "step": 22843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2359764837086635e-06,
+      "loss": 1.1199,
+      "step": 22844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.235156949104455e-06,
+      "loss": 0.9807,
+      "step": 22845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.234337434606495e-06,
+      "loss": 1.1258,
+      "step": 22846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.23351794022074e-06,
+      "loss": 0.8271,
+      "step": 22847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.232698465953148e-06,
+      "loss": 0.7103,
+      "step": 22848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.231879011809681e-06,
+      "loss": 1.0158,
+      "step": 22849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.231059577796298e-06,
+      "loss": 0.7978,
+      "step": 22850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230240163918957e-06,
+      "loss": 0.8622,
+      "step": 22851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.229420770183619e-06,
+      "loss": 0.9122,
+      "step": 22852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.22860139659624e-06,
+      "loss": 1.0793,
+      "step": 22853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.227782043162776e-06,
+      "loss": 0.7946,
+      "step": 22854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.226962709889191e-06,
+      "loss": 0.594,
+      "step": 22855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.226143396781442e-06,
+      "loss": 0.9781,
+      "step": 22856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.225324103845484e-06,
+      "loss": 0.7853,
+      "step": 22857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.224504831087278e-06,
+      "loss": 0.8092,
+      "step": 22858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.223685578512783e-06,
+      "loss": 0.9541,
+      "step": 22859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.222866346127952e-06,
+      "loss": 1.0445,
+      "step": 22860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.222047133938746e-06,
+      "loss": 1.0409,
+      "step": 22861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2212279419511265e-06,
+      "loss": 0.9941,
+      "step": 22862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.22040877017104e-06,
+      "loss": 0.5074,
+      "step": 22863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2195896186044526e-06,
+      "loss": 1.0647,
+      "step": 22864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.218770487257317e-06,
+      "loss": 0.6591,
+      "step": 22865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2179513761355905e-06,
+      "loss": 0.8956,
+      "step": 22866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.217132285245232e-06,
+      "loss": 0.7742,
+      "step": 22867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.216313214592195e-06,
+      "loss": 0.6851,
+      "step": 22868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.215494164182439e-06,
+      "loss": 0.7452,
+      "step": 22869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2146751340219165e-06,
+      "loss": 1.1226,
+      "step": 22870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2138561241165875e-06,
+      "loss": 1.0011,
+      "step": 22871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.213037134472409e-06,
+      "loss": 0.7693,
+      "step": 22872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.21221816509533e-06,
+      "loss": 0.9388,
+      "step": 22873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.211399215991313e-06,
+      "loss": 0.7281,
+      "step": 22874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.210580287166307e-06,
+      "loss": 0.9568,
+      "step": 22875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.209761378626272e-06,
+      "loss": 0.9276,
+      "step": 22876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.208942490377163e-06,
+      "loss": 0.9099,
+      "step": 22877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.208123622424934e-06,
+      "loss": 1.0095,
+      "step": 22878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2073047747755385e-06,
+      "loss": 1.0814,
+      "step": 22879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.206485947434934e-06,
+      "loss": 0.8975,
+      "step": 22880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.205667140409074e-06,
+      "loss": 0.8684,
+      "step": 22881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.204848353703916e-06,
+      "loss": 0.8566,
+      "step": 22882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.204029587325409e-06,
+      "loss": 0.6471,
+      "step": 22883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.203210841279506e-06,
+      "loss": 0.5914,
+      "step": 22884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2023921155721675e-06,
+      "loss": 0.9421,
+      "step": 22885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.201573410209344e-06,
+      "loss": 0.6563,
+      "step": 22886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.200754725196987e-06,
+      "loss": 0.8211,
+      "step": 22887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.199936060541054e-06,
+      "loss": 1.1686,
+      "step": 22888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.199117416247498e-06,
+      "loss": 0.9455,
+      "step": 22889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.198298792322271e-06,
+      "loss": 0.8637,
+      "step": 22890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.197480188771325e-06,
+      "loss": 0.5719,
+      "step": 22891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.196661605600618e-06,
+      "loss": 0.8823,
+      "step": 22892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.195843042816097e-06,
+      "loss": 1.0948,
+      "step": 22893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.195024500423713e-06,
+      "loss": 1.0257,
+      "step": 22894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.194205978429426e-06,
+      "loss": 0.5703,
+      "step": 22895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.193387476839185e-06,
+      "loss": 0.7863,
+      "step": 22896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.192568995658937e-06,
+      "loss": 1.0395,
+      "step": 22897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191750534894644e-06,
+      "loss": 1.0545,
+      "step": 22898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.190932094552252e-06,
+      "loss": 0.8918,
+      "step": 22899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.190113674637714e-06,
+      "loss": 0.5771,
+      "step": 22900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.18929527515698e-06,
+      "loss": 1.0523,
+      "step": 22901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.188476896116008e-06,
+      "loss": 0.8439,
+      "step": 22902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1876585375207386e-06,
+      "loss": 0.6587,
+      "step": 22903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.18684019937713e-06,
+      "loss": 0.8596,
+      "step": 22904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.186021881691132e-06,
+      "loss": 1.0219,
+      "step": 22905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.185203584468694e-06,
+      "loss": 0.9229,
+      "step": 22906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.184385307715769e-06,
+      "loss": 0.94,
+      "step": 22907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.183567051438308e-06,
+      "loss": 0.6973,
+      "step": 22908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.18274881564226e-06,
+      "loss": 0.9556,
+      "step": 22909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.181930600333573e-06,
+      "loss": 0.9831,
+      "step": 22910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.181112405518201e-06,
+      "loss": 0.7099,
+      "step": 22911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180294231202097e-06,
+      "loss": 0.8336,
+      "step": 22912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.179476077391201e-06,
+      "loss": 1.0993,
+      "step": 22913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1786579440914706e-06,
+      "loss": 0.6408,
+      "step": 22914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1778398313088505e-06,
+      "loss": 0.7319,
+      "step": 22915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.177021739049295e-06,
+      "loss": 1.0846,
+      "step": 22916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1762036673187485e-06,
+      "loss": 1.0894,
+      "step": 22917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.175385616123165e-06,
+      "loss": 0.9343,
+      "step": 22918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.174567585468488e-06,
+      "loss": 0.9737,
+      "step": 22919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.173749575360671e-06,
+      "loss": 0.9686,
+      "step": 22920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.172931585805662e-06,
+      "loss": 0.8929,
+      "step": 22921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.17211361680941e-06,
+      "loss": 0.6925,
+      "step": 22922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.17129566837786e-06,
+      "loss": 0.7196,
+      "step": 22923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.170477740516959e-06,
+      "loss": 0.9564,
+      "step": 22924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.169659833232661e-06,
+      "loss": 1.0732,
+      "step": 22925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168841946530911e-06,
+      "loss": 1.0478,
+      "step": 22926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168024080417656e-06,
+      "loss": 0.856,
+      "step": 22927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.167206234898843e-06,
+      "loss": 1.0326,
+      "step": 22928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.166388409980422e-06,
+      "loss": 0.5372,
+      "step": 22929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.165570605668341e-06,
+      "loss": 0.9321,
+      "step": 22930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.164752821968542e-06,
+      "loss": 0.8729,
+      "step": 22931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163935058886982e-06,
+      "loss": 0.8206,
+      "step": 22932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163117316429593e-06,
+      "loss": 1.146,
+      "step": 22933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1622995946023345e-06,
+      "loss": 0.8336,
+      "step": 22934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1614818934111475e-06,
+      "loss": 0.9327,
+      "step": 22935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.16066421286198e-06,
+      "loss": 0.8616,
+      "step": 22936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.159846552960774e-06,
+      "loss": 0.7875,
+      "step": 22937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.159028913713483e-06,
+      "loss": 0.8531,
+      "step": 22938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.158211295126048e-06,
+      "loss": 0.8624,
+      "step": 22939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.157393697204416e-06,
+      "loss": 0.9085,
+      "step": 22940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.156576119954532e-06,
+      "loss": 0.73,
+      "step": 22941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1557585633823455e-06,
+      "loss": 0.8667,
+      "step": 22942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.154941027493797e-06,
+      "loss": 0.8956,
+      "step": 22943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.154123512294833e-06,
+      "loss": 0.9937,
+      "step": 22944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.153306017791398e-06,
+      "loss": 1.2437,
+      "step": 22945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.152488543989438e-06,
+      "loss": 0.7966,
+      "step": 22946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.151671090894896e-06,
+      "loss": 0.8288,
+      "step": 22947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.150853658513721e-06,
+      "loss": 1.009,
+      "step": 22948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1500362468518514e-06,
+      "loss": 0.7056,
+      "step": 22949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1492188559152364e-06,
+      "loss": 0.7422,
+      "step": 22950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14840148570982e-06,
+      "loss": 0.8139,
+      "step": 22951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.147584136241546e-06,
+      "loss": 0.7665,
+      "step": 22952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.146766807516354e-06,
+      "loss": 0.8717,
+      "step": 22953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.145949499540192e-06,
+      "loss": 0.8934,
+      "step": 22954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.145132212318999e-06,
+      "loss": 0.9253,
+      "step": 22955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1443149458587235e-06,
+      "loss": 0.9791,
+      "step": 22956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.143497700165307e-06,
+      "loss": 0.7773,
+      "step": 22957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.142680475244691e-06,
+      "loss": 1.0404,
+      "step": 22958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.141863271102823e-06,
+      "loss": 0.8556,
+      "step": 22959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14104608774564e-06,
+      "loss": 0.9675,
+      "step": 22960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.140228925179089e-06,
+      "loss": 0.9026,
+      "step": 22961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1394117834091125e-06,
+      "loss": 0.738,
+      "step": 22962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.138594662441649e-06,
+      "loss": 0.8087,
+      "step": 22963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1377775622826416e-06,
+      "loss": 1.063,
+      "step": 22964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.136960482938035e-06,
+      "loss": 0.849,
+      "step": 22965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.136143424413768e-06,
+      "loss": 0.914,
+      "step": 22966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.135326386715784e-06,
+      "loss": 0.835,
+      "step": 22967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.134509369850027e-06,
+      "loss": 1.0033,
+      "step": 22968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133692373822436e-06,
+      "loss": 0.7541,
+      "step": 22969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132875398638952e-06,
+      "loss": 0.8832,
+      "step": 22970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132058444305514e-06,
+      "loss": 0.7321,
+      "step": 22971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.131241510828072e-06,
+      "loss": 0.7574,
+      "step": 22972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.130424598212553e-06,
+      "loss": 1.0998,
+      "step": 22973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.129607706464909e-06,
+      "loss": 1.0184,
+      "step": 22974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.128790835591075e-06,
+      "loss": 0.9025,
+      "step": 22975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.127973985596992e-06,
+      "loss": 0.6456,
+      "step": 22976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.127157156488602e-06,
+      "loss": 0.8892,
+      "step": 22977
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.126340348271845e-06,
+      "loss": 0.9775,
+      "step": 22978
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.125523560952659e-06,
+      "loss": 0.7566,
+      "step": 22979
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124706794536984e-06,
+      "loss": 1.1211,
+      "step": 22980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.123890049030763e-06,
+      "loss": 0.9303,
+      "step": 22981
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.123073324439933e-06,
+      "loss": 0.8167,
+      "step": 22982
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.122256620770433e-06,
+      "loss": 0.75,
+      "step": 22983
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.121439938028202e-06,
+      "loss": 0.8106,
+      "step": 22984
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.120623276219178e-06,
+      "loss": 0.8272,
+      "step": 22985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1198066353493024e-06,
+      "loss": 0.7585,
+      "step": 22986
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1189900154245116e-06,
+      "loss": 1.0448,
+      "step": 22987
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.118173416450747e-06,
+      "loss": 0.8919,
+      "step": 22988
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.117356838433943e-06,
+      "loss": 0.9168,
+      "step": 22989
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.116540281380042e-06,
+      "loss": 0.7159,
+      "step": 22990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1157237452949805e-06,
+      "loss": 1.2128,
+      "step": 22991
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.114907230184698e-06,
+      "loss": 0.6926,
+      "step": 22992
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11409073605513e-06,
+      "loss": 0.8566,
+      "step": 22993
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.113274262912211e-06,
+      "loss": 0.9589,
+      "step": 22994
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.112457810761883e-06,
+      "loss": 0.7853,
+      "step": 22995
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.111641379610085e-06,
+      "loss": 0.8518,
+      "step": 22996
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.110824969462751e-06,
+      "loss": 0.9121,
+      "step": 22997
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.110008580325815e-06,
+      "loss": 0.991,
+      "step": 22998
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.109192212205221e-06,
+      "loss": 0.6567,
+      "step": 22999
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.108375865106903e-06,
+      "loss": 0.5246,
+      "step": 23000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.107559539036793e-06,
+      "loss": 0.8874,
+      "step": 23001
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.106743234000837e-06,
+      "loss": 1.2029,
+      "step": 23002
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.105926950004961e-06,
+      "loss": 0.9605,
+      "step": 23003
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.105110687055108e-06,
+      "loss": 0.6485,
+      "step": 23004
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.104294445157209e-06,
+      "loss": 0.812,
+      "step": 23005
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.103478224317204e-06,
+      "loss": 0.8731,
+      "step": 23006
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.102662024541025e-06,
+      "loss": 0.7951,
+      "step": 23007
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101845845834612e-06,
+      "loss": 0.7096,
+      "step": 23008
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101029688203897e-06,
+      "loss": 0.9016,
+      "step": 23009
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.100213551654816e-06,
+      "loss": 0.6441,
+      "step": 23010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.099397436193303e-06,
+      "loss": 0.7203,
+      "step": 23011
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.098581341825297e-06,
+      "loss": 0.6918,
+      "step": 23012
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.097765268556729e-06,
+      "loss": 1.0622,
+      "step": 23013
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096949216393534e-06,
+      "loss": 0.8974,
+      "step": 23014
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096133185341645e-06,
+      "loss": 0.7441,
+      "step": 23015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0953171754069974e-06,
+      "loss": 0.5936,
+      "step": 23016
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.094501186595526e-06,
+      "loss": 1.0208,
+      "step": 23017
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.093685218913167e-06,
+      "loss": 0.843,
+      "step": 23018
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.092869272365849e-06,
+      "loss": 0.9732,
+      "step": 23019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.09205334695951e-06,
+      "loss": 0.9361,
+      "step": 23020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.091237442700082e-06,
+      "loss": 0.9102,
+      "step": 23021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0904215595935015e-06,
+      "loss": 0.9326,
+      "step": 23022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.089605697645695e-06,
+      "loss": 0.7311,
+      "step": 23023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.088789856862599e-06,
+      "loss": 0.949,
+      "step": 23024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.087974037250146e-06,
+      "loss": 0.8445,
+      "step": 23025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.08715823881427e-06,
+      "loss": 0.9352,
+      "step": 23026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0863424615609035e-06,
+      "loss": 1.1018,
+      "step": 23027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085526705495975e-06,
+      "loss": 1.0162,
+      "step": 23028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.084710970625424e-06,
+      "loss": 0.9417,
+      "step": 23029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.083895256955175e-06,
+      "loss": 1.0873,
+      "step": 23030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0830795644911664e-06,
+      "loss": 0.9977,
+      "step": 23031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.082263893239329e-06,
+      "loss": 0.6673,
+      "step": 23032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.081448243205591e-06,
+      "loss": 0.8095,
+      "step": 23033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0806326143958816e-06,
+      "loss": 0.8204,
+      "step": 23034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.079817006816139e-06,
+      "loss": 0.906,
+      "step": 23035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.079001420472291e-06,
+      "loss": 0.9824,
+      "step": 23036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.078185855370267e-06,
+      "loss": 0.8976,
+      "step": 23037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.077370311516003e-06,
+      "loss": 0.7773,
+      "step": 23038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.076554788915427e-06,
+      "loss": 0.7042,
+      "step": 23039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.075739287574467e-06,
+      "loss": 0.8526,
+      "step": 23040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.074923807499054e-06,
+      "loss": 0.931,
+      "step": 23041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.074108348695128e-06,
+      "loss": 0.8103,
+      "step": 23042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0732929111686035e-06,
+      "loss": 0.7501,
+      "step": 23043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07247749492542e-06,
+      "loss": 1.2763,
+      "step": 23044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0716620999715055e-06,
+      "loss": 1.0316,
+      "step": 23045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.070846726312787e-06,
+      "loss": 0.9722,
+      "step": 23046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0700313739551985e-06,
+      "loss": 0.726,
+      "step": 23047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.069216042904667e-06,
+      "loss": 0.7821,
+      "step": 23048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.068400733167122e-06,
+      "loss": 0.8346,
+      "step": 23049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.067585444748492e-06,
+      "loss": 0.8149,
+      "step": 23050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.066770177654707e-06,
+      "loss": 0.9929,
+      "step": 23051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.065954931891697e-06,
+      "loss": 1.4147,
+      "step": 23052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.065139707465386e-06,
+      "loss": 0.6861,
+      "step": 23053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0643245043817055e-06,
+      "loss": 0.8112,
+      "step": 23054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.063509322646581e-06,
+      "loss": 0.8999,
+      "step": 23055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.062694162265945e-06,
+      "loss": 0.9311,
+      "step": 23056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.061879023245723e-06,
+      "loss": 0.6837,
+      "step": 23057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.061063905591843e-06,
+      "loss": 1.0292,
+      "step": 23058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.060248809310231e-06,
+      "loss": 0.7131,
+      "step": 23059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0594337344068175e-06,
+      "loss": 0.8529,
+      "step": 23060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.05861868088753e-06,
+      "loss": 0.8926,
+      "step": 23061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0578036487582945e-06,
+      "loss": 0.877,
+      "step": 23062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.056988638025037e-06,
+      "loss": 0.8817,
+      "step": 23063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.056173648693681e-06,
+      "loss": 0.9232,
+      "step": 23064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.05535868077016e-06,
+      "loss": 0.8434,
+      "step": 23065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.054543734260397e-06,
+      "loss": 1.0581,
+      "step": 23066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.053728809170322e-06,
+      "loss": 0.9711,
+      "step": 23067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052913905505854e-06,
+      "loss": 0.8366,
+      "step": 23068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052099023272924e-06,
+      "loss": 0.6545,
+      "step": 23069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.051284162477459e-06,
+      "loss": 0.9245,
+      "step": 23070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.050469323125381e-06,
+      "loss": 0.8651,
+      "step": 23071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.049654505222623e-06,
+      "loss": 0.7617,
+      "step": 23072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048839708775099e-06,
+      "loss": 0.7567,
+      "step": 23073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048024933788744e-06,
+      "loss": 0.8041,
+      "step": 23074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.047210180269478e-06,
+      "loss": 1.0269,
+      "step": 23075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0463954482232275e-06,
+      "loss": 0.912,
+      "step": 23076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.045580737655915e-06,
+      "loss": 0.587,
+      "step": 23077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0447660485734705e-06,
+      "loss": 0.6568,
+      "step": 23078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0439513809818146e-06,
+      "loss": 0.9786,
+      "step": 23079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043136734886871e-06,
+      "loss": 0.7123,
+      "step": 23080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.042322110294568e-06,
+      "loss": 0.6451,
+      "step": 23081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.041507507210828e-06,
+      "loss": 0.698,
+      "step": 23082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.040692925641572e-06,
+      "loss": 0.8337,
+      "step": 23083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039878365592726e-06,
+      "loss": 1.3056,
+      "step": 23084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039063827070214e-06,
+      "loss": 0.8678,
+      "step": 23085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.038249310079955e-06,
+      "loss": 0.9363,
+      "step": 23086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.037434814627879e-06,
+      "loss": 0.9869,
+      "step": 23087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.036620340719906e-06,
+      "loss": 0.7026,
+      "step": 23088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0358058883619565e-06,
+      "loss": 1.0601,
+      "step": 23089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0349914575599585e-06,
+      "loss": 0.7859,
+      "step": 23090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.03417704831983e-06,
+      "loss": 1.0051,
+      "step": 23091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0333626606475e-06,
+      "loss": 0.8311,
+      "step": 23092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0325482945488825e-06,
+      "loss": 0.6606,
+      "step": 23093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.031733950029904e-06,
+      "loss": 0.7681,
+      "step": 23094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.030919627096483e-06,
+      "loss": 0.7584,
+      "step": 23095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.030105325754547e-06,
+      "loss": 0.6735,
+      "step": 23096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.029291046010014e-06,
+      "loss": 0.7882,
+      "step": 23097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.028476787868805e-06,
+      "loss": 0.8182,
+      "step": 23098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0276625513368445e-06,
+      "loss": 1.0931,
+      "step": 23099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.026848336420053e-06,
+      "loss": 0.6964,
+      "step": 23100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0260341431243495e-06,
+      "loss": 0.7063,
+      "step": 23101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.025219971455659e-06,
+      "loss": 0.5991,
+      "step": 23102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.024405821419896e-06,
+      "loss": 0.5698,
+      "step": 23103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.023591693022983e-06,
+      "loss": 0.8388,
+      "step": 23104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.022777586270844e-06,
+      "loss": 0.7682,
+      "step": 23105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.021963501169395e-06,
+      "loss": 0.996,
+      "step": 23106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.021149437724559e-06,
+      "loss": 0.9622,
+      "step": 23107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.020335395942255e-06,
+      "loss": 0.8396,
+      "step": 23108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.019521375828405e-06,
+      "loss": 0.7203,
+      "step": 23109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.018707377388925e-06,
+      "loss": 1.2316,
+      "step": 23110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.017893400629734e-06,
+      "loss": 1.0089,
+      "step": 23111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.017079445556759e-06,
+      "loss": 0.6918,
+      "step": 23112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.016265512175908e-06,
+      "loss": 0.7219,
+      "step": 23113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0154516004931085e-06,
+      "loss": 1.1748,
+      "step": 23114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.014637710514274e-06,
+      "loss": 0.7571,
+      "step": 23115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.013823842245326e-06,
+      "loss": 0.9203,
+      "step": 23116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.013009995692184e-06,
+      "loss": 0.8054,
+      "step": 23117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.012196170860763e-06,
+      "loss": 0.6165,
+      "step": 23118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.011382367756986e-06,
+      "loss": 1.1057,
+      "step": 23119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.010568586386766e-06,
+      "loss": 0.9253,
+      "step": 23120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.009754826756023e-06,
+      "loss": 0.9641,
+      "step": 23121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008941088870679e-06,
+      "loss": 0.4926,
+      "step": 23122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008127372736645e-06,
+      "loss": 1.033,
+      "step": 23123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0073136783598404e-06,
+      "loss": 0.9916,
+      "step": 23124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0065000057461815e-06,
+      "loss": 1.078,
+      "step": 23125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.005686354901589e-06,
+      "loss": 0.8925,
+      "step": 23126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004872725831978e-06,
+      "loss": 1.3136,
+      "step": 23127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004059118543265e-06,
+      "loss": 1.082,
+      "step": 23128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.003245533041365e-06,
+      "loss": 0.8696,
+      "step": 23129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.002431969332197e-06,
+      "loss": 0.8487,
+      "step": 23130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.001618427421679e-06,
+      "loss": 0.9215,
+      "step": 23131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.000804907315726e-06,
+      "loss": 0.711,
+      "step": 23132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99999140902025e-06,
+      "loss": 0.9258,
+      "step": 23133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.999177932541168e-06,
+      "loss": 0.8143,
+      "step": 23134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9983644778843985e-06,
+      "loss": 0.5738,
+      "step": 23135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.997551045055857e-06,
+      "loss": 0.5787,
+      "step": 23136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.996737634061457e-06,
+      "loss": 0.9537,
+      "step": 23137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.995924244907112e-06,
+      "loss": 0.8346,
+      "step": 23138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.995110877598742e-06,
+      "loss": 0.6666,
+      "step": 23139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99429753214226e-06,
+      "loss": 0.8882,
+      "step": 23140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.993484208543577e-06,
+      "loss": 0.7646,
+      "step": 23141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9926709068086176e-06,
+      "loss": 0.6113,
+      "step": 23142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.991857626943284e-06,
+      "loss": 0.9043,
+      "step": 23143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.991044368953495e-06,
+      "loss": 0.8029,
+      "step": 23144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.990231132845169e-06,
+      "loss": 0.8792,
+      "step": 23145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.989417918624214e-06,
+      "loss": 0.9902,
+      "step": 23146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9886047262965436e-06,
+      "loss": 0.933,
+      "step": 23147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9877915558680785e-06,
+      "loss": 0.7666,
+      "step": 23148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.986978407344727e-06,
+      "loss": 0.8329,
+      "step": 23149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9861652807324e-06,
+      "loss": 1.1746,
+      "step": 23150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.985352176037017e-06,
+      "loss": 0.6949,
+      "step": 23151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.984539093264489e-06,
+      "loss": 1.0162,
+      "step": 23152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.983726032420726e-06,
+      "loss": 1.0042,
+      "step": 23153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9829129935116415e-06,
+      "loss": 0.7929,
+      "step": 23154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.982099976543148e-06,
+      "loss": 0.7754,
+      "step": 23155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9812869815211595e-06,
+      "loss": 0.9047,
+      "step": 23156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.980474008451587e-06,
+      "loss": 0.8665,
+      "step": 23157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979661057340344e-06,
+      "loss": 0.7789,
+      "step": 23158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.978848128193339e-06,
+      "loss": 0.7554,
+      "step": 23159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.978035221016487e-06,
+      "loss": 1.1238,
+      "step": 23160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.977222335815699e-06,
+      "loss": 0.7519,
+      "step": 23161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9764094725968875e-06,
+      "loss": 0.8876,
+      "step": 23162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9755966313659605e-06,
+      "loss": 1.0606,
+      "step": 23163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.974783812128831e-06,
+      "loss": 0.6373,
+      "step": 23164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.973971014891408e-06,
+      "loss": 0.793,
+      "step": 23165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9731582396596034e-06,
+      "loss": 0.8273,
+      "step": 23166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.97234548643933e-06,
+      "loss": 0.9106,
+      "step": 23167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9715327552364946e-06,
+      "loss": 0.8628,
+      "step": 23168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.970720046057011e-06,
+      "loss": 1.0598,
+      "step": 23169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969907358906787e-06,
+      "loss": 0.9079,
+      "step": 23170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969094693791734e-06,
+      "loss": 1.0935,
+      "step": 23171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.968282050717763e-06,
+      "loss": 0.829,
+      "step": 23172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967469429690779e-06,
+      "loss": 0.8975,
+      "step": 23173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9666568307166916e-06,
+      "loss": 0.7906,
+      "step": 23174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965844253801416e-06,
+      "loss": 0.8687,
+      "step": 23175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965031698950857e-06,
+      "loss": 0.9743,
+      "step": 23176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964219166170922e-06,
+      "loss": 0.7352,
+      "step": 23177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.963406655467526e-06,
+      "loss": 1.0975,
+      "step": 23178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9625941668465724e-06,
+      "loss": 0.9017,
+      "step": 23179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.961781700313972e-06,
+      "loss": 0.7958,
+      "step": 23180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.96096925587563e-06,
+      "loss": 0.9082,
+      "step": 23181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.960156833537464e-06,
+      "loss": 0.9694,
+      "step": 23182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.959344433305368e-06,
+      "loss": 0.8013,
+      "step": 23183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.958532055185261e-06,
+      "loss": 1.0268,
+      "step": 23184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9577196991830445e-06,
+      "loss": 0.954,
+      "step": 23185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.956907365304626e-06,
+      "loss": 0.7252,
+      "step": 23186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.956095053555918e-06,
+      "loss": 1.0219,
+      "step": 23187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.955282763942825e-06,
+      "loss": 0.8083,
+      "step": 23188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9544704964712525e-06,
+      "loss": 0.7658,
+      "step": 23189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.953658251147109e-06,
+      "loss": 0.9766,
+      "step": 23190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.952846027976301e-06,
+      "loss": 0.8288,
+      "step": 23191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.952033826964739e-06,
+      "loss": 0.6503,
+      "step": 23192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.951221648118322e-06,
+      "loss": 0.7616,
+      "step": 23193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9504094914429595e-06,
+      "loss": 1.0302,
+      "step": 23194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.949597356944557e-06,
+      "loss": 0.7284,
+      "step": 23195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.948785244629022e-06,
+      "loss": 0.683,
+      "step": 23196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947973154502261e-06,
+      "loss": 0.7527,
+      "step": 23197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947161086570177e-06,
+      "loss": 1.2368,
+      "step": 23198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9463490408386735e-06,
+      "loss": 0.8194,
+      "step": 23199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.945537017313664e-06,
+      "loss": 0.7706,
+      "step": 23200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.944725016001046e-06,
+      "loss": 0.9511,
+      "step": 23201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.94391303690673e-06,
+      "loss": 0.7571,
+      "step": 23202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.943101080036615e-06,
+      "loss": 0.9168,
+      "step": 23203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9422891453966066e-06,
+      "loss": 0.8556,
+      "step": 23204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.941477232992614e-06,
+      "loss": 0.4339,
+      "step": 23205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.940665342830537e-06,
+      "loss": 0.9465,
+      "step": 23206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.939853474916283e-06,
+      "loss": 1.0073,
+      "step": 23207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9390416292557515e-06,
+      "loss": 0.6767,
+      "step": 23208
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.938229805854851e-06,
+      "loss": 0.9016,
+      "step": 23209
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.937418004719485e-06,
+      "loss": 0.8363,
+      "step": 23210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.936606225855551e-06,
+      "loss": 0.8128,
+      "step": 23211
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.935794469268964e-06,
+      "loss": 0.8993,
+      "step": 23212
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.934982734965614e-06,
+      "loss": 0.9864,
+      "step": 23213
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.934171022951412e-06,
+      "loss": 0.7655,
+      "step": 23214
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.933359333232258e-06,
+      "loss": 0.8073,
+      "step": 23215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.932547665814056e-06,
+      "loss": 0.8207,
+      "step": 23216
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.931736020702706e-06,
+      "loss": 0.8305,
+      "step": 23217
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.930924397904113e-06,
+      "loss": 0.9155,
+      "step": 23218
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93011279742418e-06,
+      "loss": 0.9422,
+      "step": 23219
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.929301219268806e-06,
+      "loss": 1.1414,
+      "step": 23220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.928489663443896e-06,
+      "loss": 0.5391,
+      "step": 23221
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.927678129955352e-06,
+      "loss": 1.0456,
+      "step": 23222
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.926866618809072e-06,
+      "loss": 0.9545,
+      "step": 23223
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92605513001096e-06,
+      "loss": 0.9274,
+      "step": 23224
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.925243663566915e-06,
+      "loss": 0.9583,
+      "step": 23225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.924432219482839e-06,
+      "loss": 0.8118,
+      "step": 23226
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9236207977646345e-06,
+      "loss": 0.8622,
+      "step": 23227
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.922809398418201e-06,
+      "loss": 0.83,
+      "step": 23228
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.921998021449438e-06,
+      "loss": 1.0848,
+      "step": 23229
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.921186666864248e-06,
+      "loss": 0.7483,
+      "step": 23230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.920375334668532e-06,
+      "loss": 0.965,
+      "step": 23231
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.91956402486819e-06,
+      "loss": 0.9478,
+      "step": 23232
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.918752737469118e-06,
+      "loss": 1.087,
+      "step": 23233
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917941472477218e-06,
+      "loss": 0.8232,
+      "step": 23234
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917130229898387e-06,
+      "loss": 1.1623,
+      "step": 23235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916319009738532e-06,
+      "loss": 1.0064,
+      "step": 23236
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.915507812003547e-06,
+      "loss": 0.9815,
+      "step": 23237
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.914696636699328e-06,
+      "loss": 0.8926,
+      "step": 23238
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913885483831781e-06,
+      "loss": 0.645,
+      "step": 23239
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9130743534068015e-06,
+      "loss": 0.8829,
+      "step": 23240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.912263245430289e-06,
+      "loss": 0.8656,
+      "step": 23241
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.911452159908141e-06,
+      "loss": 1.0588,
+      "step": 23242
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.910641096846255e-06,
+      "loss": 0.7911,
+      "step": 23243
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.909830056250527e-06,
+      "loss": 1.2039,
+      "step": 23244
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.909019038126861e-06,
+      "loss": 0.7489,
+      "step": 23245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9082080424811515e-06,
+      "loss": 0.9402,
+      "step": 23246
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9073970693192946e-06,
+      "loss": 1.1751,
+      "step": 23247
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9065861186471915e-06,
+      "loss": 0.7996,
+      "step": 23248
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.905775190470737e-06,
+      "loss": 0.8607,
+      "step": 23249
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.90496428479583e-06,
+      "loss": 0.9061,
+      "step": 23250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.904153401628369e-06,
+      "loss": 0.9099,
+      "step": 23251
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.903342540974244e-06,
+      "loss": 1.1779,
+      "step": 23252
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.902531702839354e-06,
+      "loss": 0.941,
+      "step": 23253
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.901720887229601e-06,
+      "loss": 0.8082,
+      "step": 23254
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.900910094150879e-06,
+      "loss": 0.5917,
+      "step": 23255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.900099323609079e-06,
+      "loss": 0.9631,
+      "step": 23256
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8992885756101025e-06,
+      "loss": 1.111,
+      "step": 23257
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8984778501598444e-06,
+      "loss": 0.8744,
+      "step": 23258
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8976671472642e-06,
+      "loss": 0.8495,
+      "step": 23259
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.896856466929062e-06,
+      "loss": 0.7764,
+      "step": 23260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.896045809160336e-06,
+      "loss": 0.9317,
+      "step": 23261
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.895235173963902e-06,
+      "loss": 0.9866,
+      "step": 23262
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.894424561345665e-06,
+      "loss": 0.9317,
+      "step": 23263
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.893613971311517e-06,
+      "loss": 0.5913,
+      "step": 23264
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.892803403867352e-06,
+      "loss": 1.0108,
+      "step": 23265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.891992859019067e-06,
+      "loss": 0.9335,
+      "step": 23266
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.891182336772556e-06,
+      "loss": 0.7657,
+      "step": 23267
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.890371837133711e-06,
+      "loss": 1.1259,
+      "step": 23268
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.889561360108427e-06,
+      "loss": 0.8669,
+      "step": 23269
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8887509057025995e-06,
+      "loss": 0.9066,
+      "step": 23270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.887940473922122e-06,
+      "loss": 0.9031,
+      "step": 23271
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.887130064772886e-06,
+      "loss": 0.7114,
+      "step": 23272
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.886319678260785e-06,
+      "loss": 1.0391,
+      "step": 23273
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.885509314391712e-06,
+      "loss": 0.548,
+      "step": 23274
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.884698973171562e-06,
+      "loss": 1.008,
+      "step": 23275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.883888654606227e-06,
+      "loss": 0.9968,
+      "step": 23276
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8830783587016015e-06,
+      "loss": 0.6404,
+      "step": 23277
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.882268085463573e-06,
+      "loss": 0.7504,
+      "step": 23278
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88145783489804e-06,
+      "loss": 0.8121,
+      "step": 23279
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8806476070108905e-06,
+      "loss": 0.953,
+      "step": 23280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8798374018080205e-06,
+      "loss": 0.849,
+      "step": 23281
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8790272192953175e-06,
+      "loss": 1.0983,
+      "step": 23282
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.878217059478673e-06,
+      "loss": 0.6189,
+      "step": 23283
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8774069223639835e-06,
+      "loss": 1.0534,
+      "step": 23284
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.876596807957137e-06,
+      "loss": 0.8263,
+      "step": 23285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.875786716264026e-06,
+      "loss": 0.9483,
+      "step": 23286
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.874976647290539e-06,
+      "loss": 0.7708,
+      "step": 23287
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.87416660104257e-06,
+      "loss": 0.7442,
+      "step": 23288
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.873356577526009e-06,
+      "loss": 0.7347,
+      "step": 23289
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.872546576746744e-06,
+      "loss": 0.9701,
+      "step": 23290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.871736598710674e-06,
+      "loss": 0.7733,
+      "step": 23291
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8709266434236765e-06,
+      "loss": 0.8556,
+      "step": 23292
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.87011671089165e-06,
+      "loss": 1.044,
+      "step": 23293
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.869306801120481e-06,
+      "loss": 0.6611,
+      "step": 23294
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.868496914116063e-06,
+      "loss": 0.7579,
+      "step": 23295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.86768704988428e-06,
+      "loss": 1.0436,
+      "step": 23296
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.866877208431027e-06,
+      "loss": 0.8595,
+      "step": 23297
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.86606738976219e-06,
+      "loss": 0.9333,
+      "step": 23298
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.865257593883658e-06,
+      "loss": 0.9248,
+      "step": 23299
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.864447820801321e-06,
+      "loss": 1.0683,
+      "step": 23300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.863638070521072e-06,
+      "loss": 1.2113,
+      "step": 23301
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.862828343048792e-06,
+      "loss": 0.7465,
+      "step": 23302
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.862018638390373e-06,
+      "loss": 0.9133,
+      "step": 23303
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8612089565517026e-06,
+      "loss": 0.9446,
+      "step": 23304
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.860399297538667e-06,
+      "loss": 0.7526,
+      "step": 23305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.85958966135716e-06,
+      "loss": 1.0227,
+      "step": 23306
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.858780048013064e-06,
+      "loss": 0.8989,
+      "step": 23307
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.857970457512266e-06,
+      "loss": 0.8276,
+      "step": 23308
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.857160889860659e-06,
+      "loss": 0.8865,
+      "step": 23309
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.856351345064127e-06,
+      "loss": 0.9577,
+      "step": 23310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.855541823128559e-06,
+      "loss": 0.8079,
+      "step": 23311
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8547323240598375e-06,
+      "loss": 1.268,
+      "step": 23312
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853922847863852e-06,
+      "loss": 0.5257,
+      "step": 23313
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853113394546487e-06,
+      "loss": 1.0513,
+      "step": 23314
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.852303964113634e-06,
+      "loss": 0.9139,
+      "step": 23315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.851494556571174e-06,
+      "loss": 0.5579,
+      "step": 23316
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.850685171924993e-06,
+      "loss": 0.9787,
+      "step": 23317
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849875810180983e-06,
+      "loss": 0.7257,
+      "step": 23318
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849066471345025e-06,
+      "loss": 0.6199,
+      "step": 23319
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8482571554230056e-06,
+      "loss": 1.5055,
+      "step": 23320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.847447862420813e-06,
+      "loss": 1.0043,
+      "step": 23321
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8466385923443266e-06,
+      "loss": 0.5733,
+      "step": 23322
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845829345199433e-06,
+      "loss": 0.8748,
+      "step": 23323
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845020120992021e-06,
+      "loss": 0.8981,
+      "step": 23324
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.844210919727971e-06,
+      "loss": 0.7583,
+      "step": 23325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.843401741413171e-06,
+      "loss": 0.9277,
+      "step": 23326
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.842592586053504e-06,
+      "loss": 0.8831,
+      "step": 23327
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.841783453654854e-06,
+      "loss": 0.7218,
+      "step": 23328
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8409743442231054e-06,
+      "loss": 0.652,
+      "step": 23329
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.840165257764142e-06,
+      "loss": 1.1207,
+      "step": 23330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.839356194283853e-06,
+      "loss": 1.252,
+      "step": 23331
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83854715378811e-06,
+      "loss": 0.7886,
+      "step": 23332
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.837738136282806e-06,
+      "loss": 0.9021,
+      "step": 23333
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.836929141773821e-06,
+      "loss": 0.8204,
+      "step": 23334
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.836120170267035e-06,
+      "loss": 1.1309,
+      "step": 23335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.835311221768338e-06,
+      "loss": 1.1127,
+      "step": 23336
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83450229628361e-06,
+      "loss": 1.0187,
+      "step": 23337
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833693393818731e-06,
+      "loss": 0.778,
+      "step": 23338
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.832884514379585e-06,
+      "loss": 0.9707,
+      "step": 23339
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8320756579720545e-06,
+      "loss": 0.8616,
+      "step": 23340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.831266824602025e-06,
+      "loss": 1.1502,
+      "step": 23341
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.830458014275374e-06,
+      "loss": 0.9215,
+      "step": 23342
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.829649226997983e-06,
+      "loss": 1.3147,
+      "step": 23343
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8288404627757324e-06,
+      "loss": 0.9155,
+      "step": 23344
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.828031721614509e-06,
+      "loss": 0.9108,
+      "step": 23345
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.827223003520191e-06,
+      "loss": 0.7603,
+      "step": 23346
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.826414308498659e-06,
+      "loss": 1.2866,
+      "step": 23347
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.825605636555793e-06,
+      "loss": 0.7755,
+      "step": 23348
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824796987697477e-06,
+      "loss": 0.7142,
+      "step": 23349
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.823988361929591e-06,
+      "loss": 0.8977,
+      "step": 23350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8231797592580165e-06,
+      "loss": 1.0575,
+      "step": 23351
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8223711796886275e-06,
+      "loss": 0.9319,
+      "step": 23352
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.821562623227307e-06,
+      "loss": 0.8863,
+      "step": 23353
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.82075408987994e-06,
+      "loss": 0.9473,
+      "step": 23354
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.819945579652401e-06,
+      "loss": 0.7487,
+      "step": 23355
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.81913709255057e-06,
+      "loss": 1.1079,
+      "step": 23356
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.818328628580325e-06,
+      "loss": 0.9349,
+      "step": 23357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.817520187747551e-06,
+      "loss": 0.6692,
+      "step": 23358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.816711770058122e-06,
+      "loss": 0.9184,
+      "step": 23359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815903375517919e-06,
+      "loss": 0.976,
+      "step": 23360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8150950041328235e-06,
+      "loss": 0.6726,
+      "step": 23361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.814286655908706e-06,
+      "loss": 0.898,
+      "step": 23362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.813478330851453e-06,
+      "loss": 0.9999,
+      "step": 23363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8126700289669365e-06,
+      "loss": 0.731,
+      "step": 23364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.81186175026104e-06,
+      "loss": 0.7362,
+      "step": 23365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.811053494739637e-06,
+      "loss": 0.9454,
+      "step": 23366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8102452624086075e-06,
+      "loss": 0.9437,
+      "step": 23367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.809437053273829e-06,
+      "loss": 0.8236,
+      "step": 23368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.808628867341177e-06,
+      "loss": 0.9834,
+      "step": 23369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.807820704616532e-06,
+      "loss": 0.9513,
+      "step": 23370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8070125651057715e-06,
+      "loss": 0.7177,
+      "step": 23371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.806204448814768e-06,
+      "loss": 0.9164,
+      "step": 23372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.805396355749401e-06,
+      "loss": 0.8925,
+      "step": 23373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.804588285915546e-06,
+      "loss": 1.0664,
+      "step": 23374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.803780239319078e-06,
+      "loss": 1.2212,
+      "step": 23375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.802972215965877e-06,
+      "loss": 0.7946,
+      "step": 23376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.802164215861817e-06,
+      "loss": 0.9994,
+      "step": 23377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8013562390127715e-06,
+      "loss": 0.8655,
+      "step": 23378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.800548285424622e-06,
+      "loss": 0.6948,
+      "step": 23379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.799740355103239e-06,
+      "loss": 0.7976,
+      "step": 23380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7989324480545025e-06,
+      "loss": 1.088,
+      "step": 23381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.798124564284283e-06,
+      "loss": 0.6965,
+      "step": 23382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.797316703798456e-06,
+      "loss": 0.8087,
+      "step": 23383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796508866602896e-06,
+      "loss": 0.7562,
+      "step": 23384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.795701052703482e-06,
+      "loss": 1.123,
+      "step": 23385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.794893262106084e-06,
+      "loss": 0.9274,
+      "step": 23386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.794085494816578e-06,
+      "loss": 0.8484,
+      "step": 23387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.79327775084084e-06,
+      "loss": 0.9296,
+      "step": 23388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.792470030184741e-06,
+      "loss": 0.8428,
+      "step": 23389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.791662332854159e-06,
+      "loss": 0.7628,
+      "step": 23390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790854658854964e-06,
+      "loss": 1.0507,
+      "step": 23391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790047008193029e-06,
+      "loss": 0.6951,
+      "step": 23392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.789239380874226e-06,
+      "loss": 0.7724,
+      "step": 23393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.788431776904434e-06,
+      "loss": 1.0438,
+      "step": 23394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.787624196289523e-06,
+      "loss": 1.117,
+      "step": 23395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.786816639035363e-06,
+      "loss": 0.8372,
+      "step": 23396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78600910514783e-06,
+      "loss": 0.8898,
+      "step": 23397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7852015946327975e-06,
+      "loss": 0.9035,
+      "step": 23398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.784394107496135e-06,
+      "loss": 0.9704,
+      "step": 23399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.783586643743714e-06,
+      "loss": 0.9523,
+      "step": 23400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.782779203381414e-06,
+      "loss": 0.9763,
+      "step": 23401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.781971786415096e-06,
+      "loss": 0.8258,
+      "step": 23402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.781164392850636e-06,
+      "loss": 0.7101,
+      "step": 23403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7803570226939085e-06,
+      "loss": 0.9094,
+      "step": 23404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.779549675950781e-06,
+      "loss": 0.7899,
+      "step": 23405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.778742352627123e-06,
+      "loss": 0.9469,
+      "step": 23406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.777935052728812e-06,
+      "loss": 0.9477,
+      "step": 23407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.777127776261715e-06,
+      "loss": 0.79,
+      "step": 23408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7763205232317e-06,
+      "loss": 0.9195,
+      "step": 23409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7755132936446425e-06,
+      "loss": 0.9928,
+      "step": 23410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.774706087506412e-06,
+      "loss": 0.7473,
+      "step": 23411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7738989048228755e-06,
+      "loss": 1.0594,
+      "step": 23412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.773091745599905e-06,
+      "loss": 0.8649,
+      "step": 23413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.772284609843367e-06,
+      "loss": 0.7393,
+      "step": 23414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7714774975591335e-06,
+      "loss": 1.1085,
+      "step": 23415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.770670408753077e-06,
+      "loss": 0.8882,
+      "step": 23416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7698633434310626e-06,
+      "loss": 0.6186,
+      "step": 23417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76905630159896e-06,
+      "loss": 0.9066,
+      "step": 23418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76824928326264e-06,
+      "loss": 0.9483,
+      "step": 23419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76744228842797e-06,
+      "loss": 0.9665,
+      "step": 23420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7666353171008205e-06,
+      "loss": 0.6974,
+      "step": 23421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7658283692870565e-06,
+      "loss": 0.9739,
+      "step": 23422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.765021444992548e-06,
+      "loss": 0.9377,
+      "step": 23423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.76421454422316e-06,
+      "loss": 0.7433,
+      "step": 23424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.763407666984765e-06,
+      "loss": 0.5648,
+      "step": 23425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.762600813283229e-06,
+      "loss": 0.666,
+      "step": 23426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.761793983124417e-06,
+      "loss": 0.5426,
+      "step": 23427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.760987176514202e-06,
+      "loss": 0.9384,
+      "step": 23428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.760180393458447e-06,
+      "loss": 0.7152,
+      "step": 23429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75937363396302e-06,
+      "loss": 1.0174,
+      "step": 23430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.758566898033791e-06,
+      "loss": 0.8702,
+      "step": 23431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.757760185676621e-06,
+      "loss": 0.9856,
+      "step": 23432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.756953496897376e-06,
+      "loss": 0.8337,
+      "step": 23433
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7561468317019275e-06,
+      "loss": 0.8068,
+      "step": 23434
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75534019009614e-06,
+      "loss": 0.8047,
+      "step": 23435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.754533572085876e-06,
+      "loss": 0.9245,
+      "step": 23436
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.753726977677007e-06,
+      "loss": 0.8434,
+      "step": 23437
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7529204068753966e-06,
+      "loss": 1.0514,
+      "step": 23438
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.752113859686909e-06,
+      "loss": 0.9247,
+      "step": 23439
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.751307336117408e-06,
+      "loss": 0.5468,
+      "step": 23440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.750500836172767e-06,
+      "loss": 0.8561,
+      "step": 23441
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.749694359858838e-06,
+      "loss": 0.9264,
+      "step": 23442
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.748887907181495e-06,
+      "loss": 0.827,
+      "step": 23443
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7480814781466e-06,
+      "loss": 0.789,
+      "step": 23444
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7472750727600155e-06,
+      "loss": 0.74,
+      "step": 23445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.746468691027611e-06,
+      "loss": 1.0309,
+      "step": 23446
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.745662332955247e-06,
+      "loss": 0.7676,
+      "step": 23447
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744855998548787e-06,
+      "loss": 0.6588,
+      "step": 23448
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744049687814094e-06,
+      "loss": 1.1197,
+      "step": 23449
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.743243400757036e-06,
+      "loss": 0.7997,
+      "step": 23450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.742437137383475e-06,
+      "loss": 0.9319,
+      "step": 23451
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.741630897699272e-06,
+      "loss": 0.9191,
+      "step": 23452
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.74082468171029e-06,
+      "loss": 1.0688,
+      "step": 23453
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.740018489422391e-06,
+      "loss": 0.7722,
+      "step": 23454
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.73921232084144e-06,
+      "loss": 0.8583,
+      "step": 23455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.738406175973301e-06,
+      "loss": 1.016,
+      "step": 23456
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7376000548238345e-06,
+      "loss": 0.6301,
+      "step": 23457
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7367939573989e-06,
+      "loss": 0.7235,
+      "step": 23458
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.735987883704364e-06,
+      "loss": 0.762,
+      "step": 23459
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.735181833746087e-06,
+      "loss": 0.9464,
+      "step": 23460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.734375807529931e-06,
+      "loss": 1.2722,
+      "step": 23461
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.733569805061757e-06,
+      "loss": 0.6092,
+      "step": 23462
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732763826347424e-06,
+      "loss": 0.8064,
+      "step": 23463
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.731957871392796e-06,
+      "loss": 1.0286,
+      "step": 23464
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.731151940203733e-06,
+      "loss": 0.6962,
+      "step": 23465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.730346032786097e-06,
+      "loss": 0.7295,
+      "step": 23466
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.729540149145746e-06,
+      "loss": 0.9048,
+      "step": 23467
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.728734289288543e-06,
+      "loss": 0.9106,
+      "step": 23468
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.727928453220347e-06,
+      "loss": 0.7915,
+      "step": 23469
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72712264094702e-06,
+      "loss": 0.7983,
+      "step": 23470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.726316852474424e-06,
+      "loss": 0.832,
+      "step": 23471
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.725511087808409e-06,
+      "loss": 0.8155,
+      "step": 23472
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.724705346954843e-06,
+      "loss": 0.8597,
+      "step": 23473
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.723899629919585e-06,
+      "loss": 1.0354,
+      "step": 23474
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7230939367084915e-06,
+      "loss": 0.9983,
+      "step": 23475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.722288267327421e-06,
+      "loss": 0.8597,
+      "step": 23476
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.721482621782236e-06,
+      "loss": 0.7427,
+      "step": 23477
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.720677000078795e-06,
+      "loss": 0.877,
+      "step": 23478
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.71987140222295e-06,
+      "loss": 1.0302,
+      "step": 23479
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.71906582822057e-06,
+      "loss": 0.9838,
+      "step": 23480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7182602780775076e-06,
+      "loss": 1.0319,
+      "step": 23481
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.717454751799619e-06,
+      "loss": 0.8786,
+      "step": 23482
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.716649249392764e-06,
+      "loss": 0.8109,
+      "step": 23483
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7158437708628e-06,
+      "loss": 0.7851,
+      "step": 23484
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.715038316215583e-06,
+      "loss": 0.6314,
+      "step": 23485
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.714232885456975e-06,
+      "loss": 0.6862,
+      "step": 23486
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.71342747859283e-06,
+      "loss": 0.8337,
+      "step": 23487
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7126220956290025e-06,
+      "loss": 0.8555,
+      "step": 23488
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.711816736571355e-06,
+      "loss": 0.9016,
+      "step": 23489
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.711011401425741e-06,
+      "loss": 1.1198,
+      "step": 23490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.710206090198019e-06,
+      "loss": 1.0861,
+      "step": 23491
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.709400802894043e-06,
+      "loss": 1.1331,
+      "step": 23492
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.708595539519669e-06,
+      "loss": 0.6231,
+      "step": 23493
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70779030008075e-06,
+      "loss": 0.9132,
+      "step": 23494
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70698508458315e-06,
+      "loss": 0.7573,
+      "step": 23495
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.706179893032718e-06,
+      "loss": 0.7772,
+      "step": 23496
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.705374725435312e-06,
+      "loss": 0.7007,
+      "step": 23497
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.704569581796787e-06,
+      "loss": 0.7732,
+      "step": 23498
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7037644621229995e-06,
+      "loss": 0.9638,
+      "step": 23499
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.702959366419801e-06,
+      "loss": 0.8141,
+      "step": 23500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70215429469305e-06,
+      "loss": 1.2007,
+      "step": 23501
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7013492469486e-06,
+      "loss": 1.0979,
+      "step": 23502
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7005442231923e-06,
+      "loss": 0.7375,
+      "step": 23503
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699739223430012e-06,
+      "loss": 0.7896,
+      "step": 23504
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.698934247667587e-06,
+      "loss": 0.7819,
+      "step": 23505
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.698129295910876e-06,
+      "loss": 0.9332,
+      "step": 23506
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.697324368165738e-06,
+      "loss": 0.668,
+      "step": 23507
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.696519464438022e-06,
+      "loss": 1.1112,
+      "step": 23508
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.695714584733586e-06,
+      "loss": 0.7998,
+      "step": 23509
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.694909729058277e-06,
+      "loss": 0.6709,
+      "step": 23510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.694104897417958e-06,
+      "loss": 1.0531,
+      "step": 23511
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.693300089818468e-06,
+      "loss": 1.1233,
+      "step": 23512
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6924953062656715e-06,
+      "loss": 0.7858,
+      "step": 23513
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.691690546765415e-06,
+      "loss": 0.7293,
+      "step": 23514
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.69088581132355e-06,
+      "loss": 0.7321,
+      "step": 23515
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.690081099945933e-06,
+      "loss": 0.9491,
+      "step": 23516
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.689276412638414e-06,
+      "loss": 0.7107,
+      "step": 23517
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.688471749406844e-06,
+      "loss": 0.6941,
+      "step": 23518
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6876671102570735e-06,
+      "loss": 0.8119,
+      "step": 23519
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.686862495194958e-06,
+      "loss": 1.0517,
+      "step": 23520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.686057904226348e-06,
+      "loss": 0.8925,
+      "step": 23521
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.685253337357092e-06,
+      "loss": 0.7825,
+      "step": 23522
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68444879459304e-06,
+      "loss": 0.867,
+      "step": 23523
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6836442759400436e-06,
+      "loss": 0.8033,
+      "step": 23524
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.682839781403956e-06,
+      "loss": 0.8653,
+      "step": 23525
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6820353109906255e-06,
+      "loss": 0.722,
+      "step": 23526
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6812308647059035e-06,
+      "loss": 0.7782,
+      "step": 23527
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.680426442555636e-06,
+      "loss": 0.6972,
+      "step": 23528
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679622044545679e-06,
+      "loss": 0.9415,
+      "step": 23529
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.678817670681879e-06,
+      "loss": 0.6461,
+      "step": 23530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6780133209700885e-06,
+      "loss": 1.0208,
+      "step": 23531
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6772089954161514e-06,
+      "loss": 0.8531,
+      "step": 23532
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.676404694025916e-06,
+      "loss": 0.9565,
+      "step": 23533
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.675600416805239e-06,
+      "loss": 0.9554,
+      "step": 23534
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6747961637599645e-06,
+      "loss": 0.6588,
+      "step": 23535
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.67399193489594e-06,
+      "loss": 1.0298,
+      "step": 23536
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.673187730219015e-06,
+      "loss": 0.8004,
+      "step": 23537
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.67238354973504e-06,
+      "loss": 1.2191,
+      "step": 23538
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.671579393449861e-06,
+      "loss": 1.0867,
+      "step": 23539
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.670775261369325e-06,
+      "loss": 0.4795,
+      "step": 23540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6699711534992865e-06,
+      "loss": 0.6864,
+      "step": 23541
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6691670698455815e-06,
+      "loss": 1.0325,
+      "step": 23542
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.668363010414065e-06,
+      "loss": 0.789,
+      "step": 23543
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.667558975210583e-06,
+      "loss": 0.6515,
+      "step": 23544
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6667549642409825e-06,
+      "loss": 1.0556,
+      "step": 23545
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.665950977511108e-06,
+      "loss": 0.8189,
+      "step": 23546
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.665147015026809e-06,
+      "loss": 0.8356,
+      "step": 23547
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.664343076793933e-06,
+      "loss": 0.7422,
+      "step": 23548
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.663539162818322e-06,
+      "loss": 0.8267,
+      "step": 23549
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662735273105827e-06,
+      "loss": 1.1822,
+      "step": 23550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.661931407662292e-06,
+      "loss": 0.7751,
+      "step": 23551
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.661127566493563e-06,
+      "loss": 0.972,
+      "step": 23552
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.660323749605484e-06,
+      "loss": 0.9673,
+      "step": 23553
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.659519957003901e-06,
+      "loss": 1.0136,
+      "step": 23554
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.658716188694659e-06,
+      "loss": 0.7848,
+      "step": 23555
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.657912444683604e-06,
+      "loss": 1.0481,
+      "step": 23556
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.657108724976582e-06,
+      "loss": 0.899,
+      "step": 23557
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.656305029579435e-06,
+      "loss": 0.6668,
+      "step": 23558
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6555013584980114e-06,
+      "loss": 0.8971,
+      "step": 23559
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.654697711738152e-06,
+      "loss": 0.9626,
+      "step": 23560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.653894089305704e-06,
+      "loss": 1.0861,
+      "step": 23561
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6530904912065085e-06,
+      "loss": 0.7371,
+      "step": 23562
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6522869174464115e-06,
+      "loss": 0.975,
+      "step": 23563
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6514833680312525e-06,
+      "loss": 1.0735,
+      "step": 23564
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.650679842966881e-06,
+      "loss": 0.5726,
+      "step": 23565
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.649876342259138e-06,
+      "loss": 0.791,
+      "step": 23566
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6490728659138636e-06,
+      "loss": 0.6933,
+      "step": 23567
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.648269413936905e-06,
+      "loss": 0.7953,
+      "step": 23568
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.647465986334105e-06,
+      "loss": 0.6307,
+      "step": 23569
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.646662583111303e-06,
+      "loss": 1.1217,
+      "step": 23570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.645859204274346e-06,
+      "loss": 0.6452,
+      "step": 23571
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.645055849829072e-06,
+      "loss": 0.9273,
+      "step": 23572
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644252519781321e-06,
+      "loss": 0.7133,
+      "step": 23573
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.643449214136941e-06,
+      "loss": 0.8932,
+      "step": 23574
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.64264593290177e-06,
+      "loss": 0.9671,
+      "step": 23575
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6418426760816505e-06,
+      "loss": 0.9823,
+      "step": 23576
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6410394436824244e-06,
+      "loss": 0.7322,
+      "step": 23577
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.640236235709934e-06,
+      "loss": 0.8421,
+      "step": 23578
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.639433052170017e-06,
+      "loss": 0.7337,
+      "step": 23579
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.638629893068516e-06,
+      "loss": 1.1144,
+      "step": 23580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.637826758411275e-06,
+      "loss": 0.939,
+      "step": 23581
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.637023648204127e-06,
+      "loss": 0.7413,
+      "step": 23582
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.636220562452918e-06,
+      "loss": 0.8298,
+      "step": 23583
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6354175011634866e-06,
+      "loss": 0.7842,
+      "step": 23584
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.634614464341671e-06,
+      "loss": 0.8189,
+      "step": 23585
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.633811451993315e-06,
+      "loss": 0.8349,
+      "step": 23586
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6330084641242556e-06,
+      "loss": 0.6196,
+      "step": 23587
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.632205500740332e-06,
+      "loss": 1.0421,
+      "step": 23588
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.631402561847383e-06,
+      "loss": 0.7079,
+      "step": 23589
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.63059964745125e-06,
+      "loss": 0.9347,
+      "step": 23590
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.629796757557773e-06,
+      "loss": 0.9911,
+      "step": 23591
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.628993892172786e-06,
+      "loss": 0.8681,
+      "step": 23592
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.62819105130213e-06,
+      "loss": 0.9172,
+      "step": 23593
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.62738823495164e-06,
+      "loss": 0.6986,
+      "step": 23594
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.62658544312716e-06,
+      "loss": 0.7932,
+      "step": 23595
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.625782675834526e-06,
+      "loss": 0.8896,
+      "step": 23596
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624979933079574e-06,
+      "loss": 0.5896,
+      "step": 23597
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624177214868142e-06,
+      "loss": 0.7067,
+      "step": 23598
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.623374521206069e-06,
+      "loss": 0.5901,
+      "step": 23599
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.622571852099191e-06,
+      "loss": 0.6439,
+      "step": 23600
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.621769207553349e-06,
+      "loss": 0.7264,
+      "step": 23601
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.620966587574374e-06,
+      "loss": 0.9927,
+      "step": 23602
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6201639921681026e-06,
+      "loss": 0.9307,
+      "step": 23603
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.619361421340375e-06,
+      "loss": 0.759,
+      "step": 23604
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.61855887509703e-06,
+      "loss": 0.7693,
+      "step": 23605
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6177563534438986e-06,
+      "loss": 0.6414,
+      "step": 23606
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616953856386815e-06,
+      "loss": 1.1832,
+      "step": 23607
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616151383931622e-06,
+      "loss": 0.6266,
+      "step": 23608
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6153489360841515e-06,
+      "loss": 0.7528,
+      "step": 23609
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.614546512850237e-06,
+      "loss": 1.1419,
+      "step": 23610
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.613744114235724e-06,
+      "loss": 0.6981,
+      "step": 23611
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.612941740246432e-06,
+      "loss": 0.873,
+      "step": 23612
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.612139390888207e-06,
+      "loss": 0.726,
+      "step": 23613
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.611337066166881e-06,
+      "loss": 0.8964,
+      "step": 23614
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.610534766088286e-06,
+      "loss": 1.1989,
+      "step": 23615
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6097324906582584e-06,
+      "loss": 0.865,
+      "step": 23616
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.608930239882634e-06,
+      "loss": 0.9567,
+      "step": 23617
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.608128013767245e-06,
+      "loss": 1.0472,
+      "step": 23618
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.607325812317924e-06,
+      "loss": 0.7557,
+      "step": 23619
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.606523635540509e-06,
+      "loss": 0.9923,
+      "step": 23620
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.605721483440833e-06,
+      "loss": 0.9189,
+      "step": 23621
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.604919356024724e-06,
+      "loss": 0.879,
+      "step": 23622
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.604117253298019e-06,
+      "loss": 0.5511,
+      "step": 23623
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6033151752665516e-06,
+      "loss": 0.6714,
+      "step": 23624
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6025131219361505e-06,
+      "loss": 0.767,
+      "step": 23625
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6017110933126536e-06,
+      "loss": 0.8288,
+      "step": 23626
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60090908940189e-06,
+      "loss": 1.0626,
+      "step": 23627
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.600107110209693e-06,
+      "loss": 1.0079,
+      "step": 23628
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.599305155741896e-06,
+      "loss": 1.0179,
+      "step": 23629
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.598503226004328e-06,
+      "loss": 0.636,
+      "step": 23630
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.597701321002825e-06,
+      "loss": 0.7576,
+      "step": 23631
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596899440743214e-06,
+      "loss": 0.7003,
+      "step": 23632
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596097585231329e-06,
+      "loss": 0.7217,
+      "step": 23633
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.595295754472997e-06,
+      "loss": 0.7792,
+      "step": 23634
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5944939484740545e-06,
+      "loss": 1.0492,
+      "step": 23635
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5936921672403305e-06,
+      "loss": 0.791,
+      "step": 23636
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.592890410777654e-06,
+      "loss": 0.8132,
+      "step": 23637
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.592088679091858e-06,
+      "loss": 0.9728,
+      "step": 23638
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.591286972188772e-06,
+      "loss": 0.8995,
+      "step": 23639
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.590485290074224e-06,
+      "loss": 0.9247,
+      "step": 23640
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5896836327540504e-06,
+      "loss": 0.8817,
+      "step": 23641
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.588882000234073e-06,
+      "loss": 0.5757,
+      "step": 23642
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.588080392520122e-06,
+      "loss": 0.7887,
+      "step": 23643
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.587278809618032e-06,
+      "loss": 0.6683,
+      "step": 23644
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.586477251533628e-06,
+      "loss": 0.4725,
+      "step": 23645
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.585675718272739e-06,
+      "loss": 0.7274,
+      "step": 23646
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.584874209841199e-06,
+      "loss": 0.8523,
+      "step": 23647
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.584072726244831e-06,
+      "loss": 0.7059,
+      "step": 23648
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.583271267489466e-06,
+      "loss": 0.936,
+      "step": 23649
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.582469833580931e-06,
+      "loss": 0.9392,
+      "step": 23650
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5816684245250595e-06,
+      "loss": 0.8697,
+      "step": 23651
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5808670403276694e-06,
+      "loss": 0.7045,
+      "step": 23652
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.580065680994596e-06,
+      "loss": 0.7534,
+      "step": 23653
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579264346531664e-06,
+      "loss": 0.7081,
+      "step": 23654
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5784630369447e-06,
+      "loss": 0.9917,
+      "step": 23655
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.577661752239535e-06,
+      "loss": 0.8209,
+      "step": 23656
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.576860492421993e-06,
+      "loss": 1.1409,
+      "step": 23657
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.576059257497902e-06,
+      "loss": 0.737,
+      "step": 23658
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.575258047473087e-06,
+      "loss": 0.8905,
+      "step": 23659
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.574456862353377e-06,
+      "loss": 0.9422,
+      "step": 23660
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.573655702144599e-06,
+      "loss": 0.8142,
+      "step": 23661
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.572854566852576e-06,
+      "loss": 0.8931,
+      "step": 23662
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.572053456483134e-06,
+      "loss": 0.7996,
+      "step": 23663
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.571252371042099e-06,
+      "loss": 1.216,
+      "step": 23664
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.570451310535299e-06,
+      "loss": 1.1711,
+      "step": 23665
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.569650274968557e-06,
+      "loss": 0.8864,
+      "step": 23666
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.568849264347701e-06,
+      "loss": 0.8712,
+      "step": 23667
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.568048278678552e-06,
+      "loss": 0.8037,
+      "step": 23668
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.567247317966938e-06,
+      "loss": 0.7493,
+      "step": 23669
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.566446382218683e-06,
+      "loss": 0.8571,
+      "step": 23670
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.565645471439613e-06,
+      "loss": 0.7338,
+      "step": 23671
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.564844585635549e-06,
+      "loss": 0.8229,
+      "step": 23672
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.564043724812315e-06,
+      "loss": 0.7228,
+      "step": 23673
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.563242888975737e-06,
+      "loss": 0.9902,
+      "step": 23674
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.562442078131639e-06,
+      "loss": 0.8564,
+      "step": 23675
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.561641292285845e-06,
+      "loss": 0.6943,
+      "step": 23676
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5608405314441744e-06,
+      "loss": 1.0616,
+      "step": 23677
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.560039795612456e-06,
+      "loss": 0.8752,
+      "step": 23678
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.55923908479651e-06,
+      "loss": 0.7542,
+      "step": 23679
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.558438399002156e-06,
+      "loss": 1.0179,
+      "step": 23680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.557637738235229e-06,
+      "loss": 0.7258,
+      "step": 23681
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.556837102501534e-06,
+      "loss": 0.6412,
+      "step": 23682
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5560364918069055e-06,
+      "loss": 0.8895,
+      "step": 23683
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.555235906157161e-06,
+      "loss": 0.7498,
+      "step": 23684
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5544353455581245e-06,
+      "loss": 0.895,
+      "step": 23685
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.553634810015614e-06,
+      "loss": 1.0099,
+      "step": 23686
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.552834299535456e-06,
+      "loss": 1.0131,
+      "step": 23687
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5520338141234705e-06,
+      "loss": 1.0567,
+      "step": 23688
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.551233353785476e-06,
+      "loss": 0.7469,
+      "step": 23689
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.550432918527296e-06,
+      "loss": 1.0092,
+      "step": 23690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.549632508354755e-06,
+      "loss": 0.86,
+      "step": 23691
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.548832123273665e-06,
+      "loss": 0.8356,
+      "step": 23692
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.548031763289853e-06,
+      "loss": 0.878,
+      "step": 23693
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.547231428409138e-06,
+      "loss": 0.8276,
+      "step": 23694
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.546431118637336e-06,
+      "loss": 1.0172,
+      "step": 23695
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.545630833980272e-06,
+      "loss": 0.7827,
+      "step": 23696
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544830574443765e-06,
+      "loss": 1.1836,
+      "step": 23697
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544030340033631e-06,
+      "loss": 0.8092,
+      "step": 23698
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.543230130755696e-06,
+      "loss": 0.9765,
+      "step": 23699
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.542429946615774e-06,
+      "loss": 0.811,
+      "step": 23700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5416297876196875e-06,
+      "loss": 0.8486,
+      "step": 23701
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5408296537732505e-06,
+      "loss": 1.0419,
+      "step": 23702
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.540029545082286e-06,
+      "loss": 0.7153,
+      "step": 23703
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.539229461552608e-06,
+      "loss": 0.643,
+      "step": 23704
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.538429403190039e-06,
+      "loss": 0.7228,
+      "step": 23705
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.537629370000397e-06,
+      "loss": 0.7876,
+      "step": 23706
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.536829361989496e-06,
+      "loss": 0.9077,
+      "step": 23707
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.536029379163158e-06,
+      "loss": 0.9401,
+      "step": 23708
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5352294215272005e-06,
+      "loss": 0.7893,
+      "step": 23709
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5344294890874386e-06,
+      "loss": 1.0984,
+      "step": 23710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.533629581849692e-06,
+      "loss": 0.8586,
+      "step": 23711
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5328296998197735e-06,
+      "loss": 0.9903,
+      "step": 23712
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.532029843003502e-06,
+      "loss": 1.0544,
+      "step": 23713
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.531230011406695e-06,
+      "loss": 0.9211,
+      "step": 23714
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.53043020503517e-06,
+      "loss": 0.974,
+      "step": 23715
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.52963042389474e-06,
+      "loss": 0.9334,
+      "step": 23716
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.528830667991224e-06,
+      "loss": 0.8624,
+      "step": 23717
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.528030937330436e-06,
+      "loss": 1.1923,
+      "step": 23718
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.527231231918194e-06,
+      "loss": 0.8875,
+      "step": 23719
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.526431551760309e-06,
+      "loss": 0.8974,
+      "step": 23720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.525631896862608e-06,
+      "loss": 0.9013,
+      "step": 23721
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524832267230889e-06,
+      "loss": 0.7386,
+      "step": 23722
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524032662870978e-06,
+      "loss": 1.201,
+      "step": 23723
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.523233083788688e-06,
+      "loss": 0.7494,
+      "step": 23724
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.522433529989831e-06,
+      "loss": 0.9102,
+      "step": 23725
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.521634001480226e-06,
+      "loss": 0.8121,
+      "step": 23726
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5208344982656845e-06,
+      "loss": 0.7138,
+      "step": 23727
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.52003502035202e-06,
+      "loss": 0.8179,
+      "step": 23728
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.519235567745046e-06,
+      "loss": 0.754,
+      "step": 23729
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5184361404505795e-06,
+      "loss": 0.6412,
+      "step": 23730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.517636738474434e-06,
+      "loss": 0.7488,
+      "step": 23731
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.516837361822419e-06,
+      "loss": 0.923,
+      "step": 23732
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5160380105003495e-06,
+      "loss": 0.4214,
+      "step": 23733
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.515238684514037e-06,
+      "loss": 1.1466,
+      "step": 23734
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.514439383869297e-06,
+      "loss": 0.7963,
+      "step": 23735
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.513640108571941e-06,
+      "loss": 0.9008,
+      "step": 23736
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.512840858627781e-06,
+      "loss": 0.8152,
+      "step": 23737
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5120416340426275e-06,
+      "loss": 0.8389,
+      "step": 23738
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.511242434822297e-06,
+      "loss": 0.7609,
+      "step": 23739
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.510443260972599e-06,
+      "loss": 0.8785,
+      "step": 23740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.509644112499347e-06,
+      "loss": 1.0374,
+      "step": 23741
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5088449894083485e-06,
+      "loss": 0.7861,
+      "step": 23742
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.508045891705415e-06,
+      "loss": 1.0101,
+      "step": 23743
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.507246819396361e-06,
+      "loss": 0.8335,
+      "step": 23744
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.506447772486997e-06,
+      "loss": 1.0522,
+      "step": 23745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505648750983131e-06,
+      "loss": 1.076,
+      "step": 23746
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.504849754890575e-06,
+      "loss": 0.9166,
+      "step": 23747
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.504050784215142e-06,
+      "loss": 0.8948,
+      "step": 23748
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.503251838962639e-06,
+      "loss": 0.8678,
+      "step": 23749
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.502452919138875e-06,
+      "loss": 0.9438,
+      "step": 23750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.501654024749668e-06,
+      "loss": 0.6952,
+      "step": 23751
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5008551558008145e-06,
+      "loss": 0.7433,
+      "step": 23752
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.500056312298134e-06,
+      "loss": 0.7261,
+      "step": 23753
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.499257494247433e-06,
+      "loss": 0.7386,
+      "step": 23754
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.49845870165452e-06,
+      "loss": 0.6463,
+      "step": 23755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4976599345252004e-06,
+      "loss": 0.857,
+      "step": 23756
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.49686119286529e-06,
+      "loss": 0.9202,
+      "step": 23757
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4960624766805946e-06,
+      "loss": 0.8872,
+      "step": 23758
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.49526378597692e-06,
+      "loss": 0.6943,
+      "step": 23759
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4944651207600765e-06,
+      "loss": 1.0313,
+      "step": 23760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.493666481035876e-06,
+      "loss": 0.9651,
+      "step": 23761
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4928678668101195e-06,
+      "loss": 1.0127,
+      "step": 23762
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.492069278088617e-06,
+      "loss": 0.9224,
+      "step": 23763
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.491270714877176e-06,
+      "loss": 1.2794,
+      "step": 23764
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.490472177181602e-06,
+      "loss": 0.95,
+      "step": 23765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4896736650077075e-06,
+      "loss": 0.878,
+      "step": 23766
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4888751783612945e-06,
+      "loss": 1.0458,
+      "step": 23767
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4880767172481695e-06,
+      "loss": 0.702,
+      "step": 23768
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.487278281674143e-06,
+      "loss": 0.6712,
+      "step": 23769
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.486479871645017e-06,
+      "loss": 0.8267,
+      "step": 23770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4856814871666045e-06,
+      "loss": 0.782,
+      "step": 23771
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.484883128244704e-06,
+      "loss": 1.0108,
+      "step": 23772
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4840847948851215e-06,
+      "loss": 0.7707,
+      "step": 23773
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.483286487093664e-06,
+      "loss": 0.9468,
+      "step": 23774
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4824882048761406e-06,
+      "loss": 0.9355,
+      "step": 23775
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.481689948238352e-06,
+      "loss": 1.0108,
+      "step": 23776
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.480891717186104e-06,
+      "loss": 0.7516,
+      "step": 23777
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.480093511725203e-06,
+      "loss": 0.7111,
+      "step": 23778
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.479295331861455e-06,
+      "loss": 0.8194,
+      "step": 23779
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.47849717760066e-06,
+      "loss": 0.7232,
+      "step": 23780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.477699048948627e-06,
+      "loss": 1.0715,
+      "step": 23781
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476900945911158e-06,
+      "loss": 0.7557,
+      "step": 23782
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476102868494052e-06,
+      "loss": 0.8398,
+      "step": 23783
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.47530481670312e-06,
+      "loss": 0.9315,
+      "step": 23784
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.474506790544162e-06,
+      "loss": 0.8283,
+      "step": 23785
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.473708790022981e-06,
+      "loss": 1.0754,
+      "step": 23786
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.472910815145383e-06,
+      "loss": 0.6615,
+      "step": 23787
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4721128659171705e-06,
+      "loss": 0.8518,
+      "step": 23788
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.471314942344143e-06,
+      "loss": 0.8686,
+      "step": 23789
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.470517044432104e-06,
+      "loss": 0.769,
+      "step": 23790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469719172186864e-06,
+      "loss": 0.7011,
+      "step": 23791
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46892132561421e-06,
+      "loss": 0.6474,
+      "step": 23792
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4681235047199565e-06,
+      "loss": 0.8179,
+      "step": 23793
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.467325709509899e-06,
+      "loss": 0.7134,
+      "step": 23794
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.466527939989841e-06,
+      "loss": 0.901,
+      "step": 23795
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.465730196165585e-06,
+      "loss": 0.7327,
+      "step": 23796
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.464932478042931e-06,
+      "loss": 0.8143,
+      "step": 23797
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.464134785627683e-06,
+      "loss": 0.5966,
+      "step": 23798
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.463337118925636e-06,
+      "loss": 0.947,
+      "step": 23799
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4625394779425955e-06,
+      "loss": 0.7539,
+      "step": 23800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.461741862684363e-06,
+      "loss": 1.0003,
+      "step": 23801
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.460944273156735e-06,
+      "loss": 0.8973,
+      "step": 23802
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4601467093655135e-06,
+      "loss": 1.0624,
+      "step": 23803
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4593491713164954e-06,
+      "loss": 0.7593,
+      "step": 23804
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.458551659015486e-06,
+      "loss": 0.9323,
+      "step": 23805
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.457754172468283e-06,
+      "loss": 0.8504,
+      "step": 23806
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4569567116806845e-06,
+      "loss": 0.719,
+      "step": 23807
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.456159276658487e-06,
+      "loss": 0.993,
+      "step": 23808
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.455361867407497e-06,
+      "loss": 0.7357,
+      "step": 23809
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.454564483933507e-06,
+      "loss": 0.9511,
+      "step": 23810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45376712624232e-06,
+      "loss": 0.4627,
+      "step": 23811
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4529697943397316e-06,
+      "loss": 1.0929,
+      "step": 23812
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.452172488231538e-06,
+      "loss": 1.0102,
+      "step": 23813
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.451375207923541e-06,
+      "loss": 0.7855,
+      "step": 23814
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.450577953421539e-06,
+      "loss": 0.582,
+      "step": 23815
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.449780724731328e-06,
+      "loss": 0.9355,
+      "step": 23816
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.448983521858704e-06,
+      "loss": 0.7032,
+      "step": 23817
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.448186344809467e-06,
+      "loss": 0.8134,
+      "step": 23818
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.447389193589415e-06,
+      "loss": 0.9252,
+      "step": 23819
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.446592068204341e-06,
+      "loss": 0.7743,
+      "step": 23820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4457949686600485e-06,
+      "loss": 0.8014,
+      "step": 23821
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.444997894962325e-06,
+      "loss": 0.7718,
+      "step": 23822
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.444200847116973e-06,
+      "loss": 0.9897,
+      "step": 23823
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.443403825129788e-06,
+      "loss": 0.8857,
+      "step": 23824
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.442606829006564e-06,
+      "loss": 1.0023,
+      "step": 23825
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.441809858753098e-06,
+      "loss": 0.9055,
+      "step": 23826
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.441012914375186e-06,
+      "loss": 0.7382,
+      "step": 23827
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.440215995878625e-06,
+      "loss": 1.1336,
+      "step": 23828
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.439419103269204e-06,
+      "loss": 0.824,
+      "step": 23829
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.438622236552727e-06,
+      "loss": 0.8149,
+      "step": 23830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.437825395734987e-06,
+      "loss": 1.0563,
+      "step": 23831
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.437028580821774e-06,
+      "loss": 0.933,
+      "step": 23832
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.436231791818884e-06,
+      "loss": 0.9735,
+      "step": 23833
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.435435028732111e-06,
+      "loss": 0.9697,
+      "step": 23834
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43463829156725e-06,
+      "loss": 0.84,
+      "step": 23835
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.433841580330097e-06,
+      "loss": 1.0318,
+      "step": 23836
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4330448950264435e-06,
+      "loss": 1.0544,
+      "step": 23837
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.432248235662082e-06,
+      "loss": 0.6456,
+      "step": 23838
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.431451602242809e-06,
+      "loss": 0.8982,
+      "step": 23839
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.430654994774418e-06,
+      "loss": 0.6338,
+      "step": 23840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4298584132627e-06,
+      "loss": 0.7005,
+      "step": 23841
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.429061857713448e-06,
+      "loss": 0.7311,
+      "step": 23842
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4282653281324534e-06,
+      "loss": 0.8387,
+      "step": 23843
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.427468824525509e-06,
+      "loss": 1.2887,
+      "step": 23844
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.426672346898408e-06,
+      "loss": 0.8273,
+      "step": 23845
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425875895256946e-06,
+      "loss": 0.8629,
+      "step": 23846
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425079469606906e-06,
+      "loss": 1.038,
+      "step": 23847
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.424283069954089e-06,
+      "loss": 0.8037,
+      "step": 23848
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.423486696304282e-06,
+      "loss": 1.0093,
+      "step": 23849
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.422690348663276e-06,
+      "loss": 0.9166,
+      "step": 23850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.421894027036865e-06,
+      "loss": 0.9279,
+      "step": 23851
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.421097731430839e-06,
+      "loss": 0.8054,
+      "step": 23852
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.420301461850983e-06,
+      "loss": 0.8466,
+      "step": 23853
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.419505218303095e-06,
+      "loss": 0.7517,
+      "step": 23854
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.418709000792962e-06,
+      "loss": 1.0306,
+      "step": 23855
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.417912809326375e-06,
+      "loss": 0.8703,
+      "step": 23856
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4171166439091224e-06,
+      "loss": 0.8401,
+      "step": 23857
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4163205045469975e-06,
+      "loss": 1.1243,
+      "step": 23858
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4155243912457885e-06,
+      "loss": 0.8431,
+      "step": 23859
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.414728304011281e-06,
+      "loss": 1.0975,
+      "step": 23860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.413932242849274e-06,
+      "loss": 0.6862,
+      "step": 23861
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.413136207765543e-06,
+      "loss": 1.061,
+      "step": 23862
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.412340198765887e-06,
+      "loss": 0.8981,
+      "step": 23863
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.411544215856091e-06,
+      "loss": 0.5418,
+      "step": 23864
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410748259041941e-06,
+      "loss": 1.4315,
+      "step": 23865
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.409952328329232e-06,
+      "loss": 0.8737,
+      "step": 23866
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.409156423723747e-06,
+      "loss": 0.7288,
+      "step": 23867
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.408360545231275e-06,
+      "loss": 0.84,
+      "step": 23868
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.407564692857603e-06,
+      "loss": 0.7393,
+      "step": 23869
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.406768866608522e-06,
+      "loss": 0.7797,
+      "step": 23870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.405973066489817e-06,
+      "loss": 0.9277,
+      "step": 23871
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.405177292507274e-06,
+      "loss": 0.7379,
+      "step": 23872
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.404381544666681e-06,
+      "loss": 0.7872,
+      "step": 23873
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.403585822973822e-06,
+      "loss": 0.8322,
+      "step": 23874
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.402790127434489e-06,
+      "loss": 0.8551,
+      "step": 23875
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.401994458054466e-06,
+      "loss": 0.8043,
+      "step": 23876
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4011988148395375e-06,
+      "loss": 0.8306,
+      "step": 23877
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40040319779549e-06,
+      "loss": 0.8982,
+      "step": 23878
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.399607606928112e-06,
+      "loss": 0.9491,
+      "step": 23879
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398812042243187e-06,
+      "loss": 0.9623,
+      "step": 23880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398016503746503e-06,
+      "loss": 0.785,
+      "step": 23881
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.397220991443841e-06,
+      "loss": 0.7796,
+      "step": 23882
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.396425505340987e-06,
+      "loss": 0.8094,
+      "step": 23883
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.395630045443726e-06,
+      "loss": 0.8751,
+      "step": 23884
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.394834611757845e-06,
+      "loss": 0.6887,
+      "step": 23885
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.394039204289128e-06,
+      "loss": 1.0689,
+      "step": 23886
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.393243823043356e-06,
+      "loss": 0.7431,
+      "step": 23887
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.392448468026317e-06,
+      "loss": 0.972,
+      "step": 23888
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.391653139243794e-06,
+      "loss": 0.7293,
+      "step": 23889
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.390857836701567e-06,
+      "loss": 0.5724,
+      "step": 23890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.39006256040543e-06,
+      "loss": 0.8914,
+      "step": 23891
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.389267310361151e-06,
+      "loss": 0.8866,
+      "step": 23892
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.388472086574524e-06,
+      "loss": 0.5364,
+      "step": 23893
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.38767688905133e-06,
+      "loss": 0.8505,
+      "step": 23894
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3868817177973505e-06,
+      "loss": 0.906,
+      "step": 23895
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.386086572818365e-06,
+      "loss": 0.7742,
+      "step": 23896
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.385291454120163e-06,
+      "loss": 0.9209,
+      "step": 23897
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.384496361708521e-06,
+      "loss": 0.7789,
+      "step": 23898
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.383701295589222e-06,
+      "loss": 0.6968,
+      "step": 23899
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.382906255768051e-06,
+      "loss": 0.9104,
+      "step": 23900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3821112422507905e-06,
+      "loss": 0.5943,
+      "step": 23901
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.381316255043215e-06,
+      "loss": 1.0105,
+      "step": 23902
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3805212941511095e-06,
+      "loss": 0.983,
+      "step": 23903
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.379726359580256e-06,
+      "loss": 0.903,
+      "step": 23904
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.378931451336432e-06,
+      "loss": 1.0659,
+      "step": 23905
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.378136569425423e-06,
+      "loss": 0.7232,
+      "step": 23906
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.377341713853007e-06,
+      "loss": 0.6242,
+      "step": 23907
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.376546884624961e-06,
+      "loss": 1.1548,
+      "step": 23908
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.375752081747072e-06,
+      "loss": 0.9214,
+      "step": 23909
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3749573052251155e-06,
+      "loss": 0.9239,
+      "step": 23910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374162555064875e-06,
+      "loss": 1.1046,
+      "step": 23911
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3733678312721234e-06,
+      "loss": 0.8362,
+      "step": 23912
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.372573133852645e-06,
+      "loss": 0.8515,
+      "step": 23913
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3717784628122145e-06,
+      "loss": 0.85,
+      "step": 23914
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.370983818156617e-06,
+      "loss": 0.7485,
+      "step": 23915
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3701891998916274e-06,
+      "loss": 0.6768,
+      "step": 23916
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.369394608023026e-06,
+      "loss": 1.0597,
+      "step": 23917
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.368600042556587e-06,
+      "loss": 0.8275,
+      "step": 23918
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.367805503498095e-06,
+      "loss": 0.7127,
+      "step": 23919
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.367010990853324e-06,
+      "loss": 0.8889,
+      "step": 23920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.366216504628055e-06,
+      "loss": 0.9983,
+      "step": 23921
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36542204482806e-06,
+      "loss": 0.9756,
+      "step": 23922
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3646276114591184e-06,
+      "loss": 0.7483,
+      "step": 23923
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36383320452701e-06,
+      "loss": 0.9537,
+      "step": 23924
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.363038824037511e-06,
+      "loss": 0.8211,
+      "step": 23925
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.362244469996397e-06,
+      "loss": 0.7014,
+      "step": 23926
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3614501424094445e-06,
+      "loss": 0.9128,
+      "step": 23927
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.360655841282431e-06,
+      "loss": 0.8184,
+      "step": 23928
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.359861566621133e-06,
+      "loss": 0.6431,
+      "step": 23929
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.359067318431324e-06,
+      "loss": 0.9935,
+      "step": 23930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3582730967187875e-06,
+      "loss": 0.8402,
+      "step": 23931
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.357478901489287e-06,
+      "loss": 0.783,
+      "step": 23932
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.356684732748607e-06,
+      "loss": 0.7684,
+      "step": 23933
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3558905905025205e-06,
+      "loss": 0.8697,
+      "step": 23934
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.355096474756801e-06,
+      "loss": 0.9099,
+      "step": 23935
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.354302385517224e-06,
+      "loss": 0.8981,
+      "step": 23936
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.353508322789566e-06,
+      "loss": 0.588,
+      "step": 23937
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.352714286579601e-06,
+      "loss": 0.7059,
+      "step": 23938
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351920276893101e-06,
+      "loss": 1.271,
+      "step": 23939
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351126293735843e-06,
+      "loss": 0.9507,
+      "step": 23940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.350332337113603e-06,
+      "loss": 0.9191,
+      "step": 23941
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.349538407032148e-06,
+      "loss": 0.7391,
+      "step": 23942
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.348744503497256e-06,
+      "loss": 0.6729,
+      "step": 23943
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3479506265147e-06,
+      "loss": 0.9281,
+      "step": 23944
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3471567760902496e-06,
+      "loss": 1.0077,
+      "step": 23945
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.346362952229683e-06,
+      "loss": 0.846,
+      "step": 23946
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.345569154938771e-06,
+      "loss": 0.7868,
+      "step": 23947
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.344775384223283e-06,
+      "loss": 0.8666,
+      "step": 23948
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.343981640088998e-06,
+      "loss": 0.8533,
+      "step": 23949
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.343187922541682e-06,
+      "loss": 0.8768,
+      "step": 23950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3423942315871145e-06,
+      "loss": 0.9204,
+      "step": 23951
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.341600567231059e-06,
+      "loss": 0.8294,
+      "step": 23952
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.34080692947929e-06,
+      "loss": 1.2104,
+      "step": 23953
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.340013318337577e-06,
+      "loss": 0.7799,
+      "step": 23954
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.339219733811697e-06,
+      "loss": 0.6622,
+      "step": 23955
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.338426175907416e-06,
+      "loss": 0.9624,
+      "step": 23956
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337632644630506e-06,
+      "loss": 0.8273,
+      "step": 23957
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.336839139986739e-06,
+      "loss": 1.0244,
+      "step": 23958
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.336045661981884e-06,
+      "loss": 0.7458,
+      "step": 23959
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.335252210621714e-06,
+      "loss": 0.9653,
+      "step": 23960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.334458785911998e-06,
+      "loss": 0.78,
+      "step": 23961
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.333665387858502e-06,
+      "loss": 0.8993,
+      "step": 23962
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3328720164669965e-06,
+      "loss": 0.9245,
+      "step": 23963
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3320786717432555e-06,
+      "loss": 0.9233,
+      "step": 23964
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.331285353693046e-06,
+      "loss": 1.1497,
+      "step": 23965
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.330492062322133e-06,
+      "loss": 0.9329,
+      "step": 23966
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.329698797636293e-06,
+      "loss": 1.0236,
+      "step": 23967
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.32890555964129e-06,
+      "loss": 0.9984,
+      "step": 23968
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.328112348342894e-06,
+      "loss": 0.8948,
+      "step": 23969
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.327319163746871e-06,
+      "loss": 0.858,
+      "step": 23970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.326526005858997e-06,
+      "loss": 0.7541,
+      "step": 23971
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.325732874685027e-06,
+      "loss": 0.8725,
+      "step": 23972
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.324939770230738e-06,
+      "loss": 0.8425,
+      "step": 23973
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3241466925018955e-06,
+      "loss": 0.8041,
+      "step": 23974
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.323353641504264e-06,
+      "loss": 0.8438,
+      "step": 23975
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3225606172436135e-06,
+      "loss": 1.1949,
+      "step": 23976
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.321767619725713e-06,
+      "loss": 0.9069,
+      "step": 23977
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.320974648956326e-06,
+      "loss": 0.7016,
+      "step": 23978
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.320181704941218e-06,
+      "loss": 0.6034,
+      "step": 23979
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.319388787686158e-06,
+      "loss": 0.9726,
+      "step": 23980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3185958971969155e-06,
+      "loss": 0.8901,
+      "step": 23981
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317803033479248e-06,
+      "loss": 1.1452,
+      "step": 23982
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317010196538929e-06,
+      "loss": 0.7372,
+      "step": 23983
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.316217386381717e-06,
+      "loss": 0.712,
+      "step": 23984
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.315424603013382e-06,
+      "loss": 0.9632,
+      "step": 23985
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.314631846439691e-06,
+      "loss": 0.8859,
+      "step": 23986
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3138391166664045e-06,
+      "loss": 0.7084,
+      "step": 23987
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3130464136992884e-06,
+      "loss": 0.7134,
+      "step": 23988
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.312253737544109e-06,
+      "loss": 1.1061,
+      "step": 23989
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.311461088206632e-06,
+      "loss": 0.9604,
+      "step": 23990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3106684656926195e-06,
+      "loss": 0.7337,
+      "step": 23991
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3098758700078345e-06,
+      "loss": 0.7565,
+      "step": 23992
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30908330115804e-06,
+      "loss": 0.8443,
+      "step": 23993
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.308290759149003e-06,
+      "loss": 0.7815,
+      "step": 23994
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.307498243986487e-06,
+      "loss": 1.0896,
+      "step": 23995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.306705755676252e-06,
+      "loss": 0.7567,
+      "step": 23996
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.305913294224062e-06,
+      "loss": 0.9556,
+      "step": 23997
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3051208596356825e-06,
+      "loss": 0.925,
+      "step": 23998
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.304328451916875e-06,
+      "loss": 0.8931,
+      "step": 23999
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.303536071073397e-06,
+      "loss": 1.068,
+      "step": 24000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.302743717111024e-06,
+      "loss": 0.8192,
+      "step": 24001
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3019513900355026e-06,
+      "loss": 0.9774,
+      "step": 24002
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.301159089852602e-06,
+      "loss": 0.9987,
+      "step": 24003
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3003668165680844e-06,
+      "loss": 0.886,
+      "step": 24004
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29957457018771e-06,
+      "loss": 0.5178,
+      "step": 24005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.298782350717238e-06,
+      "loss": 0.7623,
+      "step": 24006
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.297990158162434e-06,
+      "loss": 1.1268,
+      "step": 24007
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2971979925290576e-06,
+      "loss": 0.9659,
+      "step": 24008
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.296405853822865e-06,
+      "loss": 0.8212,
+      "step": 24009
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.295613742049622e-06,
+      "loss": 0.672,
+      "step": 24010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.294821657215091e-06,
+      "loss": 0.8112,
+      "step": 24011
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.294029599325026e-06,
+      "loss": 0.8955,
+      "step": 24012
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.293237568385189e-06,
+      "loss": 0.8052,
+      "step": 24013
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.292445564401339e-06,
+      "loss": 0.9692,
+      "step": 24014
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.291653587379236e-06,
+      "loss": 0.6698,
+      "step": 24015
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.290861637324641e-06,
+      "loss": 0.853,
+      "step": 24016
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.290069714243312e-06,
+      "loss": 0.9275,
+      "step": 24017
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.289277818141006e-06,
+      "loss": 1.1172,
+      "step": 24018
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.288485949023486e-06,
+      "loss": 0.5852,
+      "step": 24019
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.287694106896509e-06,
+      "loss": 0.7292,
+      "step": 24020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.286902291765834e-06,
+      "loss": 0.8221,
+      "step": 24021
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2861105036372165e-06,
+      "loss": 1.1846,
+      "step": 24022
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.285318742516415e-06,
+      "loss": 1.0114,
+      "step": 24023
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.284527008409187e-06,
+      "loss": 0.7106,
+      "step": 24024
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283735301321292e-06,
+      "loss": 0.9882,
+      "step": 24025
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.282943621258488e-06,
+      "loss": 1.1084,
+      "step": 24026
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.282151968226527e-06,
+      "loss": 0.6553,
+      "step": 24027
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.281360342231173e-06,
+      "loss": 0.8056,
+      "step": 24028
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2805687432781795e-06,
+      "loss": 0.6112,
+      "step": 24029
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2797771713733025e-06,
+      "loss": 0.7342,
+      "step": 24030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.278985626522302e-06,
+      "loss": 1.001,
+      "step": 24031
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.278194108730928e-06,
+      "loss": 0.6955,
+      "step": 24032
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.277402618004939e-06,
+      "loss": 0.9427,
+      "step": 24033
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.276611154350093e-06,
+      "loss": 1.2686,
+      "step": 24034
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.275819717772145e-06,
+      "loss": 0.9548,
+      "step": 24035
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.275028308276845e-06,
+      "loss": 0.6659,
+      "step": 24036
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.274236925869956e-06,
+      "loss": 0.7226,
+      "step": 24037
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.273445570557231e-06,
+      "loss": 0.4473,
+      "step": 24038
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.272654242344422e-06,
+      "loss": 0.7578,
+      "step": 24039
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2718629412372854e-06,
+      "loss": 0.6092,
+      "step": 24040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.27107166724158e-06,
+      "loss": 0.7659,
+      "step": 24041
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.270280420363051e-06,
+      "loss": 0.7358,
+      "step": 24042
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.269489200607457e-06,
+      "loss": 1.0451,
+      "step": 24043
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.268698007980552e-06,
+      "loss": 0.7518,
+      "step": 24044
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267906842488088e-06,
+      "loss": 0.6347,
+      "step": 24045
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2671157041358225e-06,
+      "loss": 1.1663,
+      "step": 24046
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.266324592929505e-06,
+      "loss": 0.705,
+      "step": 24047
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2655335088748906e-06,
+      "loss": 0.8275,
+      "step": 24048
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.264742451977729e-06,
+      "loss": 0.7072,
+      "step": 24049
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.263951422243777e-06,
+      "loss": 0.8873,
+      "step": 24050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.263160419678787e-06,
+      "loss": 0.9174,
+      "step": 24051
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.262369444288507e-06,
+      "loss": 0.7512,
+      "step": 24052
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.261578496078692e-06,
+      "loss": 1.0405,
+      "step": 24053
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.260787575055092e-06,
+      "loss": 0.8811,
+      "step": 24054
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.25999668122346e-06,
+      "loss": 0.8791,
+      "step": 24055
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.259205814589549e-06,
+      "loss": 0.9566,
+      "step": 24056
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.258414975159108e-06,
+      "loss": 0.8704,
+      "step": 24057
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.257624162937888e-06,
+      "loss": 0.8311,
+      "step": 24058
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.256833377931642e-06,
+      "loss": 0.7105,
+      "step": 24059
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.256042620146119e-06,
+      "loss": 1.053,
+      "step": 24060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.25525188958707e-06,
+      "loss": 0.6923,
+      "step": 24061
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.254461186260245e-06,
+      "loss": 0.8334,
+      "step": 24062
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.253670510171394e-06,
+      "loss": 0.728,
+      "step": 24063
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2528798613262665e-06,
+      "loss": 1.1325,
+      "step": 24064
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2520892397306145e-06,
+      "loss": 0.7156,
+      "step": 24065
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251298645390185e-06,
+      "loss": 1.1367,
+      "step": 24066
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.250508078310725e-06,
+      "loss": 0.8784,
+      "step": 24067
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2497175384979905e-06,
+      "loss": 0.8573,
+      "step": 24068
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2489270259577254e-06,
+      "loss": 0.9371,
+      "step": 24069
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.248136540695677e-06,
+      "loss": 0.9001,
+      "step": 24070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2473460827176025e-06,
+      "loss": 0.8227,
+      "step": 24071
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.246555652029238e-06,
+      "loss": 0.994,
+      "step": 24072
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.245765248636341e-06,
+      "loss": 1.0336,
+      "step": 24073
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2449748725446556e-06,
+      "loss": 0.7414,
+      "step": 24074
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2441845237599285e-06,
+      "loss": 0.7852,
+      "step": 24075
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.243394202287907e-06,
+      "loss": 0.7461,
+      "step": 24076
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.242603908134344e-06,
+      "loss": 0.6165,
+      "step": 24077
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.24181364130498e-06,
+      "loss": 0.7741,
+      "step": 24078
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241023401805564e-06,
+      "loss": 0.967,
+      "step": 24079
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.240233189641844e-06,
+      "loss": 0.8418,
+      "step": 24080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23944300481957e-06,
+      "loss": 0.8848,
+      "step": 24081
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23865284734448e-06,
+      "loss": 1.0446,
+      "step": 24082
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.237862717222325e-06,
+      "loss": 0.9204,
+      "step": 24083
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23707261445885e-06,
+      "loss": 1.1573,
+      "step": 24084
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.236282539059799e-06,
+      "loss": 0.82,
+      "step": 24085
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.235492491030921e-06,
+      "loss": 1.1584,
+      "step": 24086
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23470247037796e-06,
+      "loss": 0.8443,
+      "step": 24087
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.233912477106658e-06,
+      "loss": 0.4994,
+      "step": 24088
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.233122511222767e-06,
+      "loss": 0.6984,
+      "step": 24089
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.232332572732025e-06,
+      "loss": 1.1178,
+      "step": 24090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231542661640182e-06,
+      "loss": 0.8369,
+      "step": 24091
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.230752777952975e-06,
+      "loss": 0.75,
+      "step": 24092
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.229962921676156e-06,
+      "loss": 0.8957,
+      "step": 24093
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.229173092815461e-06,
+      "loss": 0.7324,
+      "step": 24094
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.228383291376641e-06,
+      "loss": 1.0414,
+      "step": 24095
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.227593517365436e-06,
+      "loss": 0.7203,
+      "step": 24096
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2268037707875875e-06,
+      "loss": 0.7746,
+      "step": 24097
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.226014051648843e-06,
+      "loss": 0.71,
+      "step": 24098
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.225224359954943e-06,
+      "loss": 0.6066,
+      "step": 24099
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.224434695711632e-06,
+      "loss": 0.8121,
+      "step": 24100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223645058924652e-06,
+      "loss": 1.0767,
+      "step": 24101
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222855449599743e-06,
+      "loss": 0.9346,
+      "step": 24102
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222065867742647e-06,
+      "loss": 0.788,
+      "step": 24103
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.221276313359109e-06,
+      "loss": 0.8585,
+      "step": 24104
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.22048678645487e-06,
+      "loss": 0.6584,
+      "step": 24105
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2196972870356685e-06,
+      "loss": 0.8113,
+      "step": 24106
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.218907815107249e-06,
+      "loss": 0.8028,
+      "step": 24107
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.218118370675352e-06,
+      "loss": 0.9815,
+      "step": 24108
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.21732895374572e-06,
+      "loss": 0.9147,
+      "step": 24109
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.216539564324089e-06,
+      "loss": 0.8506,
+      "step": 24110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2157502024162085e-06,
+      "loss": 0.5202,
+      "step": 24111
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.214960868027806e-06,
+      "loss": 1.0217,
+      "step": 24112
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.21417156116463e-06,
+      "loss": 0.9678,
+      "step": 24113
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.213382281832421e-06,
+      "loss": 1.0842,
+      "step": 24114
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.212593030036912e-06,
+      "loss": 0.9217,
+      "step": 24115
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.211803805783851e-06,
+      "loss": 0.9276,
+      "step": 24116
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.211014609078972e-06,
+      "loss": 0.9172,
+      "step": 24117
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.210225439928016e-06,
+      "loss": 0.5787,
+      "step": 24118
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20943629833672e-06,
+      "loss": 0.906,
+      "step": 24119
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.208647184310826e-06,
+      "loss": 0.9443,
+      "step": 24120
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.207858097856072e-06,
+      "loss": 0.7152,
+      "step": 24121
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2070690389781926e-06,
+      "loss": 0.7766,
+      "step": 24122
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20628000768293e-06,
+      "loss": 0.7084,
+      "step": 24123
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.205491003976017e-06,
+      "loss": 1.1582,
+      "step": 24124
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2047020278631985e-06,
+      "loss": 0.9781,
+      "step": 24125
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.203913079350206e-06,
+      "loss": 0.7684,
+      "step": 24126
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.203124158442781e-06,
+      "loss": 0.9243,
+      "step": 24127
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202335265146656e-06,
+      "loss": 0.9045,
+      "step": 24128
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2015463994675725e-06,
+      "loss": 0.8727,
+      "step": 24129
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2007575614112655e-06,
+      "loss": 0.7586,
+      "step": 24130
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.199968750983474e-06,
+      "loss": 0.6502,
+      "step": 24131
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.19917996818993e-06,
+      "loss": 0.8428,
+      "step": 24132
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.198391213036367e-06,
+      "loss": 1.1781,
+      "step": 24133
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.197602485528531e-06,
+      "loss": 0.6542,
+      "step": 24134
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196813785672149e-06,
+      "loss": 1.0238,
+      "step": 24135
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196025113472961e-06,
+      "loss": 1.0727,
+      "step": 24136
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.195236468936698e-06,
+      "loss": 1.0336,
+      "step": 24137
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1944478520691e-06,
+      "loss": 0.9246,
+      "step": 24138
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.193659262875899e-06,
+      "loss": 0.8882,
+      "step": 24139
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.19287070136283e-06,
+      "loss": 1.0291,
+      "step": 24140
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.192082167535634e-06,
+      "loss": 1.0905,
+      "step": 24141
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.191293661400033e-06,
+      "loss": 0.8542,
+      "step": 24142
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.190505182961768e-06,
+      "loss": 0.6536,
+      "step": 24143
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.189716732226574e-06,
+      "loss": 0.8987,
+      "step": 24144
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188928309200183e-06,
+      "loss": 0.7422,
+      "step": 24145
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188139913888327e-06,
+      "loss": 0.7606,
+      "step": 24146
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.187351546296741e-06,
+      "loss": 0.8926,
+      "step": 24147
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.18656320643116e-06,
+      "loss": 0.9959,
+      "step": 24148
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.185774894297311e-06,
+      "loss": 0.9476,
+      "step": 24149
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184986609900934e-06,
+      "loss": 0.6223,
+      "step": 24150
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184198353247759e-06,
+      "loss": 0.6434,
+      "step": 24151
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.183410124343515e-06,
+      "loss": 0.8331,
+      "step": 24152
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1826219231939366e-06,
+      "loss": 1.0833,
+      "step": 24153
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.181833749804757e-06,
+      "loss": 0.7745,
+      "step": 24154
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.181045604181703e-06,
+      "loss": 0.7246,
+      "step": 24155
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.180257486330512e-06,
+      "loss": 0.8054,
+      "step": 24156
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179469396256912e-06,
+      "loss": 0.7109,
+      "step": 24157
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.178681333966634e-06,
+      "loss": 1.1085,
+      "step": 24158
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.177893299465411e-06,
+      "loss": 0.5325,
+      "step": 24159
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.177105292758971e-06,
+      "loss": 1.0587,
+      "step": 24160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.176317313853051e-06,
+      "loss": 0.94,
+      "step": 24161
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.175529362753372e-06,
+      "loss": 0.9558,
+      "step": 24162
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.174741439465668e-06,
+      "loss": 1.0134,
+      "step": 24163
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.173953543995668e-06,
+      "loss": 0.9173,
+      "step": 24164
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.173165676349103e-06,
+      "loss": 0.8992,
+      "step": 24165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.172377836531705e-06,
+      "loss": 1.0037,
+      "step": 24166
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1715900245491956e-06,
+      "loss": 0.8961,
+      "step": 24167
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.170802240407312e-06,
+      "loss": 0.7534,
+      "step": 24168
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.170014484111779e-06,
+      "loss": 0.8082,
+      "step": 24169
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169226755668329e-06,
+      "loss": 0.7987,
+      "step": 24170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.168439055082685e-06,
+      "loss": 0.6159,
+      "step": 24171
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.167651382360578e-06,
+      "loss": 0.6809,
+      "step": 24172
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.166863737507733e-06,
+      "loss": 1.0919,
+      "step": 24173
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1660761205298825e-06,
+      "loss": 1.0142,
+      "step": 24174
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1652885314327535e-06,
+      "loss": 0.8766,
+      "step": 24175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.164500970222069e-06,
+      "loss": 0.7512,
+      "step": 24176
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.163713436903562e-06,
+      "loss": 1.1325,
+      "step": 24177
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.162925931482956e-06,
+      "loss": 0.7872,
+      "step": 24178
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.16213845396598e-06,
+      "loss": 0.8523,
+      "step": 24179
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.16135100435836e-06,
+      "loss": 1.0415,
+      "step": 24180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160563582665821e-06,
+      "loss": 0.8484,
+      "step": 24181
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.159776188894087e-06,
+      "loss": 0.969,
+      "step": 24182
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1589888230488895e-06,
+      "loss": 0.9096,
+      "step": 24183
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.158201485135952e-06,
+      "loss": 0.681,
+      "step": 24184
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.157414175160997e-06,
+      "loss": 1.0071,
+      "step": 24185
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.156626893129756e-06,
+      "loss": 1.0733,
+      "step": 24186
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155839639047949e-06,
+      "loss": 1.053,
+      "step": 24187
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155052412921305e-06,
+      "loss": 0.7522,
+      "step": 24188
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.154265214755545e-06,
+      "loss": 0.9705,
+      "step": 24189
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1534780445564e-06,
+      "loss": 1.0685,
+      "step": 24190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152690902329585e-06,
+      "loss": 0.7419,
+      "step": 24191
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.151903788080831e-06,
+      "loss": 0.569,
+      "step": 24192
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.151116701815859e-06,
+      "loss": 0.7762,
+      "step": 24193
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150329643540393e-06,
+      "loss": 0.8956,
+      "step": 24194
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.149542613260157e-06,
+      "loss": 0.8525,
+      "step": 24195
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.148755610980877e-06,
+      "loss": 0.6864,
+      "step": 24196
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.147968636708273e-06,
+      "loss": 1.2182,
+      "step": 24197
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.147181690448067e-06,
+      "loss": 1.0492,
+      "step": 24198
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.146394772205986e-06,
+      "loss": 0.9551,
+      "step": 24199
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1456078819877516e-06,
+      "loss": 0.9202,
+      "step": 24200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144821019799083e-06,
+      "loss": 0.6247,
+      "step": 24201
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144034185645704e-06,
+      "loss": 0.659,
+      "step": 24202
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.143247379533335e-06,
+      "loss": 0.7194,
+      "step": 24203
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1424606014677e-06,
+      "loss": 1.0177,
+      "step": 24204
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.141673851454521e-06,
+      "loss": 1.0289,
+      "step": 24205
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1408871294995175e-06,
+      "loss": 1.0383,
+      "step": 24206
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140100435608409e-06,
+      "loss": 0.8529,
+      "step": 24207
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139313769786923e-06,
+      "loss": 1.0728,
+      "step": 24208
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.138527132040775e-06,
+      "loss": 0.678,
+      "step": 24209
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.137740522375687e-06,
+      "loss": 0.8774,
+      "step": 24210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.136953940797377e-06,
+      "loss": 0.6981,
+      "step": 24211
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.136167387311567e-06,
+      "loss": 0.658,
+      "step": 24212
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.135380861923976e-06,
+      "loss": 0.4946,
+      "step": 24213
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.134594364640327e-06,
+      "loss": 0.9829,
+      "step": 24214
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.133807895466337e-06,
+      "loss": 0.5287,
+      "step": 24215
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1330214544077215e-06,
+      "loss": 0.9133,
+      "step": 24216
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.132235041470207e-06,
+      "loss": 0.878,
+      "step": 24217
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.131448656659508e-06,
+      "loss": 0.691,
+      "step": 24218
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.130662299981341e-06,
+      "loss": 0.9339,
+      "step": 24219
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.129875971441434e-06,
+      "loss": 0.8505,
+      "step": 24220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.129089671045494e-06,
+      "loss": 0.9443,
+      "step": 24221
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.128303398799246e-06,
+      "loss": 1.1229,
+      "step": 24222
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1275171547084045e-06,
+      "loss": 0.761,
+      "step": 24223
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126730938778688e-06,
+      "loss": 0.9192,
+      "step": 24224
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1259447510158136e-06,
+      "loss": 0.9596,
+      "step": 24225
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1251585914255e-06,
+      "loss": 1.1992,
+      "step": 24226
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.124372460013463e-06,
+      "loss": 1.1279,
+      "step": 24227
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.123586356785418e-06,
+      "loss": 0.8292,
+      "step": 24228
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.122800281747085e-06,
+      "loss": 0.855,
+      "step": 24229
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.122014234904184e-06,
+      "loss": 0.6796,
+      "step": 24230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.121228216262421e-06,
+      "loss": 1.1395,
+      "step": 24231
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.120442225827519e-06,
+      "loss": 0.8566,
+      "step": 24232
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.11965626360519e-06,
+      "loss": 1.2105,
+      "step": 24233
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1188703296011505e-06,
+      "loss": 0.8426,
+      "step": 24234
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.118084423821119e-06,
+      "loss": 0.6603,
+      "step": 24235
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.117298546270809e-06,
+      "loss": 1.184,
+      "step": 24236
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.116512696955933e-06,
+      "loss": 1.0063,
+      "step": 24237
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.115726875882209e-06,
+      "loss": 0.6445,
+      "step": 24238
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1149410830553525e-06,
+      "loss": 0.9625,
+      "step": 24239
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.114155318481076e-06,
+      "loss": 0.99,
+      "step": 24240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1133695821650916e-06,
+      "loss": 0.7368,
+      "step": 24241
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.112583874113116e-06,
+      "loss": 0.7663,
+      "step": 24242
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.11179819433086e-06,
+      "loss": 1.0012,
+      "step": 24243
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1110125428240415e-06,
+      "loss": 0.8773,
+      "step": 24244
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.110226919598371e-06,
+      "loss": 1.039,
+      "step": 24245
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.109441324659561e-06,
+      "loss": 1.0667,
+      "step": 24246
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.108655758013327e-06,
+      "loss": 0.9178,
+      "step": 24247
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.107870219665382e-06,
+      "loss": 0.8217,
+      "step": 24248
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.107084709621435e-06,
+      "loss": 0.6903,
+      "step": 24249
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.106299227887203e-06,
+      "loss": 0.8367,
+      "step": 24250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1055137744683936e-06,
+      "loss": 0.7855,
+      "step": 24251
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.104728349370719e-06,
+      "loss": 0.9586,
+      "step": 24252
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1039429525998925e-06,
+      "loss": 0.7032,
+      "step": 24253
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.103157584161627e-06,
+      "loss": 0.7531,
+      "step": 24254
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.102372244061631e-06,
+      "loss": 1.0113,
+      "step": 24255
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.101586932305618e-06,
+      "loss": 0.9078,
+      "step": 24256
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1008016488992976e-06,
+      "loss": 0.8547,
+      "step": 24257
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.10001639384838e-06,
+      "loss": 0.6187,
+      "step": 24258
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.099231167158575e-06,
+      "loss": 0.9465,
+      "step": 24259
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0984459688356e-06,
+      "loss": 0.8383,
+      "step": 24260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.097660798885153e-06,
+      "loss": 0.6475,
+      "step": 24261
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0968756573129525e-06,
+      "loss": 0.9085,
+      "step": 24262
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.096090544124706e-06,
+      "loss": 1.0072,
+      "step": 24263
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09530545932612e-06,
+      "loss": 0.7545,
+      "step": 24264
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0945204029229076e-06,
+      "loss": 0.7125,
+      "step": 24265
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.093735374920777e-06,
+      "loss": 0.7347,
+      "step": 24266
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.092950375325437e-06,
+      "loss": 1.4639,
+      "step": 24267
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.092165404142594e-06,
+      "loss": 0.8016,
+      "step": 24268
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09138046137796e-06,
+      "loss": 1.1622,
+      "step": 24269
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.090595547037242e-06,
+      "loss": 0.82,
+      "step": 24270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.089810661126146e-06,
+      "loss": 0.8608,
+      "step": 24271
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.089025803650382e-06,
+      "loss": 1.1148,
+      "step": 24272
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.088240974615653e-06,
+      "loss": 0.9982,
+      "step": 24273
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.087456174027674e-06,
+      "loss": 0.7418,
+      "step": 24274
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.086671401892148e-06,
+      "loss": 0.9473,
+      "step": 24275
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0858866582147814e-06,
+      "loss": 0.4565,
+      "step": 24276
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.085101943001279e-06,
+      "loss": 0.954,
+      "step": 24277
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.084317256257353e-06,
+      "loss": 0.8467,
+      "step": 24278
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.083532597988706e-06,
+      "loss": 0.9851,
+      "step": 24279
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.082747968201047e-06,
+      "loss": 1.0778,
+      "step": 24280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081963366900079e-06,
+      "loss": 0.7738,
+      "step": 24281
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081178794091504e-06,
+      "loss": 0.8554,
+      "step": 24282
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.080394249781036e-06,
+      "loss": 0.8793,
+      "step": 24283
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0796097339743766e-06,
+      "loss": 0.8113,
+      "step": 24284
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078825246677229e-06,
+      "loss": 0.6948,
+      "step": 24285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0780407878952985e-06,
+      "loss": 1.0092,
+      "step": 24286
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.077256357634292e-06,
+      "loss": 1.1439,
+      "step": 24287
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.076471955899914e-06,
+      "loss": 0.7862,
+      "step": 24288
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0756875826978655e-06,
+      "loss": 0.7248,
+      "step": 24289
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.074903238033858e-06,
+      "loss": 0.8701,
+      "step": 24290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0741189219135855e-06,
+      "loss": 0.8709,
+      "step": 24291
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.073334634342757e-06,
+      "loss": 0.8227,
+      "step": 24292
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.072550375327075e-06,
+      "loss": 0.7794,
+      "step": 24293
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.071766144872243e-06,
+      "loss": 0.957,
+      "step": 24294
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0709819429839625e-06,
+      "loss": 0.7213,
+      "step": 24295
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.070197769667938e-06,
+      "loss": 1.1509,
+      "step": 24296
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.069413624929873e-06,
+      "loss": 1.013,
+      "step": 24297
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.068629508775465e-06,
+      "loss": 0.7075,
+      "step": 24298
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.067845421210424e-06,
+      "loss": 0.7954,
+      "step": 24299
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.06706136224045e-06,
+      "loss": 0.8403,
+      "step": 24300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0662773318712385e-06,
+      "loss": 0.882,
+      "step": 24301
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.065493330108496e-06,
+      "loss": 0.8295,
+      "step": 24302
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064709356957923e-06,
+      "loss": 0.7773,
+      "step": 24303
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.063925412425218e-06,
+      "loss": 0.6024,
+      "step": 24304
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.063141496516087e-06,
+      "loss": 1.0602,
+      "step": 24305
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.062357609236228e-06,
+      "loss": 0.5889,
+      "step": 24306
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0615737505913395e-06,
+      "loss": 0.7224,
+      "step": 24307
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060789920587127e-06,
+      "loss": 0.8385,
+      "step": 24308
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060006119229287e-06,
+      "loss": 0.9386,
+      "step": 24309
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.059222346523521e-06,
+      "loss": 0.8986,
+      "step": 24310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.058438602475527e-06,
+      "loss": 0.7364,
+      "step": 24311
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.057654887091006e-06,
+      "loss": 0.978,
+      "step": 24312
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056871200375653e-06,
+      "loss": 0.9549,
+      "step": 24313
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056087542335173e-06,
+      "loss": 0.6027,
+      "step": 24314
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.055303912975261e-06,
+      "loss": 0.9177,
+      "step": 24315
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.054520312301617e-06,
+      "loss": 1.0681,
+      "step": 24316
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.053736740319941e-06,
+      "loss": 1.0116,
+      "step": 24317
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.052953197035929e-06,
+      "loss": 0.7442,
+      "step": 24318
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.05216968245528e-06,
+      "loss": 0.9054,
+      "step": 24319
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.051386196583693e-06,
+      "loss": 0.8731,
+      "step": 24320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.050602739426863e-06,
+      "loss": 0.9084,
+      "step": 24321
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049819310990486e-06,
+      "loss": 0.677,
+      "step": 24322
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049035911280263e-06,
+      "loss": 0.6402,
+      "step": 24323
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.048252540301891e-06,
+      "loss": 0.9127,
+      "step": 24324
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.047469198061062e-06,
+      "loss": 0.8604,
+      "step": 24325
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0466858845634795e-06,
+      "loss": 0.8242,
+      "step": 24326
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0459025998148345e-06,
+      "loss": 0.9933,
+      "step": 24327
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.045119343820824e-06,
+      "loss": 0.8693,
+      "step": 24328
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.044336116587146e-06,
+      "loss": 0.8155,
+      "step": 24329
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0435529181195e-06,
+      "loss": 0.7551,
+      "step": 24330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.042769748423569e-06,
+      "loss": 0.9305,
+      "step": 24331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.04198660750506e-06,
+      "loss": 0.9947,
+      "step": 24332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.041203495369661e-06,
+      "loss": 0.8456,
+      "step": 24333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0404204120230695e-06,
+      "loss": 0.8666,
+      "step": 24334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.039637357470982e-06,
+      "loss": 0.7072,
+      "step": 24335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.038854331719093e-06,
+      "loss": 0.7139,
+      "step": 24336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.038071334773093e-06,
+      "loss": 0.9596,
+      "step": 24337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.037288366638678e-06,
+      "loss": 0.8737,
+      "step": 24338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.036505427321543e-06,
+      "loss": 0.7813,
+      "step": 24339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.035722516827382e-06,
+      "loss": 0.7443,
+      "step": 24340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.034939635161887e-06,
+      "loss": 0.7662,
+      "step": 24341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.034156782330751e-06,
+      "loss": 0.8337,
+      "step": 24342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033373958339663e-06,
+      "loss": 0.9579,
+      "step": 24343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0325911631943245e-06,
+      "loss": 0.9398,
+      "step": 24344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.031808396900422e-06,
+      "loss": 0.8722,
+      "step": 24345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.031025659463652e-06,
+      "loss": 0.6973,
+      "step": 24346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0302429508897e-06,
+      "loss": 0.9354,
+      "step": 24347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.029460271184264e-06,
+      "loss": 0.802,
+      "step": 24348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.028677620353034e-06,
+      "loss": 0.8836,
+      "step": 24349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027894998401705e-06,
+      "loss": 0.6692,
+      "step": 24350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027112405335961e-06,
+      "loss": 0.726,
+      "step": 24351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.026329841161496e-06,
+      "loss": 0.7334,
+      "step": 24352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.025547305884001e-06,
+      "loss": 0.7685,
+      "step": 24353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0247647995091695e-06,
+      "loss": 1.0703,
+      "step": 24354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.02398232204269e-06,
+      "loss": 0.8365,
+      "step": 24355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.02319987349025e-06,
+      "loss": 0.7824,
+      "step": 24356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0224174538575444e-06,
+      "loss": 1.003,
+      "step": 24357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.021635063150259e-06,
+      "loss": 0.7515,
+      "step": 24358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020852701374086e-06,
+      "loss": 0.9194,
+      "step": 24359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020070368534719e-06,
+      "loss": 1.1163,
+      "step": 24360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.019288064637836e-06,
+      "loss": 1.0538,
+      "step": 24361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.018505789689134e-06,
+      "loss": 0.8723,
+      "step": 24362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.017723543694301e-06,
+      "loss": 0.578,
+      "step": 24363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.016941326659022e-06,
+      "loss": 0.8202,
+      "step": 24364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0161591385889885e-06,
+      "loss": 0.9803,
+      "step": 24365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0153769794898905e-06,
+      "loss": 1.0242,
+      "step": 24366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.014594849367413e-06,
+      "loss": 0.6964,
+      "step": 24367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013812748227241e-06,
+      "loss": 1.0028,
+      "step": 24368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013030676075069e-06,
+      "loss": 0.8107,
+      "step": 24369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.012248632916582e-06,
+      "loss": 0.7414,
+      "step": 24370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.011466618757463e-06,
+      "loss": 1.088,
+      "step": 24371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.010684633603403e-06,
+      "loss": 0.8632,
+      "step": 24372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.009902677460086e-06,
+      "loss": 0.561,
+      "step": 24373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0091207503332e-06,
+      "loss": 1.0124,
+      "step": 24374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.00833885222843e-06,
+      "loss": 1.002,
+      "step": 24375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.007556983151466e-06,
+      "loss": 1.0778,
+      "step": 24376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0067751431079865e-06,
+      "loss": 0.6254,
+      "step": 24377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.005993332103686e-06,
+      "loss": 0.9578,
+      "step": 24378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.005211550144243e-06,
+      "loss": 0.6891,
+      "step": 24379
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.004429797235349e-06,
+      "loss": 0.7917,
+      "step": 24380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.003648073382683e-06,
+      "loss": 0.8449,
+      "step": 24381
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0028663785919316e-06,
+      "loss": 0.6254,
+      "step": 24382
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.002084712868779e-06,
+      "loss": 0.827,
+      "step": 24383
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.001303076218911e-06,
+      "loss": 0.6754,
+      "step": 24384
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.000521468648011e-06,
+      "loss": 0.8168,
+      "step": 24385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.999739890161761e-06,
+      "loss": 0.7473,
+      "step": 24386
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.998958340765849e-06,
+      "loss": 1.1306,
+      "step": 24387
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.998176820465956e-06,
+      "loss": 0.678,
+      "step": 24388
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.997395329267766e-06,
+      "loss": 0.7791,
+      "step": 24389
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.996613867176964e-06,
+      "loss": 0.8131,
+      "step": 24390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995832434199228e-06,
+      "loss": 0.9487,
+      "step": 24391
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995051030340242e-06,
+      "loss": 0.6986,
+      "step": 24392
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9942696556056905e-06,
+      "loss": 0.8642,
+      "step": 24393
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.993488310001254e-06,
+      "loss": 1.0735,
+      "step": 24394
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.992706993532616e-06,
+      "loss": 0.709,
+      "step": 24395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991925706205458e-06,
+      "loss": 0.929,
+      "step": 24396
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991144448025462e-06,
+      "loss": 1.041,
+      "step": 24397
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.990363218998308e-06,
+      "loss": 0.9197,
+      "step": 24398
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.989582019129676e-06,
+      "loss": 0.8758,
+      "step": 24399
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.988800848425256e-06,
+      "loss": 0.7122,
+      "step": 24400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9880197068907155e-06,
+      "loss": 1.0555,
+      "step": 24401
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.987238594531743e-06,
+      "loss": 0.7064,
+      "step": 24402
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.986457511354016e-06,
+      "loss": 0.7804,
+      "step": 24403
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.985676457363214e-06,
+      "loss": 0.8407,
+      "step": 24404
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.984895432565022e-06,
+      "loss": 0.7085,
+      "step": 24405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.984114436965116e-06,
+      "loss": 0.9543,
+      "step": 24406
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983333470569174e-06,
+      "loss": 0.8128,
+      "step": 24407
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.982552533382877e-06,
+      "loss": 1.1883,
+      "step": 24408
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.981771625411905e-06,
+      "loss": 0.5847,
+      "step": 24409
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9809907466619386e-06,
+      "loss": 0.9846,
+      "step": 24410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.980209897138651e-06,
+      "loss": 0.8575,
+      "step": 24411
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.979429076847724e-06,
+      "loss": 0.6294,
+      "step": 24412
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.978648285794832e-06,
+      "loss": 0.631,
+      "step": 24413
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.97786752398566e-06,
+      "loss": 0.6937,
+      "step": 24414
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.977086791425881e-06,
+      "loss": 0.8985,
+      "step": 24415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.976306088121174e-06,
+      "loss": 0.8899,
+      "step": 24416
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9755254140772145e-06,
+      "loss": 1.1596,
+      "step": 24417
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.974744769299682e-06,
+      "loss": 0.8163,
+      "step": 24418
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973964153794252e-06,
+      "loss": 0.8098,
+      "step": 24419
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.973183567566605e-06,
+      "loss": 0.9646,
+      "step": 24420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.972403010622412e-06,
+      "loss": 0.8367,
+      "step": 24421
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.971622482967352e-06,
+      "loss": 0.8631,
+      "step": 24422
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9708419846070985e-06,
+      "loss": 1.1622,
+      "step": 24423
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.970061515547331e-06,
+      "loss": 0.7848,
+      "step": 24424
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.969281075793724e-06,
+      "loss": 0.7215,
+      "step": 24425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9685006653519505e-06,
+      "loss": 0.8968,
+      "step": 24426
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.967720284227689e-06,
+      "loss": 0.7297,
+      "step": 24427
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.966939932426614e-06,
+      "loss": 0.9534,
+      "step": 24428
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9661596099544e-06,
+      "loss": 0.7858,
+      "step": 24429
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.965379316816724e-06,
+      "loss": 1.1259,
+      "step": 24430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.964599053019254e-06,
+      "loss": 0.9476,
+      "step": 24431
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.963818818567667e-06,
+      "loss": 0.7652,
+      "step": 24432
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.96303861346764e-06,
+      "loss": 0.8497,
+      "step": 24433
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.962258437724844e-06,
+      "loss": 0.8857,
+      "step": 24434
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.96147829134495e-06,
+      "loss": 0.8865,
+      "step": 24435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.960698174333637e-06,
+      "loss": 0.8099,
+      "step": 24436
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.959918086696577e-06,
+      "loss": 1.0414,
+      "step": 24437
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9591380284394396e-06,
+      "loss": 0.8807,
+      "step": 24438
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.958357999567896e-06,
+      "loss": 0.6651,
+      "step": 24439
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.957578000087629e-06,
+      "loss": 0.9091,
+      "step": 24440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9567980300043e-06,
+      "loss": 0.8905,
+      "step": 24441
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9560180893235845e-06,
+      "loss": 1.1513,
+      "step": 24442
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.955238178051154e-06,
+      "loss": 0.8996,
+      "step": 24443
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.95445829619268e-06,
+      "loss": 0.9112,
+      "step": 24444
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.953678443753836e-06,
+      "loss": 0.9223,
+      "step": 24445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9528986207402906e-06,
+      "loss": 0.6927,
+      "step": 24446
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.952118827157718e-06,
+      "loss": 0.862,
+      "step": 24447
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.951339063011783e-06,
+      "loss": 1.0721,
+      "step": 24448
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.950559328308163e-06,
+      "loss": 0.7747,
+      "step": 24449
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.949779623052526e-06,
+      "loss": 1.3422,
+      "step": 24450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.94899994725054e-06,
+      "loss": 0.9596,
+      "step": 24451
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.948220300907877e-06,
+      "loss": 0.9782,
+      "step": 24452
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947440684030203e-06,
+      "loss": 0.9997,
+      "step": 24453
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.946661096623193e-06,
+      "loss": 0.8794,
+      "step": 24454
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.945881538692513e-06,
+      "loss": 0.9371,
+      "step": 24455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.945102010243833e-06,
+      "loss": 0.4477,
+      "step": 24456
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.944322511282819e-06,
+      "loss": 0.592,
+      "step": 24457
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.943543041815145e-06,
+      "loss": 0.8161,
+      "step": 24458
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.942763601846476e-06,
+      "loss": 0.8208,
+      "step": 24459
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.941984191382483e-06,
+      "loss": 0.9078,
+      "step": 24460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9412048104288285e-06,
+      "loss": 0.6142,
+      "step": 24461
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9404254589911824e-06,
+      "loss": 0.8596,
+      "step": 24462
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.939646137075215e-06,
+      "loss": 0.9675,
+      "step": 24463
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.938866844686592e-06,
+      "loss": 0.4512,
+      "step": 24464
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9380875818309805e-06,
+      "loss": 0.7526,
+      "step": 24465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.937308348514044e-06,
+      "loss": 1.0741,
+      "step": 24466
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.936529144741454e-06,
+      "loss": 0.715,
+      "step": 24467
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.935749970518876e-06,
+      "loss": 0.8533,
+      "step": 24468
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934970825851975e-06,
+      "loss": 0.6295,
+      "step": 24469
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934191710746422e-06,
+      "loss": 0.7973,
+      "step": 24470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.933412625207873e-06,
+      "loss": 0.9016,
+      "step": 24471
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.932633569242e-06,
+      "loss": 0.7432,
+      "step": 24472
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.931854542854467e-06,
+      "loss": 0.8437,
+      "step": 24473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.931075546050942e-06,
+      "loss": 0.9463,
+      "step": 24474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.930296578837083e-06,
+      "loss": 0.999,
+      "step": 24475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.929517641218563e-06,
+      "loss": 0.8491,
+      "step": 24476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.928738733201042e-06,
+      "loss": 1.0064,
+      "step": 24477
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.927959854790182e-06,
+      "loss": 0.7618,
+      "step": 24478
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.927181005991654e-06,
+      "loss": 0.811,
+      "step": 24479
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.926402186811118e-06,
+      "loss": 0.6036,
+      "step": 24480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.925623397254237e-06,
+      "loss": 0.7707,
+      "step": 24481
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.924844637326674e-06,
+      "loss": 0.8747,
+      "step": 24482
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9240659070340955e-06,
+      "loss": 1.0713,
+      "step": 24483
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.923287206382159e-06,
+      "loss": 0.9232,
+      "step": 24484
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.922508535376532e-06,
+      "loss": 0.9687,
+      "step": 24485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9217298940228765e-06,
+      "loss": 0.9375,
+      "step": 24486
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920951282326852e-06,
+      "loss": 0.9135,
+      "step": 24487
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920172700294125e-06,
+      "loss": 0.8281,
+      "step": 24488
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.919394147930355e-06,
+      "loss": 1.3551,
+      "step": 24489
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.918615625241206e-06,
+      "loss": 0.9015,
+      "step": 24490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.917837132232334e-06,
+      "loss": 0.8661,
+      "step": 24491
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.917058668909406e-06,
+      "loss": 1.2314,
+      "step": 24492
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9162802352780775e-06,
+      "loss": 0.7309,
+      "step": 24493
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.915501831344015e-06,
+      "loss": 1.1438,
+      "step": 24494
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.914723457112877e-06,
+      "loss": 0.6167,
+      "step": 24495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9139451125903205e-06,
+      "loss": 1.0033,
+      "step": 24496
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.913166797782011e-06,
+      "loss": 0.8746,
+      "step": 24497
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9123885126936075e-06,
+      "loss": 0.9176,
+      "step": 24498
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.911610257330768e-06,
+      "loss": 0.7065,
+      "step": 24499
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910832031699154e-06,
+      "loss": 0.6222,
+      "step": 24500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910053835804424e-06,
+      "loss": 0.733,
+      "step": 24501
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.909275669652232e-06,
+      "loss": 0.9749,
+      "step": 24502
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9084975332482456e-06,
+      "loss": 0.7578,
+      "step": 24503
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.907719426598117e-06,
+      "loss": 1.2009,
+      "step": 24504
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906941349707506e-06,
+      "loss": 1.1301,
+      "step": 24505
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906163302582076e-06,
+      "loss": 0.8798,
+      "step": 24506
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.905385285227481e-06,
+      "loss": 0.843,
+      "step": 24507
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904607297649376e-06,
+      "loss": 1.1271,
+      "step": 24508
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.903829339853422e-06,
+      "loss": 0.7143,
+      "step": 24509
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.903051411845282e-06,
+      "loss": 0.6412,
+      "step": 24510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.902273513630598e-06,
+      "loss": 0.9624,
+      "step": 24511
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.901495645215042e-06,
+      "loss": 0.9695,
+      "step": 24512
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.900717806604263e-06,
+      "loss": 0.8742,
+      "step": 24513
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8999399978039175e-06,
+      "loss": 0.972,
+      "step": 24514
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8991622188196655e-06,
+      "loss": 0.65,
+      "step": 24515
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.898384469657162e-06,
+      "loss": 0.6913,
+      "step": 24516
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.89760675032206e-06,
+      "loss": 1.0374,
+      "step": 24517
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.896829060820018e-06,
+      "loss": 1.2212,
+      "step": 24518
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8960514011566905e-06,
+      "loss": 0.9184,
+      "step": 24519
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.895273771337735e-06,
+      "loss": 0.977,
+      "step": 24520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.894496171368804e-06,
+      "loss": 0.6871,
+      "step": 24521
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8937186012555516e-06,
+      "loss": 0.9856,
+      "step": 24522
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892941061003632e-06,
+      "loss": 0.7639,
+      "step": 24523
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892163550618704e-06,
+      "loss": 1.012,
+      "step": 24524
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8913860701064175e-06,
+      "loss": 1.0055,
+      "step": 24525
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8906086194724276e-06,
+      "loss": 0.6437,
+      "step": 24526
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889831198722387e-06,
+      "loss": 0.8057,
+      "step": 24527
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889053807861952e-06,
+      "loss": 1.0262,
+      "step": 24528
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8882764468967744e-06,
+      "loss": 0.97,
+      "step": 24529
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.887499115832509e-06,
+      "loss": 0.9926,
+      "step": 24530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.886721814674804e-06,
+      "loss": 0.8736,
+      "step": 24531
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.885944543429314e-06,
+      "loss": 0.9367,
+      "step": 24532
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.885167302101693e-06,
+      "loss": 0.7169,
+      "step": 24533
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.884390090697593e-06,
+      "loss": 0.9254,
+      "step": 24534
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.883612909222665e-06,
+      "loss": 0.9648,
+      "step": 24535
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.882835757682559e-06,
+      "loss": 0.6169,
+      "step": 24536
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88205863608293e-06,
+      "loss": 0.7053,
+      "step": 24537
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8812815444294295e-06,
+      "loss": 0.8528,
+      "step": 24538
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.880504482727704e-06,
+      "loss": 0.8064,
+      "step": 24539
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.879727450983412e-06,
+      "loss": 0.6902,
+      "step": 24540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.878950449202196e-06,
+      "loss": 0.8003,
+      "step": 24541
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.878173477389711e-06,
+      "loss": 0.8446,
+      "step": 24542
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.877396535551606e-06,
+      "loss": 0.7963,
+      "step": 24543
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8766196236935314e-06,
+      "loss": 0.8053,
+      "step": 24544
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.875842741821134e-06,
+      "loss": 0.5252,
+      "step": 24545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8750658899400695e-06,
+      "loss": 0.7873,
+      "step": 24546
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8742890680559825e-06,
+      "loss": 0.7469,
+      "step": 24547
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8735122761745225e-06,
+      "loss": 1.0415,
+      "step": 24548
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.872735514301342e-06,
+      "loss": 1.0169,
+      "step": 24549
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.871958782442089e-06,
+      "loss": 0.7731,
+      "step": 24550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8711820806024065e-06,
+      "loss": 1.0565,
+      "step": 24551
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870405408787947e-06,
+      "loss": 0.9657,
+      "step": 24552
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.869628767004358e-06,
+      "loss": 0.893,
+      "step": 24553
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.868852155257287e-06,
+      "loss": 0.9188,
+      "step": 24554
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.868075573552383e-06,
+      "loss": 0.8083,
+      "step": 24555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.867299021895292e-06,
+      "loss": 1.0632,
+      "step": 24556
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866522500291658e-06,
+      "loss": 0.6574,
+      "step": 24557
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.865746008747135e-06,
+      "loss": 0.8252,
+      "step": 24558
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8649695472673655e-06,
+      "loss": 0.753,
+      "step": 24559
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.864193115857999e-06,
+      "loss": 1.0087,
+      "step": 24560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8634167145246775e-06,
+      "loss": 1.0227,
+      "step": 24561
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8626403432730494e-06,
+      "loss": 0.6313,
+      "step": 24562
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.861864002108757e-06,
+      "loss": 0.7393,
+      "step": 24563
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8610876910374524e-06,
+      "loss": 1.1083,
+      "step": 24564
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.860311410064776e-06,
+      "loss": 0.9165,
+      "step": 24565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.859535159196374e-06,
+      "loss": 0.8436,
+      "step": 24566
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.858758938437894e-06,
+      "loss": 0.9285,
+      "step": 24567
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857982747794979e-06,
+      "loss": 0.7142,
+      "step": 24568
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857206587273273e-06,
+      "loss": 0.7131,
+      "step": 24569
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.856430456878424e-06,
+      "loss": 0.8167,
+      "step": 24570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.85565435661607e-06,
+      "loss": 0.9078,
+      "step": 24571
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854878286491854e-06,
+      "loss": 0.6804,
+      "step": 24572
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854102246511427e-06,
+      "loss": 0.6628,
+      "step": 24573
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.85332623668043e-06,
+      "loss": 0.8708,
+      "step": 24574
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852550257004501e-06,
+      "loss": 0.786,
+      "step": 24575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.851774307489291e-06,
+      "loss": 0.8059,
+      "step": 24576
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850998388140436e-06,
+      "loss": 0.7066,
+      "step": 24577
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850222498963585e-06,
+      "loss": 0.7121,
+      "step": 24578
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.849446639964372e-06,
+      "loss": 1.1906,
+      "step": 24579
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8486708111484515e-06,
+      "loss": 0.7258,
+      "step": 24580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.847895012521452e-06,
+      "loss": 0.9883,
+      "step": 24581
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.847119244089021e-06,
+      "loss": 0.4563,
+      "step": 24582
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.846343505856802e-06,
+      "loss": 0.9414,
+      "step": 24583
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8455677978304315e-06,
+      "loss": 0.7852,
+      "step": 24584
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.844792120015556e-06,
+      "loss": 0.8505,
+      "step": 24585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8440164724178125e-06,
+      "loss": 0.646,
+      "step": 24586
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8432408550428445e-06,
+      "loss": 1.0629,
+      "step": 24587
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.842465267896287e-06,
+      "loss": 0.8943,
+      "step": 24588
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.841689710983787e-06,
+      "loss": 0.7831,
+      "step": 24589
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.840914184310981e-06,
+      "loss": 0.8909,
+      "step": 24590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.84013868788351e-06,
+      "loss": 0.7583,
+      "step": 24591
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.839363221707011e-06,
+      "loss": 0.6526,
+      "step": 24592
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838587785787122e-06,
+      "loss": 1.0638,
+      "step": 24593
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.837812380129488e-06,
+      "loss": 1.1011,
+      "step": 24594
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.83703700473974e-06,
+      "loss": 1.1115,
+      "step": 24595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.836261659623526e-06,
+      "loss": 1.2049,
+      "step": 24596
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8354863447864784e-06,
+      "loss": 0.9146,
+      "step": 24597
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.834711060234237e-06,
+      "loss": 0.721,
+      "step": 24598
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833935805972438e-06,
+      "loss": 0.896,
+      "step": 24599
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833160582006722e-06,
+      "loss": 0.8416,
+      "step": 24600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.832385388342725e-06,
+      "loss": 0.6696,
+      "step": 24601
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.831610224986083e-06,
+      "loss": 0.7164,
+      "step": 24602
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.830835091942435e-06,
+      "loss": 0.9456,
+      "step": 24603
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.830059989217417e-06,
+      "loss": 0.8298,
+      "step": 24604
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.829284916816666e-06,
+      "loss": 0.9412,
+      "step": 24605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.828509874745818e-06,
+      "loss": 0.7601,
+      "step": 24606
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.827734863010506e-06,
+      "loss": 0.9108,
+      "step": 24607
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826959881616373e-06,
+      "loss": 0.7691,
+      "step": 24608
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826184930569052e-06,
+      "loss": 0.6092,
+      "step": 24609
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.825410009874181e-06,
+      "loss": 0.869,
+      "step": 24610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8246351195373875e-06,
+      "loss": 0.9856,
+      "step": 24611
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.823860259564311e-06,
+      "loss": 0.7051,
+      "step": 24612
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.823085429960583e-06,
+      "loss": 0.9859,
+      "step": 24613
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.822310630731846e-06,
+      "loss": 0.7937,
+      "step": 24614
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.821535861883729e-06,
+      "loss": 0.8469,
+      "step": 24615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.820761123421867e-06,
+      "loss": 0.9977,
+      "step": 24616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.819986415351896e-06,
+      "loss": 0.95,
+      "step": 24617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8192117376794464e-06,
+      "loss": 1.0619,
+      "step": 24618
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.818437090410154e-06,
+      "loss": 1.0877,
+      "step": 24619
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8176624735496515e-06,
+      "loss": 1.0451,
+      "step": 24620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.816887887103572e-06,
+      "loss": 1.036,
+      "step": 24621
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.816113331077548e-06,
+      "loss": 1.0051,
+      "step": 24622
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.815338805477212e-06,
+      "loss": 1.0494,
+      "step": 24623
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.814564310308197e-06,
+      "loss": 0.7247,
+      "step": 24624
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.813789845576132e-06,
+      "loss": 0.8695,
+      "step": 24625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.813015411286656e-06,
+      "loss": 0.9038,
+      "step": 24626
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.812241007445396e-06,
+      "loss": 0.7707,
+      "step": 24627
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.811466634057985e-06,
+      "loss": 0.7686,
+      "step": 24628
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.810692291130054e-06,
+      "loss": 0.7386,
+      "step": 24629
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8099179786672365e-06,
+      "loss": 0.9857,
+      "step": 24630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.809143696675155e-06,
+      "loss": 0.6795,
+      "step": 24631
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.808369445159447e-06,
+      "loss": 0.8632,
+      "step": 24632
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8075952241257434e-06,
+      "loss": 0.7671,
+      "step": 24633
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.806821033579673e-06,
+      "loss": 0.8618,
+      "step": 24634
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.806046873526867e-06,
+      "loss": 0.732,
+      "step": 24635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.805272743972952e-06,
+      "loss": 0.8647,
+      "step": 24636
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.80449864492356e-06,
+      "loss": 0.6937,
+      "step": 24637
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.803724576384315e-06,
+      "loss": 0.8442,
+      "step": 24638
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8029505383608565e-06,
+      "loss": 0.581,
+      "step": 24639
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.802176530858812e-06,
+      "loss": 0.8445,
+      "step": 24640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8014025538837996e-06,
+      "loss": 0.7062,
+      "step": 24641
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.800628607441457e-06,
+      "loss": 1.0558,
+      "step": 24642
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.799854691537402e-06,
+      "loss": 0.8848,
+      "step": 24643
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7990808061772765e-06,
+      "loss": 0.9664,
+      "step": 24644
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.798306951366701e-06,
+      "loss": 0.8173,
+      "step": 24645
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.797533127111303e-06,
+      "loss": 0.5418,
+      "step": 24646
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.796759333416713e-06,
+      "loss": 0.7062,
+      "step": 24647
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7959855702885545e-06,
+      "loss": 0.683,
+      "step": 24648
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.795211837732457e-06,
+      "loss": 0.6325,
+      "step": 24649
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.794438135754045e-06,
+      "loss": 0.9358,
+      "step": 24650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.793664464358945e-06,
+      "loss": 0.7838,
+      "step": 24651
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.792890823552786e-06,
+      "loss": 0.967,
+      "step": 24652
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7921172133411904e-06,
+      "loss": 1.0421,
+      "step": 24653
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.791343633729787e-06,
+      "loss": 0.7943,
+      "step": 24654
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7905700847242e-06,
+      "loss": 0.6397,
+      "step": 24655
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.78979656633005e-06,
+      "loss": 1.0464,
+      "step": 24656
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789023078552971e-06,
+      "loss": 0.7841,
+      "step": 24657
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.788249621398585e-06,
+      "loss": 0.6417,
+      "step": 24658
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787476194872514e-06,
+      "loss": 1.0076,
+      "step": 24659
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.786702798980388e-06,
+      "loss": 0.6928,
+      "step": 24660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.785929433727819e-06,
+      "loss": 0.8849,
+      "step": 24661
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.785156099120443e-06,
+      "loss": 0.6376,
+      "step": 24662
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.784382795163882e-06,
+      "loss": 1.1012,
+      "step": 24663
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.783609521863755e-06,
+      "loss": 0.7792,
+      "step": 24664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.78283627922569e-06,
+      "loss": 0.9497,
+      "step": 24665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.782063067255306e-06,
+      "loss": 0.7566,
+      "step": 24666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7812898859582285e-06,
+      "loss": 0.8949,
+      "step": 24667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.780516735340076e-06,
+      "loss": 1.0513,
+      "step": 24668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.779743615406477e-06,
+      "loss": 0.8629,
+      "step": 24669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7789705261630565e-06,
+      "loss": 1.1184,
+      "step": 24670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.778197467615425e-06,
+      "loss": 0.7162,
+      "step": 24671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777424439769213e-06,
+      "loss": 0.7648,
+      "step": 24672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.776651442630037e-06,
+      "loss": 0.8028,
+      "step": 24673
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.775878476203518e-06,
+      "loss": 0.7541,
+      "step": 24674
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.775105540495284e-06,
+      "loss": 0.7654,
+      "step": 24675
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7743326355109504e-06,
+      "loss": 0.6977,
+      "step": 24676
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.77355976125614e-06,
+      "loss": 0.8597,
+      "step": 24677
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.772786917736473e-06,
+      "loss": 1.0159,
+      "step": 24678
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7720141049575685e-06,
+      "loss": 0.765,
+      "step": 24679
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771241322925046e-06,
+      "loss": 0.5282,
+      "step": 24680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.770468571644527e-06,
+      "loss": 0.7249,
+      "step": 24681
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.769695851121631e-06,
+      "loss": 0.8539,
+      "step": 24682
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.768923161361975e-06,
+      "loss": 0.8028,
+      "step": 24683
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76815050237118e-06,
+      "loss": 0.6497,
+      "step": 24684
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.767377874154865e-06,
+      "loss": 0.8697,
+      "step": 24685
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.766605276718649e-06,
+      "loss": 0.9268,
+      "step": 24686
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.765832710068143e-06,
+      "loss": 0.9251,
+      "step": 24687
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.765060174208978e-06,
+      "loss": 0.7436,
+      "step": 24688
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.764287669146764e-06,
+      "loss": 0.6167,
+      "step": 24689
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.763515194887126e-06,
+      "loss": 0.803,
+      "step": 24690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76274275143567e-06,
+      "loss": 0.8207,
+      "step": 24691
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7619703387980175e-06,
+      "loss": 0.86,
+      "step": 24692
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.761197956979791e-06,
+      "loss": 0.916,
+      "step": 24693
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7604256059866015e-06,
+      "loss": 0.7839,
+      "step": 24694
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.75965328582407e-06,
+      "loss": 0.868,
+      "step": 24695
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.75888099649781e-06,
+      "loss": 0.647,
+      "step": 24696
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.758108738013438e-06,
+      "loss": 0.442,
+      "step": 24697
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.75733651037657e-06,
+      "loss": 0.6727,
+      "step": 24698
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.756564313592819e-06,
+      "loss": 1.064,
+      "step": 24699
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.755792147667811e-06,
+      "loss": 1.1168,
+      "step": 24700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.755020012607151e-06,
+      "loss": 1.085,
+      "step": 24701
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.754247908416455e-06,
+      "loss": 0.8288,
+      "step": 24702
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753475835101341e-06,
+      "loss": 1.0555,
+      "step": 24703
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.752703792667422e-06,
+      "loss": 0.9608,
+      "step": 24704
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.751931781120308e-06,
+      "loss": 0.6908,
+      "step": 24705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7511598004656224e-06,
+      "loss": 0.797,
+      "step": 24706
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.750387850708974e-06,
+      "loss": 0.876,
+      "step": 24707
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.749615931855979e-06,
+      "loss": 0.8373,
+      "step": 24708
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7488440439122466e-06,
+      "loss": 0.9566,
+      "step": 24709
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.748072186883393e-06,
+      "loss": 0.8111,
+      "step": 24710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.747300360775031e-06,
+      "loss": 0.789,
+      "step": 24711
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.746528565592772e-06,
+      "loss": 0.8756,
+      "step": 24712
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.745756801342231e-06,
+      "loss": 0.712,
+      "step": 24713
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.744985068029019e-06,
+      "loss": 0.7392,
+      "step": 24714
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.744213365658747e-06,
+      "loss": 0.8226,
+      "step": 24715
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.743441694237028e-06,
+      "loss": 0.6858,
+      "step": 24716
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.742670053769472e-06,
+      "loss": 0.9713,
+      "step": 24717
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.741898444261693e-06,
+      "loss": 0.8482,
+      "step": 24718
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.741126865719303e-06,
+      "loss": 0.8805,
+      "step": 24719
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740355318147916e-06,
+      "loss": 0.9009,
+      "step": 24720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.739583801553132e-06,
+      "loss": 1.0127,
+      "step": 24721
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.73881231594057e-06,
+      "loss": 0.8754,
+      "step": 24722
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.738040861315832e-06,
+      "loss": 1.1473,
+      "step": 24723
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.737269437684539e-06,
+      "loss": 0.9907,
+      "step": 24724
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.736498045052296e-06,
+      "loss": 1.0814,
+      "step": 24725
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.735726683424714e-06,
+      "loss": 1.1937,
+      "step": 24726
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734955352807399e-06,
+      "loss": 0.7528,
+      "step": 24727
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734184053205964e-06,
+      "loss": 1.0558,
+      "step": 24728
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.733412784626016e-06,
+      "loss": 1.0463,
+      "step": 24729
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.732641547073163e-06,
+      "loss": 1.2685,
+      "step": 24730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.731870340553014e-06,
+      "loss": 0.6378,
+      "step": 24731
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.731099165071179e-06,
+      "loss": 1.1072,
+      "step": 24732
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.730328020633266e-06,
+      "loss": 0.8955,
+      "step": 24733
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7295569072448796e-06,
+      "loss": 0.6466,
+      "step": 24734
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.728785824911627e-06,
+      "loss": 0.7494,
+      "step": 24735
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.728014773639121e-06,
+      "loss": 0.887,
+      "step": 24736
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7272437534329674e-06,
+      "loss": 0.8186,
+      "step": 24737
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.726472764298771e-06,
+      "loss": 0.9714,
+      "step": 24738
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.725701806242139e-06,
+      "loss": 1.0283,
+      "step": 24739
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.724930879268682e-06,
+      "loss": 0.7274,
+      "step": 24740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.724159983383996e-06,
+      "loss": 0.642,
+      "step": 24741
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.723389118593696e-06,
+      "loss": 1.0441,
+      "step": 24742
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.722618284903385e-06,
+      "loss": 1.0572,
+      "step": 24743
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.72184748231867e-06,
+      "loss": 0.9109,
+      "step": 24744
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.721076710845154e-06,
+      "loss": 1.1754,
+      "step": 24745
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.720305970488444e-06,
+      "loss": 0.9543,
+      "step": 24746
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.719535261254143e-06,
+      "loss": 0.8254,
+      "step": 24747
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.718764583147855e-06,
+      "loss": 0.8871,
+      "step": 24748
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.717993936175188e-06,
+      "loss": 0.7116,
+      "step": 24749
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.717223320341751e-06,
+      "loss": 0.8483,
+      "step": 24750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.716452735653135e-06,
+      "loss": 1.0112,
+      "step": 24751
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.715682182114952e-06,
+      "loss": 0.6804,
+      "step": 24752
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7149116597328006e-06,
+      "loss": 0.8578,
+      "step": 24753
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71414116851229e-06,
+      "loss": 0.9058,
+      "step": 24754
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.713370708459022e-06,
+      "loss": 1.2546,
+      "step": 24755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.712600279578597e-06,
+      "loss": 0.8554,
+      "step": 24756
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71182988187662e-06,
+      "loss": 0.6143,
+      "step": 24757
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.711059515358692e-06,
+      "loss": 1.1962,
+      "step": 24758
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.710289180030415e-06,
+      "loss": 0.8987,
+      "step": 24759
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.709518875897392e-06,
+      "loss": 0.7556,
+      "step": 24760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.708748602965224e-06,
+      "loss": 0.8558,
+      "step": 24761
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.707978361239512e-06,
+      "loss": 0.9941,
+      "step": 24762
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.707208150725858e-06,
+      "loss": 0.8395,
+      "step": 24763
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.706437971429865e-06,
+      "loss": 0.8359,
+      "step": 24764
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.70566782335713e-06,
+      "loss": 0.8698,
+      "step": 24765
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.704897706513251e-06,
+      "loss": 0.9805,
+      "step": 24766
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.704127620903838e-06,
+      "loss": 0.9538,
+      "step": 24767
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.703357566534486e-06,
+      "loss": 0.6768,
+      "step": 24768
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7025875434107945e-06,
+      "loss": 0.9126,
+      "step": 24769
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701817551538367e-06,
+      "loss": 0.9699,
+      "step": 24770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701047590922793e-06,
+      "loss": 1.1812,
+      "step": 24771
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.700277661569682e-06,
+      "loss": 0.7384,
+      "step": 24772
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.699507763484629e-06,
+      "loss": 0.9024,
+      "step": 24773
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.698737896673232e-06,
+      "loss": 0.903,
+      "step": 24774
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.697968061141092e-06,
+      "loss": 1.0162,
+      "step": 24775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.697198256893806e-06,
+      "loss": 1.0504,
+      "step": 24776
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696428483936972e-06,
+      "loss": 0.4099,
+      "step": 24777
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.695658742276184e-06,
+      "loss": 0.8672,
+      "step": 24778
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.694889031917047e-06,
+      "loss": 0.8159,
+      "step": 24779
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.69411935286516e-06,
+      "loss": 0.9779,
+      "step": 24780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.69334970512611e-06,
+      "loss": 0.9626,
+      "step": 24781
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6925800887055e-06,
+      "loss": 0.8266,
+      "step": 24782
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.691810503608925e-06,
+      "loss": 0.6694,
+      "step": 24783
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.691040949841979e-06,
+      "loss": 0.7234,
+      "step": 24784
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6902714274102665e-06,
+      "loss": 0.6095,
+      "step": 24785
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.689501936319379e-06,
+      "loss": 0.8884,
+      "step": 24786
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.68873247657491e-06,
+      "loss": 0.8344,
+      "step": 24787
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.687963048182459e-06,
+      "loss": 0.9978,
+      "step": 24788
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.687193651147619e-06,
+      "loss": 1.0051,
+      "step": 24789
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.686424285475985e-06,
+      "loss": 0.8921,
+      "step": 24790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.685654951173154e-06,
+      "loss": 0.5208,
+      "step": 24791
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.684885648244718e-06,
+      "loss": 0.8789,
+      "step": 24792
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.684116376696273e-06,
+      "loss": 0.9833,
+      "step": 24793
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.683347136533413e-06,
+      "loss": 0.877,
+      "step": 24794
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.682577927761732e-06,
+      "loss": 0.6301,
+      "step": 24795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6818087503868215e-06,
+      "loss": 0.9735,
+      "step": 24796
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681039604414279e-06,
+      "loss": 0.7725,
+      "step": 24797
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.680270489849697e-06,
+      "loss": 0.9538,
+      "step": 24798
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.679501406698667e-06,
+      "loss": 0.4845,
+      "step": 24799
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.678732354966787e-06,
+      "loss": 1.0366,
+      "step": 24800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.677963334659642e-06,
+      "loss": 0.768,
+      "step": 24801
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6771943457828225e-06,
+      "loss": 1.2076,
+      "step": 24802
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.676425388341931e-06,
+      "loss": 0.8818,
+      "step": 24803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.675656462342553e-06,
+      "loss": 0.7558,
+      "step": 24804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.674887567790282e-06,
+      "loss": 0.7794,
+      "step": 24805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.674118704690708e-06,
+      "loss": 1.1275,
+      "step": 24806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.673349873049424e-06,
+      "loss": 0.9661,
+      "step": 24807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.67258107287202e-06,
+      "loss": 0.6828,
+      "step": 24808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671812304164084e-06,
+      "loss": 0.4406,
+      "step": 24809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671043566931216e-06,
+      "loss": 0.9973,
+      "step": 24810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.670274861178996e-06,
+      "loss": 0.8864,
+      "step": 24811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.669506186913018e-06,
+      "loss": 0.9519,
+      "step": 24812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668737544138874e-06,
+      "loss": 0.9514,
+      "step": 24813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.667968932862144e-06,
+      "loss": 0.721,
+      "step": 24814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6672003530884316e-06,
+      "loss": 0.7309,
+      "step": 24815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.666431804823318e-06,
+      "loss": 0.9024,
+      "step": 24816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.665663288072395e-06,
+      "loss": 0.7928,
+      "step": 24817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6648948028412496e-06,
+      "loss": 0.9001,
+      "step": 24818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6641263491354705e-06,
+      "loss": 0.688,
+      "step": 24819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6633579269606445e-06,
+      "loss": 1.1107,
+      "step": 24820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.662589536322363e-06,
+      "loss": 0.729,
+      "step": 24821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.661821177226211e-06,
+      "loss": 0.8597,
+      "step": 24822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.661052849677777e-06,
+      "loss": 0.5481,
+      "step": 24823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.660284553682649e-06,
+      "loss": 1.0765,
+      "step": 24824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.659516289246414e-06,
+      "loss": 0.8132,
+      "step": 24825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.658748056374659e-06,
+      "loss": 0.8163,
+      "step": 24826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.657979855072966e-06,
+      "loss": 0.957,
+      "step": 24827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.657211685346928e-06,
+      "loss": 0.885,
+      "step": 24828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.65644354720213e-06,
+      "loss": 0.9081,
+      "step": 24829
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6556754406441615e-06,
+      "loss": 1.0462,
+      "step": 24830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.654907365678599e-06,
+      "loss": 0.8282,
+      "step": 24831
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.654139322311029e-06,
+      "loss": 0.8188,
+      "step": 24832
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.653371310547045e-06,
+      "loss": 0.8319,
+      "step": 24833
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652603330392225e-06,
+      "loss": 0.5969,
+      "step": 24834
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.651835381852158e-06,
+      "loss": 0.9739,
+      "step": 24835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.651067464932427e-06,
+      "loss": 0.5971,
+      "step": 24836
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6502995796386165e-06,
+      "loss": 0.8577,
+      "step": 24837
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.64953172597631e-06,
+      "loss": 0.9015,
+      "step": 24838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6487639039510885e-06,
+      "loss": 0.8318,
+      "step": 24839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.647996113568547e-06,
+      "loss": 0.6785,
+      "step": 24840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.647228354834256e-06,
+      "loss": 0.9684,
+      "step": 24841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.646460627753805e-06,
+      "loss": 0.9173,
+      "step": 24842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.645692932332776e-06,
+      "loss": 0.5145,
+      "step": 24843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.644925268576751e-06,
+      "loss": 0.7821,
+      "step": 24844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.644157636491307e-06,
+      "loss": 0.7479,
+      "step": 24845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.643390036082039e-06,
+      "loss": 1.0661,
+      "step": 24846
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.642622467354522e-06,
+      "loss": 1.0315,
+      "step": 24847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.641854930314338e-06,
+      "loss": 1.0362,
+      "step": 24848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.641087424967069e-06,
+      "loss": 0.8095,
+      "step": 24849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6403199513182964e-06,
+      "loss": 0.9897,
+      "step": 24850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6395525093736006e-06,
+      "loss": 0.7069,
+      "step": 24851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638785099138564e-06,
+      "loss": 0.7823,
+      "step": 24852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638017720618766e-06,
+      "loss": 0.8747,
+      "step": 24853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.637250373819787e-06,
+      "loss": 0.7997,
+      "step": 24854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.636483058747209e-06,
+      "loss": 0.9135,
+      "step": 24855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.635715775406609e-06,
+      "loss": 0.7799,
+      "step": 24856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6349485238035674e-06,
+      "loss": 0.7355,
+      "step": 24857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6341813039436665e-06,
+      "loss": 0.8302,
+      "step": 24858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.633414115832485e-06,
+      "loss": 0.8436,
+      "step": 24859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.632646959475604e-06,
+      "loss": 0.5992,
+      "step": 24860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.631879834878596e-06,
+      "loss": 0.9316,
+      "step": 24861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6311127420470425e-06,
+      "loss": 0.5884,
+      "step": 24862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.630345680986519e-06,
+      "loss": 0.8643,
+      "step": 24863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.62957865170261e-06,
+      "loss": 1.0908,
+      "step": 24864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628811654200891e-06,
+      "loss": 0.9519,
+      "step": 24865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6280446884869385e-06,
+      "loss": 0.584,
+      "step": 24866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.627277754566331e-06,
+      "loss": 0.8406,
+      "step": 24867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.626510852444645e-06,
+      "loss": 0.9146,
+      "step": 24868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.625743982127459e-06,
+      "loss": 0.5717,
+      "step": 24869
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.624977143620347e-06,
+      "loss": 0.7463,
+      "step": 24870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.624210336928887e-06,
+      "loss": 0.8094,
+      "step": 24871
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.623443562058656e-06,
+      "loss": 1.0427,
+      "step": 24872
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.622676819015229e-06,
+      "loss": 1.1524,
+      "step": 24873
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.621910107804183e-06,
+      "loss": 0.8231,
+      "step": 24874
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.621143428431089e-06,
+      "loss": 1.0126,
+      "step": 24875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.620376780901529e-06,
+      "loss": 1.0625,
+      "step": 24876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6196101652210755e-06,
+      "loss": 0.9651,
+      "step": 24877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618843581395303e-06,
+      "loss": 1.079,
+      "step": 24878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618077029429788e-06,
+      "loss": 0.7614,
+      "step": 24879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.617310509330107e-06,
+      "loss": 0.8842,
+      "step": 24880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.616544021101822e-06,
+      "loss": 0.7442,
+      "step": 24881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6157775647505195e-06,
+      "loss": 1.1818,
+      "step": 24882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.615011140281771e-06,
+      "loss": 0.8696,
+      "step": 24883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.614244747701148e-06,
+      "loss": 0.863,
+      "step": 24884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.613478387014223e-06,
+      "loss": 0.8213,
+      "step": 24885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.61271205822657e-06,
+      "loss": 0.8797,
+      "step": 24886
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611945761343765e-06,
+      "loss": 0.9283,
+      "step": 24887
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6111794963713705e-06,
+      "loss": 0.6832,
+      "step": 24888
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.610413263314971e-06,
+      "loss": 0.589,
+      "step": 24889
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.609647062180139e-06,
+      "loss": 0.8039,
+      "step": 24890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608880892972435e-06,
+      "loss": 0.9995,
+      "step": 24891
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608114755697438e-06,
+      "loss": 1.1319,
+      "step": 24892
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.607348650360715e-06,
+      "loss": 1.0489,
+      "step": 24893
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.606582576967843e-06,
+      "loss": 0.9249,
+      "step": 24894
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.605816535524391e-06,
+      "loss": 0.6911,
+      "step": 24895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.605050526035929e-06,
+      "loss": 0.9077,
+      "step": 24896
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.604284548508028e-06,
+      "loss": 1.1109,
+      "step": 24897
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.603518602946256e-06,
+      "loss": 0.8335,
+      "step": 24898
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.602752689356188e-06,
+      "loss": 0.6986,
+      "step": 24899
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.601986807743388e-06,
+      "loss": 0.7503,
+      "step": 24900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.601220958113429e-06,
+      "loss": 0.6351,
+      "step": 24901
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.600455140471881e-06,
+      "loss": 0.9354,
+      "step": 24902
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.599689354824311e-06,
+      "loss": 0.9849,
+      "step": 24903
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.598923601176288e-06,
+      "loss": 0.7852,
+      "step": 24904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.598157879533383e-06,
+      "loss": 0.9607,
+      "step": 24905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.597392189901156e-06,
+      "loss": 0.9603,
+      "step": 24906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5966265322851875e-06,
+      "loss": 0.9704,
+      "step": 24907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.595860906691041e-06,
+      "loss": 0.8662,
+      "step": 24908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.595095313124281e-06,
+      "loss": 0.9748,
+      "step": 24909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.594329751590483e-06,
+      "loss": 0.8288,
+      "step": 24910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.593564222095198e-06,
+      "loss": 1.0759,
+      "step": 24911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5927987246440085e-06,
+      "loss": 1.0986,
+      "step": 24912
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.592033259242477e-06,
+      "loss": 0.947,
+      "step": 24913
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5912678258961674e-06,
+      "loss": 0.86,
+      "step": 24914
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5905024246106485e-06,
+      "loss": 0.9115,
+      "step": 24915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.589737055391486e-06,
+      "loss": 0.7782,
+      "step": 24916
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.588971718244245e-06,
+      "loss": 0.9126,
+      "step": 24917
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.588206413174488e-06,
+      "loss": 0.9268,
+      "step": 24918
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.587441140187788e-06,
+      "loss": 1.0434,
+      "step": 24919
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.586675899289708e-06,
+      "loss": 0.6704,
+      "step": 24920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.58591069048581e-06,
+      "loss": 0.8081,
+      "step": 24921
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.585145513781658e-06,
+      "loss": 0.8075,
+      "step": 24922
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.584380369182818e-06,
+      "loss": 0.9038,
+      "step": 24923
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5836152566948496e-06,
+      "loss": 1.0551,
+      "step": 24924
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.582850176323327e-06,
+      "loss": 1.0106,
+      "step": 24925
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.582085128073808e-06,
+      "loss": 0.4769,
+      "step": 24926
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.581320111951856e-06,
+      "loss": 0.6369,
+      "step": 24927
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.580555127963034e-06,
+      "loss": 0.6484,
+      "step": 24928
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.579790176112908e-06,
+      "loss": 0.9047,
+      "step": 24929
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.579025256407038e-06,
+      "loss": 0.8261,
+      "step": 24930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5782603688509876e-06,
+      "loss": 1.2093,
+      "step": 24931
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.577495513450317e-06,
+      "loss": 1.0165,
+      "step": 24932
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.576730690210592e-06,
+      "loss": 0.9786,
+      "step": 24933
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.575965899137372e-06,
+      "loss": 0.6302,
+      "step": 24934
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.57520114023622e-06,
+      "loss": 0.6232,
+      "step": 24935
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5744364135126915e-06,
+      "loss": 0.8471,
+      "step": 24936
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.573671718972356e-06,
+      "loss": 0.7699,
+      "step": 24937
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572907056620773e-06,
+      "loss": 0.7487,
+      "step": 24938
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572142426463501e-06,
+      "loss": 0.8447,
+      "step": 24939
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.571377828506105e-06,
+      "loss": 0.9696,
+      "step": 24940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.570613262754137e-06,
+      "loss": 0.7855,
+      "step": 24941
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.569848729213157e-06,
+      "loss": 0.964,
+      "step": 24942
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.569084227888734e-06,
+      "loss": 0.9131,
+      "step": 24943
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568319758786422e-06,
+      "loss": 0.6708,
+      "step": 24944
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.567555321911782e-06,
+      "loss": 0.9171,
+      "step": 24945
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.56679091727037e-06,
+      "loss": 1.05,
+      "step": 24946
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.566026544867747e-06,
+      "loss": 0.7646,
+      "step": 24947
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.565262204709472e-06,
+      "loss": 0.9308,
+      "step": 24948
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.564497896801099e-06,
+      "loss": 1.1687,
+      "step": 24949
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.563733621148197e-06,
+      "loss": 0.8564,
+      "step": 24950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.562969377756313e-06,
+      "loss": 0.6567,
+      "step": 24951
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.562205166631008e-06,
+      "loss": 0.9021,
+      "step": 24952
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.561440987777841e-06,
+      "loss": 1.0158,
+      "step": 24953
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.560676841202364e-06,
+      "loss": 0.6613,
+      "step": 24954
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559912726910141e-06,
+      "loss": 0.739,
+      "step": 24955
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559148644906726e-06,
+      "loss": 0.8742,
+      "step": 24956
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.558384595197675e-06,
+      "loss": 0.5096,
+      "step": 24957
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.557620577788545e-06,
+      "loss": 0.7192,
+      "step": 24958
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5568565926848916e-06,
+      "loss": 0.714,
+      "step": 24959
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55609263989227e-06,
+      "loss": 1.0671,
+      "step": 24960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.555328719416235e-06,
+      "loss": 0.8956,
+      "step": 24961
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554564831262346e-06,
+      "loss": 0.8114,
+      "step": 24962
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5538009754361524e-06,
+      "loss": 0.7158,
+      "step": 24963
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.553037151943213e-06,
+      "loss": 0.7619,
+      "step": 24964
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5522733607890825e-06,
+      "loss": 1.0731,
+      "step": 24965
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.551509601979312e-06,
+      "loss": 1.0521,
+      "step": 24966
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5507458755194545e-06,
+      "loss": 0.8402,
+      "step": 24967
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549982181415071e-06,
+      "loss": 0.841,
+      "step": 24968
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549218519671711e-06,
+      "loss": 0.7608,
+      "step": 24969
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.548454890294932e-06,
+      "loss": 0.6792,
+      "step": 24970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.547691293290281e-06,
+      "loss": 0.8128,
+      "step": 24971
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.546927728663307e-06,
+      "loss": 0.8538,
+      "step": 24972
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.546164196419575e-06,
+      "loss": 0.9172,
+      "step": 24973
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.545400696564632e-06,
+      "loss": 1.0282,
+      "step": 24974
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.544637229104027e-06,
+      "loss": 0.9056,
+      "step": 24975
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.543873794043317e-06,
+      "loss": 0.9647,
+      "step": 24976
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.543110391388052e-06,
+      "loss": 0.5696,
+      "step": 24977
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.542347021143781e-06,
+      "loss": 0.9537,
+      "step": 24978
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.541583683316054e-06,
+      "loss": 1.0077,
+      "step": 24979
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.540820377910435e-06,
+      "loss": 0.7824,
+      "step": 24980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.540057104932461e-06,
+      "loss": 0.8463,
+      "step": 24981
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.539293864387687e-06,
+      "loss": 0.9311,
+      "step": 24982
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.538530656281662e-06,
+      "loss": 1.0036,
+      "step": 24983
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537767480619939e-06,
+      "loss": 0.6941,
+      "step": 24984
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537004337408062e-06,
+      "loss": 0.8066,
+      "step": 24985
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5362412266515884e-06,
+      "loss": 0.9525,
+      "step": 24986
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.535478148356065e-06,
+      "loss": 0.8651,
+      "step": 24987
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53471510252704e-06,
+      "loss": 1.1427,
+      "step": 24988
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.533952089170063e-06,
+      "loss": 1.0617,
+      "step": 24989
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.533189108290682e-06,
+      "loss": 0.8421,
+      "step": 24990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.532426159894446e-06,
+      "loss": 0.7202,
+      "step": 24991
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.531663243986902e-06,
+      "loss": 0.8822,
+      "step": 24992
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530900360573601e-06,
+      "loss": 0.8202,
+      "step": 24993
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530137509660087e-06,
+      "loss": 0.9795,
+      "step": 24994
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.529374691251911e-06,
+      "loss": 0.8538,
+      "step": 24995
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.528611905354618e-06,
+      "loss": 1.0789,
+      "step": 24996
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.527849151973752e-06,
+      "loss": 0.9437,
+      "step": 24997
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.527086431114867e-06,
+      "loss": 0.8296,
+      "step": 24998
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.526323742783507e-06,
+      "loss": 0.7632,
+      "step": 24999
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.52556108698522e-06,
+      "loss": 0.7298,
+      "step": 25000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.524798463725547e-06,
+      "loss": 0.5246,
+      "step": 25001
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5240358730100365e-06,
+      "loss": 0.6972,
+      "step": 25002
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523273314844231e-06,
+      "loss": 0.8219,
+      "step": 25003
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.522510789233681e-06,
+      "loss": 0.5624,
+      "step": 25004
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5217482961839305e-06,
+      "loss": 1.302,
+      "step": 25005
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520985835700526e-06,
+      "loss": 0.8755,
+      "step": 25006
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520223407789007e-06,
+      "loss": 0.6535,
+      "step": 25007
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.519461012454921e-06,
+      "loss": 0.9861,
+      "step": 25008
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.518698649703813e-06,
+      "loss": 1.0497,
+      "step": 25009
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.517936319541226e-06,
+      "loss": 0.8254,
+      "step": 25010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.517174021972705e-06,
+      "loss": 0.916,
+      "step": 25011
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.51641175700379e-06,
+      "loss": 0.806,
+      "step": 25012
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.515649524640028e-06,
+      "loss": 0.6982,
+      "step": 25013
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.514887324886959e-06,
+      "loss": 0.7335,
+      "step": 25014
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5141251577501256e-06,
+      "loss": 1.0608,
+      "step": 25015
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.513363023235074e-06,
+      "loss": 0.8976,
+      "step": 25016
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.512600921347346e-06,
+      "loss": 0.629,
+      "step": 25017
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511838852092483e-06,
+      "loss": 0.8391,
+      "step": 25018
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511076815476026e-06,
+      "loss": 0.8396,
+      "step": 25019
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.51031481150352e-06,
+      "loss": 1.0683,
+      "step": 25020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5095528401804966e-06,
+      "loss": 0.7799,
+      "step": 25021
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.508790901512506e-06,
+      "loss": 0.6654,
+      "step": 25022
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.508028995505088e-06,
+      "loss": 0.8412,
+      "step": 25023
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.507267122163781e-06,
+      "loss": 0.9788,
+      "step": 25024
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5065052814941265e-06,
+      "loss": 0.8421,
+      "step": 25025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5057434735016655e-06,
+      "loss": 0.7045,
+      "step": 25026
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.504981698191936e-06,
+      "loss": 1.0769,
+      "step": 25027
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.504219955570475e-06,
+      "loss": 0.8865,
+      "step": 25028
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.50345824564283e-06,
+      "loss": 0.9149,
+      "step": 25029
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5026965684145406e-06,
+      "loss": 0.7847,
+      "step": 25030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.501934923891136e-06,
+      "loss": 0.6612,
+      "step": 25031
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5011733120781605e-06,
+      "loss": 0.7372,
+      "step": 25032
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.500411732981148e-06,
+      "loss": 0.914,
+      "step": 25033
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.499650186605645e-06,
+      "loss": 0.8491,
+      "step": 25034
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.498888672957187e-06,
+      "loss": 0.8783,
+      "step": 25035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.498127192041309e-06,
+      "loss": 0.9277,
+      "step": 25036
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49736574386355e-06,
+      "loss": 0.9617,
+      "step": 25037
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.496604328429448e-06,
+      "loss": 0.678,
+      "step": 25038
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.495842945744539e-06,
+      "loss": 0.5888,
+      "step": 25039
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49508159581436e-06,
+      "loss": 0.6304,
+      "step": 25040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.494320278644449e-06,
+      "loss": 0.8734,
+      "step": 25041
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49355899424034e-06,
+      "loss": 0.753,
+      "step": 25042
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4927977426075715e-06,
+      "loss": 0.7611,
+      "step": 25043
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492036523751678e-06,
+      "loss": 0.7568,
+      "step": 25044
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.491275337678195e-06,
+      "loss": 0.7288,
+      "step": 25045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490514184392655e-06,
+      "loss": 0.9101,
+      "step": 25046
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4897530639006e-06,
+      "loss": 0.4291,
+      "step": 25047
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.488991976207562e-06,
+      "loss": 1.0865,
+      "step": 25048
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.488230921319076e-06,
+      "loss": 0.7571,
+      "step": 25049
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487469899240678e-06,
+      "loss": 0.8373,
+      "step": 25050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.486708909977894e-06,
+      "loss": 1.0968,
+      "step": 25051
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.485947953536266e-06,
+      "loss": 0.8661,
+      "step": 25052
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.485187029921326e-06,
+      "loss": 0.8443,
+      "step": 25053
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.484426139138608e-06,
+      "loss": 1.0586,
+      "step": 25054
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4836652811936456e-06,
+      "loss": 0.6462,
+      "step": 25055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.482904456091969e-06,
+      "loss": 0.7093,
+      "step": 25056
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.482143663839115e-06,
+      "loss": 0.8059,
+      "step": 25057
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4813829044406075e-06,
+      "loss": 0.9952,
+      "step": 25058
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.480622177901991e-06,
+      "loss": 0.8298,
+      "step": 25059
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.479861484228795e-06,
+      "loss": 0.905,
+      "step": 25060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4791008234265445e-06,
+      "loss": 1.0577,
+      "step": 25061
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.478340195500774e-06,
+      "loss": 0.6538,
+      "step": 25062
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.477579600457017e-06,
+      "loss": 0.4818,
+      "step": 25063
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476819038300798e-06,
+      "loss": 0.612,
+      "step": 25064
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476058509037658e-06,
+      "loss": 0.9962,
+      "step": 25065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4752980126731234e-06,
+      "loss": 0.8486,
+      "step": 25066
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.474537549212724e-06,
+      "loss": 0.8807,
+      "step": 25067
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4737771186619895e-06,
+      "loss": 0.874,
+      "step": 25068
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.47301672102645e-06,
+      "loss": 1.0527,
+      "step": 25069
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.472256356311637e-06,
+      "loss": 0.694,
+      "step": 25070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.471496024523078e-06,
+      "loss": 0.9243,
+      "step": 25071
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.470735725666303e-06,
+      "loss": 0.8321,
+      "step": 25072
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.46997545974684e-06,
+      "loss": 1.0464,
+      "step": 25073
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.469215226770219e-06,
+      "loss": 0.7237,
+      "step": 25074
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.468455026741968e-06,
+      "loss": 0.7979,
+      "step": 25075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.467694859667611e-06,
+      "loss": 0.8339,
+      "step": 25076
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.466934725552684e-06,
+      "loss": 0.8663,
+      "step": 25077
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.466174624402712e-06,
+      "loss": 0.8702,
+      "step": 25078
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.465414556223221e-06,
+      "loss": 0.918,
+      "step": 25079
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4646545210197435e-06,
+      "loss": 0.9437,
+      "step": 25080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463894518797798e-06,
+      "loss": 0.8891,
+      "step": 25081
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463134549562909e-06,
+      "loss": 0.578,
+      "step": 25082
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.462374613320616e-06,
+      "loss": 0.6864,
+      "step": 25083
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4616147100764385e-06,
+      "loss": 0.7434,
+      "step": 25084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.460854839835902e-06,
+      "loss": 0.559,
+      "step": 25085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.460095002604533e-06,
+      "loss": 0.797,
+      "step": 25086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4593351983878585e-06,
+      "loss": 0.9397,
+      "step": 25087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.458575427191402e-06,
+      "loss": 0.6133,
+      "step": 25088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.457815689020689e-06,
+      "loss": 0.9801,
+      "step": 25089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4570559838812445e-06,
+      "loss": 0.9885,
+      "step": 25090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456296311778593e-06,
+      "loss": 0.59,
+      "step": 25091
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45553667271826e-06,
+      "loss": 0.714,
+      "step": 25092
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45477706670577e-06,
+      "loss": 0.7336,
+      "step": 25093
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45401749374664e-06,
+      "loss": 0.9454,
+      "step": 25094
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.453257953846405e-06,
+      "loss": 0.7667,
+      "step": 25095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.452498447010581e-06,
+      "loss": 0.9697,
+      "step": 25096
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.451738973244696e-06,
+      "loss": 0.4237,
+      "step": 25097
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.450979532554268e-06,
+      "loss": 0.8796,
+      "step": 25098
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.450220124944827e-06,
+      "loss": 0.4901,
+      "step": 25099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.449460750421884e-06,
+      "loss": 0.5276,
+      "step": 25100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.448701408990969e-06,
+      "loss": 1.0239,
+      "step": 25101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.447942100657604e-06,
+      "loss": 1.0779,
+      "step": 25102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.44718282542731e-06,
+      "loss": 1.0234,
+      "step": 25103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4464235833056075e-06,
+      "loss": 0.6927,
+      "step": 25104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.445664374298018e-06,
+      "loss": 0.935,
+      "step": 25105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4449051984100645e-06,
+      "loss": 0.8697,
+      "step": 25106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.444146055647261e-06,
+      "loss": 0.9818,
+      "step": 25107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.443386946015137e-06,
+      "loss": 0.9559,
+      "step": 25108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.442627869519214e-06,
+      "loss": 0.6653,
+      "step": 25109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.441868826165002e-06,
+      "loss": 0.65,
+      "step": 25110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.441109815958028e-06,
+      "loss": 0.9461,
+      "step": 25111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.440350838903804e-06,
+      "loss": 0.9027,
+      "step": 25112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.43959189500786e-06,
+      "loss": 0.9274,
+      "step": 25113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.438832984275709e-06,
+      "loss": 0.942,
+      "step": 25114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.438074106712872e-06,
+      "loss": 1.0532,
+      "step": 25115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.437315262324867e-06,
+      "loss": 0.9122,
+      "step": 25116
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.436556451117211e-06,
+      "loss": 0.7749,
+      "step": 25117
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.435797673095424e-06,
+      "loss": 1.1665,
+      "step": 25118
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.435038928265023e-06,
+      "loss": 1.1944,
+      "step": 25119
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.434280216631526e-06,
+      "loss": 0.87,
+      "step": 25120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.43352153820045e-06,
+      "loss": 0.8991,
+      "step": 25121
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.432762892977312e-06,
+      "loss": 0.9433,
+      "step": 25122
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.43200428096763e-06,
+      "loss": 0.7414,
+      "step": 25123
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.43124570217692e-06,
+      "loss": 0.9248,
+      "step": 25124
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.430487156610695e-06,
+      "loss": 0.8995,
+      "step": 25125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.429728644274479e-06,
+      "loss": 0.9759,
+      "step": 25126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.428970165173783e-06,
+      "loss": 0.7889,
+      "step": 25127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.428211719314124e-06,
+      "loss": 1.0071,
+      "step": 25128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4274533067010205e-06,
+      "loss": 0.9357,
+      "step": 25129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.426694927339977e-06,
+      "loss": 0.5402,
+      "step": 25130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.425936581236519e-06,
+      "loss": 0.94,
+      "step": 25131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4251782683961595e-06,
+      "loss": 0.9782,
+      "step": 25132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.424419988824411e-06,
+      "loss": 1.0112,
+      "step": 25133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.423661742526788e-06,
+      "loss": 0.8203,
+      "step": 25134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.422903529508805e-06,
+      "loss": 0.6256,
+      "step": 25135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.422145349775976e-06,
+      "loss": 0.7279,
+      "step": 25136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.421387203333811e-06,
+      "loss": 0.8294,
+      "step": 25137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.42062909018783e-06,
+      "loss": 0.6322,
+      "step": 25138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.419871010343547e-06,
+      "loss": 0.8931,
+      "step": 25139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.419112963806468e-06,
+      "loss": 0.9182,
+      "step": 25140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4183549505821055e-06,
+      "loss": 0.7344,
+      "step": 25141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417596970675976e-06,
+      "loss": 1.0358,
+      "step": 25142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.416839024093587e-06,
+      "loss": 0.5396,
+      "step": 25143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4160811108404566e-06,
+      "loss": 0.8346,
+      "step": 25144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.415323230922094e-06,
+      "loss": 0.9819,
+      "step": 25145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4145653843440085e-06,
+      "loss": 1.0377,
+      "step": 25146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.413807571111714e-06,
+      "loss": 0.9326,
+      "step": 25147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.41304979123072e-06,
+      "loss": 0.6136,
+      "step": 25148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4122920447065375e-06,
+      "loss": 0.7398,
+      "step": 25149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411534331544676e-06,
+      "loss": 0.8052,
+      "step": 25150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4107766517506466e-06,
+      "loss": 0.7221,
+      "step": 25151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.410019005329958e-06,
+      "loss": 0.7092,
+      "step": 25152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.409261392288123e-06,
+      "loss": 0.9367,
+      "step": 25153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.408503812630648e-06,
+      "loss": 1.068,
+      "step": 25154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407746266363039e-06,
+      "loss": 0.8324,
+      "step": 25155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4069887534908135e-06,
+      "loss": 0.926,
+      "step": 25156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4062312740194756e-06,
+      "loss": 0.8159,
+      "step": 25157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.405473827954533e-06,
+      "loss": 0.8816,
+      "step": 25158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4047164153015006e-06,
+      "loss": 0.6749,
+      "step": 25159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.403959036065876e-06,
+      "loss": 0.9655,
+      "step": 25160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.403201690253168e-06,
+      "loss": 0.8239,
+      "step": 25161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.40244437786889e-06,
+      "loss": 1.1707,
+      "step": 25162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.401687098918549e-06,
+      "loss": 1.0244,
+      "step": 25163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.400929853407651e-06,
+      "loss": 0.8516,
+      "step": 25164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.400172641341699e-06,
+      "loss": 0.8819,
+      "step": 25165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.399415462726203e-06,
+      "loss": 0.8308,
+      "step": 25166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.398658317566669e-06,
+      "loss": 0.7474,
+      "step": 25167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3979012058686006e-06,
+      "loss": 0.8883,
+      "step": 25168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.397144127637512e-06,
+      "loss": 0.7088,
+      "step": 25169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3963870828788975e-06,
+      "loss": 0.7572,
+      "step": 25170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.395630071598269e-06,
+      "loss": 0.7947,
+      "step": 25171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.394873093801129e-06,
+      "loss": 0.7484,
+      "step": 25172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3941161494929806e-06,
+      "loss": 0.8621,
+      "step": 25173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.393359238679334e-06,
+      "loss": 0.5211,
+      "step": 25174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.39260236136569e-06,
+      "loss": 0.8092,
+      "step": 25175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391845517557553e-06,
+      "loss": 0.9602,
+      "step": 25176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391088707260428e-06,
+      "loss": 0.5898,
+      "step": 25177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390331930479817e-06,
+      "loss": 0.906,
+      "step": 25178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3895751872212254e-06,
+      "loss": 0.5572,
+      "step": 25179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.388818477490154e-06,
+      "loss": 0.9124,
+      "step": 25180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.388061801292109e-06,
+      "loss": 0.7558,
+      "step": 25181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3873051586325874e-06,
+      "loss": 0.7136,
+      "step": 25182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3865485495170965e-06,
+      "loss": 0.7938,
+      "step": 25183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3857919739511364e-06,
+      "loss": 0.6744,
+      "step": 25184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3850354319402095e-06,
+      "loss": 0.668,
+      "step": 25185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.384278923489814e-06,
+      "loss": 0.7387,
+      "step": 25186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.383522448605459e-06,
+      "loss": 0.7417,
+      "step": 25187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.38276600729264e-06,
+      "loss": 0.8608,
+      "step": 25188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.382009599556864e-06,
+      "loss": 1.001,
+      "step": 25189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.381253225403624e-06,
+      "loss": 0.601,
+      "step": 25190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.380496884838422e-06,
+      "loss": 0.7902,
+      "step": 25191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.379740577866757e-06,
+      "loss": 0.8256,
+      "step": 25192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.378984304494135e-06,
+      "loss": 1.1441,
+      "step": 25193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.378228064726051e-06,
+      "loss": 0.951,
+      "step": 25194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.377471858568007e-06,
+      "loss": 0.9567,
+      "step": 25195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.376715686025501e-06,
+      "loss": 0.4754,
+      "step": 25196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3759595471040315e-06,
+      "loss": 0.9192,
+      "step": 25197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3752034418090985e-06,
+      "loss": 0.7316,
+      "step": 25198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.374447370146198e-06,
+      "loss": 1.0587,
+      "step": 25199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.373691332120832e-06,
+      "loss": 0.7786,
+      "step": 25200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.372935327738495e-06,
+      "loss": 0.5599,
+      "step": 25201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.372179357004685e-06,
+      "loss": 1.2043,
+      "step": 25202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.371423419924903e-06,
+      "loss": 0.7287,
+      "step": 25203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.37066751650464e-06,
+      "loss": 0.8441,
+      "step": 25204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3699116467494e-06,
+      "loss": 0.5094,
+      "step": 25205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369155810664675e-06,
+      "loss": 0.8957,
+      "step": 25206
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.368400008255965e-06,
+      "loss": 0.8484,
+      "step": 25207
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.367644239528764e-06,
+      "loss": 0.8437,
+      "step": 25208
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.366888504488572e-06,
+      "loss": 0.6667,
+      "step": 25209
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.366132803140874e-06,
+      "loss": 1.0444,
+      "step": 25210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365377135491177e-06,
+      "loss": 0.6226,
+      "step": 25211
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.364621501544972e-06,
+      "loss": 0.9279,
+      "step": 25212
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3638659013077545e-06,
+      "loss": 1.1359,
+      "step": 25213
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.363110334785017e-06,
+      "loss": 0.9895,
+      "step": 25214
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.362354801982259e-06,
+      "loss": 0.5473,
+      "step": 25215
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.36159930290497e-06,
+      "loss": 0.6788,
+      "step": 25216
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.360843837558643e-06,
+      "loss": 0.7732,
+      "step": 25217
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.360088405948778e-06,
+      "loss": 1.1526,
+      "step": 25218
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.35933300808087e-06,
+      "loss": 0.9159,
+      "step": 25219
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.358577643960403e-06,
+      "loss": 0.6607,
+      "step": 25220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357822313592875e-06,
+      "loss": 0.8847,
+      "step": 25221
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357067016983774e-06,
+      "loss": 0.8147,
+      "step": 25222
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3563117541386025e-06,
+      "loss": 0.5983,
+      "step": 25223
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355556525062847e-06,
+      "loss": 1.1105,
+      "step": 25224
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.354801329762e-06,
+      "loss": 1.0254,
+      "step": 25225
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.354046168241553e-06,
+      "loss": 0.8849,
+      "step": 25226
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.353291040507001e-06,
+      "loss": 0.8153,
+      "step": 25227
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.35253594656383e-06,
+      "loss": 0.9755,
+      "step": 25228
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3517808864175335e-06,
+      "loss": 0.7877,
+      "step": 25229
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.351025860073604e-06,
+      "loss": 0.9669,
+      "step": 25230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.35027086753753e-06,
+      "loss": 0.941,
+      "step": 25231
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3495159088148015e-06,
+      "loss": 0.6849,
+      "step": 25232
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.348760983910912e-06,
+      "loss": 0.8141,
+      "step": 25233
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.348006092831348e-06,
+      "loss": 0.9783,
+      "step": 25234
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.347251235581596e-06,
+      "loss": 1.0178,
+      "step": 25235
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.346496412167154e-06,
+      "loss": 0.7227,
+      "step": 25236
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.345741622593507e-06,
+      "loss": 0.8278,
+      "step": 25237
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.344986866866143e-06,
+      "loss": 1.0603,
+      "step": 25238
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.344232144990555e-06,
+      "loss": 0.8106,
+      "step": 25239
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.343477456972221e-06,
+      "loss": 0.8793,
+      "step": 25240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.34272280281664e-06,
+      "loss": 0.9315,
+      "step": 25241
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.341968182529297e-06,
+      "loss": 0.9395,
+      "step": 25242
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.341213596115677e-06,
+      "loss": 0.7095,
+      "step": 25243
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.34045904358127e-06,
+      "loss": 0.8837,
+      "step": 25244
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3397045249315615e-06,
+      "loss": 0.9279,
+      "step": 25245
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.338950040172039e-06,
+      "loss": 0.9117,
+      "step": 25246
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.338195589308186e-06,
+      "loss": 0.6015,
+      "step": 25247
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.337441172345497e-06,
+      "loss": 0.5801,
+      "step": 25248
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.336686789289456e-06,
+      "loss": 1.0551,
+      "step": 25249
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.335932440145544e-06,
+      "loss": 1.0277,
+      "step": 25250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.335178124919249e-06,
+      "loss": 0.823,
+      "step": 25251
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.334423843616056e-06,
+      "loss": 0.8874,
+      "step": 25252
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.333669596241446e-06,
+      "loss": 0.8114,
+      "step": 25253
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.332915382800915e-06,
+      "loss": 0.719,
+      "step": 25254
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.332161203299941e-06,
+      "loss": 0.9601,
+      "step": 25255
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.331407057744009e-06,
+      "loss": 0.7369,
+      "step": 25256
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.330652946138602e-06,
+      "loss": 0.8161,
+      "step": 25257
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.329898868489207e-06,
+      "loss": 0.9865,
+      "step": 25258
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3291448248013045e-06,
+      "loss": 0.9278,
+      "step": 25259
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.328390815080381e-06,
+      "loss": 0.6677,
+      "step": 25260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3276368393319174e-06,
+      "loss": 0.6804,
+      "step": 25261
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326882897561398e-06,
+      "loss": 0.6719,
+      "step": 25262
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326128989774305e-06,
+      "loss": 0.7672,
+      "step": 25263
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.325375115976121e-06,
+      "loss": 0.6478,
+      "step": 25264
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.324621276172324e-06,
+      "loss": 0.945,
+      "step": 25265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.323867470368404e-06,
+      "loss": 0.7797,
+      "step": 25266
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.323113698569839e-06,
+      "loss": 0.8166,
+      "step": 25267
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.322359960782112e-06,
+      "loss": 0.8904,
+      "step": 25268
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.321606257010705e-06,
+      "loss": 0.6726,
+      "step": 25269
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.320852587261094e-06,
+      "loss": 0.5574,
+      "step": 25270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.320098951538759e-06,
+      "loss": 0.8604,
+      "step": 25271
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319345349849188e-06,
+      "loss": 1.0049,
+      "step": 25272
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.318591782197856e-06,
+      "loss": 0.917,
+      "step": 25273
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3178382485902456e-06,
+      "loss": 0.8583,
+      "step": 25274
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.317084749031835e-06,
+      "loss": 1.055,
+      "step": 25275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.316331283528104e-06,
+      "loss": 0.8335,
+      "step": 25276
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.315577852084532e-06,
+      "loss": 1.2247,
+      "step": 25277
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.314824454706594e-06,
+      "loss": 0.9008,
+      "step": 25278
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.314071091399782e-06,
+      "loss": 0.9429,
+      "step": 25279
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3133177621695585e-06,
+      "loss": 0.7056,
+      "step": 25280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.312564467021412e-06,
+      "loss": 0.7636,
+      "step": 25281
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.311811205960815e-06,
+      "loss": 1.0113,
+      "step": 25282
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.311057978993245e-06,
+      "loss": 0.6838,
+      "step": 25283
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.310304786124185e-06,
+      "loss": 0.8571,
+      "step": 25284
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30955162735911e-06,
+      "loss": 0.977,
+      "step": 25285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.308798502703496e-06,
+      "loss": 0.9465,
+      "step": 25286
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.308045412162821e-06,
+      "loss": 0.9527,
+      "step": 25287
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.307292355742559e-06,
+      "loss": 0.6999,
+      "step": 25288
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.306539333448191e-06,
+      "loss": 1.001,
+      "step": 25289
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3057863452851875e-06,
+      "loss": 0.8139,
+      "step": 25290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.305033391259028e-06,
+      "loss": 0.7933,
+      "step": 25291
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3042804713751875e-06,
+      "loss": 0.9774,
+      "step": 25292
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30352758563914e-06,
+      "loss": 0.8398,
+      "step": 25293
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.302774734056362e-06,
+      "loss": 0.9434,
+      "step": 25294
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3020219166323276e-06,
+      "loss": 0.9779,
+      "step": 25295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3012691333725066e-06,
+      "loss": 0.6049,
+      "step": 25296
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3005163842823835e-06,
+      "loss": 1.1063,
+      "step": 25297
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.299763669367426e-06,
+      "loss": 0.8318,
+      "step": 25298
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.299010988633114e-06,
+      "loss": 1.0237,
+      "step": 25299
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.298258342084912e-06,
+      "loss": 0.899,
+      "step": 25300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.297505729728294e-06,
+      "loss": 0.8831,
+      "step": 25301
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.29675315156874e-06,
+      "loss": 1.0988,
+      "step": 25302
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.296000607611719e-06,
+      "loss": 0.9123,
+      "step": 25303
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.295248097862704e-06,
+      "loss": 0.7775,
+      "step": 25304
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.294495622327167e-06,
+      "loss": 1.0106,
+      "step": 25305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.293743181010581e-06,
+      "loss": 0.8919,
+      "step": 25306
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.292990773918417e-06,
+      "loss": 0.9758,
+      "step": 25307
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.292238401056144e-06,
+      "loss": 0.7979,
+      "step": 25308
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.291486062429242e-06,
+      "loss": 1.1153,
+      "step": 25309
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.290733758043174e-06,
+      "loss": 0.8473,
+      "step": 25310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.289981487903412e-06,
+      "loss": 0.999,
+      "step": 25311
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.289229252015428e-06,
+      "loss": 0.7658,
+      "step": 25312
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.288477050384692e-06,
+      "loss": 0.8914,
+      "step": 25313
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2877248830166696e-06,
+      "loss": 0.7727,
+      "step": 25314
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286972749916839e-06,
+      "loss": 0.8702,
+      "step": 25315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286220651090666e-06,
+      "loss": 1.015,
+      "step": 25316
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2854685865436205e-06,
+      "loss": 1.1396,
+      "step": 25317
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.284716556281171e-06,
+      "loss": 0.7933,
+      "step": 25318
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2839645603087875e-06,
+      "loss": 0.7195,
+      "step": 25319
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.283212598631935e-06,
+      "loss": 0.6584,
+      "step": 25320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.282460671256085e-06,
+      "loss": 0.9054,
+      "step": 25321
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2817087781867064e-06,
+      "loss": 1.1669,
+      "step": 25322
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.280956919429263e-06,
+      "loss": 0.7823,
+      "step": 25323
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.280205094989227e-06,
+      "loss": 0.8435,
+      "step": 25324
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.279453304872063e-06,
+      "loss": 0.8193,
+      "step": 25325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.278701549083234e-06,
+      "loss": 0.9931,
+      "step": 25326
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277949827628217e-06,
+      "loss": 0.9453,
+      "step": 25327
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2771981405124715e-06,
+      "loss": 0.5467,
+      "step": 25328
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.276446487741471e-06,
+      "loss": 0.8777,
+      "step": 25329
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.275694869320672e-06,
+      "loss": 1.0193,
+      "step": 25330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.274943285255545e-06,
+      "loss": 0.6848,
+      "step": 25331
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.27419173555155e-06,
+      "loss": 1.0465,
+      "step": 25332
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.273440220214162e-06,
+      "loss": 0.8834,
+      "step": 25333
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.272688739248842e-06,
+      "loss": 0.6888,
+      "step": 25334
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.271937292661054e-06,
+      "loss": 0.7467,
+      "step": 25335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.271185880456262e-06,
+      "loss": 0.8876,
+      "step": 25336
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.270434502639933e-06,
+      "loss": 0.7359,
+      "step": 25337
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.269683159217529e-06,
+      "loss": 1.1032,
+      "step": 25338
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268931850194514e-06,
+      "loss": 1.1192,
+      "step": 25339
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268180575576353e-06,
+      "loss": 0.7735,
+      "step": 25340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.267429335368507e-06,
+      "loss": 0.8724,
+      "step": 25341
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.266678129576441e-06,
+      "loss": 0.741,
+      "step": 25342
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.265926958205617e-06,
+      "loss": 1.1311,
+      "step": 25343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.265175821261495e-06,
+      "loss": 1.028,
+      "step": 25344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.264424718749544e-06,
+      "loss": 0.9927,
+      "step": 25345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.263673650675221e-06,
+      "loss": 0.8582,
+      "step": 25346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.262922617043988e-06,
+      "loss": 0.9939,
+      "step": 25347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2621716178613114e-06,
+      "loss": 0.9056,
+      "step": 25348
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.261420653132649e-06,
+      "loss": 0.826,
+      "step": 25349
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.260669722863457e-06,
+      "loss": 0.6882,
+      "step": 25350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.259918827059205e-06,
+      "loss": 0.9088,
+      "step": 25351
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.259167965725348e-06,
+      "loss": 0.5949,
+      "step": 25352
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.258417138867349e-06,
+      "loss": 0.9868,
+      "step": 25353
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.257666346490666e-06,
+      "loss": 1.0171,
+      "step": 25354
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.256915588600761e-06,
+      "loss": 0.8309,
+      "step": 25355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.256164865203093e-06,
+      "loss": 0.7131,
+      "step": 25356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.255414176303115e-06,
+      "loss": 0.7846,
+      "step": 25357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.254663521906298e-06,
+      "loss": 1.2636,
+      "step": 25358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.253912902018097e-06,
+      "loss": 1.0801,
+      "step": 25359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.253162316643965e-06,
+      "loss": 1.0852,
+      "step": 25360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2524117657893645e-06,
+      "loss": 0.8298,
+      "step": 25361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2516612494597496e-06,
+      "loss": 0.7782,
+      "step": 25362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.250910767660584e-06,
+      "loss": 0.8707,
+      "step": 25363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.250160320397321e-06,
+      "loss": 0.997,
+      "step": 25364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.249409907675422e-06,
+      "loss": 0.7466,
+      "step": 25365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.248659529500342e-06,
+      "loss": 0.8573,
+      "step": 25366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247909185877538e-06,
+      "loss": 0.7904,
+      "step": 25367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247158876812464e-06,
+      "loss": 0.9005,
+      "step": 25368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.24640860231058e-06,
+      "loss": 0.697,
+      "step": 25369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.24565836237734e-06,
+      "loss": 0.9747,
+      "step": 25370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2449081570181996e-06,
+      "loss": 0.9986,
+      "step": 25371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.244157986238615e-06,
+      "loss": 0.9028,
+      "step": 25372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.243407850044042e-06,
+      "loss": 1.0256,
+      "step": 25373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2426577484399365e-06,
+      "loss": 0.8388,
+      "step": 25374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.241907681431747e-06,
+      "loss": 0.668,
+      "step": 25375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.241157649024937e-06,
+      "loss": 0.844,
+      "step": 25376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.240407651224958e-06,
+      "loss": 0.8148,
+      "step": 25377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.239657688037263e-06,
+      "loss": 0.6218,
+      "step": 25378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.238907759467309e-06,
+      "loss": 0.9695,
+      "step": 25379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.238157865520539e-06,
+      "loss": 0.7448,
+      "step": 25380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237408006202418e-06,
+      "loss": 0.6854,
+      "step": 25381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.236658181518395e-06,
+      "loss": 0.8361,
+      "step": 25382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.235908391473923e-06,
+      "loss": 1.0306,
+      "step": 25383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.235158636074453e-06,
+      "loss": 0.7787,
+      "step": 25384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.23440891532544e-06,
+      "loss": 0.6056,
+      "step": 25385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.233659229232333e-06,
+      "loss": 0.688,
+      "step": 25386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.232909577800582e-06,
+      "loss": 0.9589,
+      "step": 25387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.232159961035645e-06,
+      "loss": 0.669,
+      "step": 25388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231410378942976e-06,
+      "loss": 0.8232,
+      "step": 25389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2306608315280146e-06,
+      "loss": 1.0285,
+      "step": 25390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.229911318796219e-06,
+      "loss": 0.6822,
+      "step": 25391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.229161840753036e-06,
+      "loss": 0.5337,
+      "step": 25392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.228412397403916e-06,
+      "loss": 0.9244,
+      "step": 25393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.227662988754314e-06,
+      "loss": 0.6768,
+      "step": 25394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226913614809677e-06,
+      "loss": 0.876,
+      "step": 25395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226164275575454e-06,
+      "loss": 0.7173,
+      "step": 25396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.225414971057094e-06,
+      "loss": 1.0228,
+      "step": 25397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.224665701260045e-06,
+      "loss": 0.7413,
+      "step": 25398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.22391646618976e-06,
+      "loss": 0.7531,
+      "step": 25399
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.223167265851684e-06,
+      "loss": 0.7195,
+      "step": 25400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.222418100251267e-06,
+      "loss": 1.0006,
+      "step": 25401
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.221668969393955e-06,
+      "loss": 0.8462,
+      "step": 25402
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2209198732851975e-06,
+      "loss": 0.6738,
+      "step": 25403
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.220170811930441e-06,
+      "loss": 0.757,
+      "step": 25404
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.219421785335128e-06,
+      "loss": 0.8136,
+      "step": 25405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.218672793504715e-06,
+      "loss": 1.1008,
+      "step": 25406
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2179238364446475e-06,
+      "loss": 0.6951,
+      "step": 25407
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.217174914160365e-06,
+      "loss": 0.8247,
+      "step": 25408
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.216426026657324e-06,
+      "loss": 0.7289,
+      "step": 25409
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.215677173940959e-06,
+      "loss": 1.005,
+      "step": 25410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214928356016718e-06,
+      "loss": 0.7611,
+      "step": 25411
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214179572890052e-06,
+      "loss": 0.6695,
+      "step": 25412
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.213430824566403e-06,
+      "loss": 0.8718,
+      "step": 25413
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.212682111051219e-06,
+      "loss": 0.8689,
+      "step": 25414
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.211933432349941e-06,
+      "loss": 1.0409,
+      "step": 25415
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.211184788468016e-06,
+      "loss": 0.7627,
+      "step": 25416
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.210436179410885e-06,
+      "loss": 1.1096,
+      "step": 25417
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.209687605183991e-06,
+      "loss": 0.9341,
+      "step": 25418
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208939065792789e-06,
+      "loss": 0.9884,
+      "step": 25419
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208190561242708e-06,
+      "loss": 0.7384,
+      "step": 25420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2074420915392e-06,
+      "loss": 0.8104,
+      "step": 25421
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.206693656687705e-06,
+      "loss": 1.0798,
+      "step": 25422
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.205945256693661e-06,
+      "loss": 0.657,
+      "step": 25423
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.205196891562519e-06,
+      "loss": 0.7876,
+      "step": 25424
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.204448561299718e-06,
+      "loss": 0.8317,
+      "step": 25425
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2037002659107004e-06,
+      "loss": 0.7304,
+      "step": 25426
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202952005400908e-06,
+      "loss": 0.5935,
+      "step": 25427
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202203779775778e-06,
+      "loss": 0.9584,
+      "step": 25428
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.201455589040758e-06,
+      "loss": 0.6659,
+      "step": 25429
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.200707433201284e-06,
+      "loss": 1.0768,
+      "step": 25430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1999593122628e-06,
+      "loss": 0.6935,
+      "step": 25431
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1992112262307445e-06,
+      "loss": 0.9498,
+      "step": 25432
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1984631751105575e-06,
+      "loss": 0.6572,
+      "step": 25433
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.197715158907679e-06,
+      "loss": 0.8877,
+      "step": 25434
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.196967177627551e-06,
+      "loss": 0.7342,
+      "step": 25435
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.196219231275605e-06,
+      "loss": 0.7218,
+      "step": 25436
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1954713198572925e-06,
+      "loss": 0.8516,
+      "step": 25437
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.194723443378044e-06,
+      "loss": 0.9171,
+      "step": 25438
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193975601843305e-06,
+      "loss": 1.0839,
+      "step": 25439
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193227795258505e-06,
+      "loss": 0.8657,
+      "step": 25440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.192480023629083e-06,
+      "loss": 0.822,
+      "step": 25441
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.191732286960482e-06,
+      "loss": 1.1219,
+      "step": 25442
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.190984585258141e-06,
+      "loss": 0.9467,
+      "step": 25443
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.190236918527492e-06,
+      "loss": 0.8268,
+      "step": 25444
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.189489286773974e-06,
+      "loss": 0.541,
+      "step": 25445
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.188741690003025e-06,
+      "loss": 0.8062,
+      "step": 25446
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1879941282200795e-06,
+      "loss": 0.8299,
+      "step": 25447
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.187246601430573e-06,
+      "loss": 0.9173,
+      "step": 25448
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.186499109639951e-06,
+      "loss": 0.7303,
+      "step": 25449
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.185751652853636e-06,
+      "loss": 0.8418,
+      "step": 25450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1850042310770725e-06,
+      "loss": 0.85,
+      "step": 25451
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.184256844315692e-06,
+      "loss": 1.3051,
+      "step": 25452
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.183509492574929e-06,
+      "loss": 0.7354,
+      "step": 25453
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182762175860217e-06,
+      "loss": 0.9811,
+      "step": 25454
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182014894176999e-06,
+      "loss": 0.744,
+      "step": 25455
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181267647530701e-06,
+      "loss": 0.8456,
+      "step": 25456
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.180520435926761e-06,
+      "loss": 0.6221,
+      "step": 25457
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.17977325937061e-06,
+      "loss": 1.0622,
+      "step": 25458
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.179026117867686e-06,
+      "loss": 1.1199,
+      "step": 25459
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.178279011423417e-06,
+      "loss": 1.1023,
+      "step": 25460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.177531940043241e-06,
+      "loss": 0.8549,
+      "step": 25461
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176784903732585e-06,
+      "loss": 0.885,
+      "step": 25462
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176037902496888e-06,
+      "loss": 1.0736,
+      "step": 25463
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1752909363415774e-06,
+      "loss": 0.7993,
+      "step": 25464
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.174544005272086e-06,
+      "loss": 0.6849,
+      "step": 25465
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.173797109293844e-06,
+      "loss": 0.6614,
+      "step": 25466
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1730502484122895e-06,
+      "loss": 0.6888,
+      "step": 25467
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.172303422632849e-06,
+      "loss": 1.113,
+      "step": 25468
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.171556631960958e-06,
+      "loss": 1.0581,
+      "step": 25469
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.170809876402039e-06,
+      "loss": 0.9771,
+      "step": 25470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1700631559615265e-06,
+      "loss": 0.9716,
+      "step": 25471
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.169316470644849e-06,
+      "loss": 0.6559,
+      "step": 25472
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.168569820457441e-06,
+      "loss": 0.7424,
+      "step": 25473
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16782320540473e-06,
+      "loss": 0.7737,
+      "step": 25474
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.167076625492147e-06,
+      "loss": 0.8626,
+      "step": 25475
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.166330080725117e-06,
+      "loss": 0.8614,
+      "step": 25476
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1655835711090715e-06,
+      "loss": 0.873,
+      "step": 25477
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.164837096649441e-06,
+      "loss": 0.9626,
+      "step": 25478
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1640906573516505e-06,
+      "loss": 0.8099,
+      "step": 25479
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1633442532211295e-06,
+      "loss": 1.0177,
+      "step": 25480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.162597884263308e-06,
+      "loss": 0.5502,
+      "step": 25481
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.161851550483611e-06,
+      "loss": 0.8286,
+      "step": 25482
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.161105251887466e-06,
+      "loss": 0.6189,
+      "step": 25483
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.160358988480299e-06,
+      "loss": 0.6872,
+      "step": 25484
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159612760267541e-06,
+      "loss": 0.6837,
+      "step": 25485
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.158866567254617e-06,
+      "loss": 0.8357,
+      "step": 25486
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.158120409446952e-06,
+      "loss": 1.0481,
+      "step": 25487
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.157374286849974e-06,
+      "loss": 0.978,
+      "step": 25488
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156628199469112e-06,
+      "loss": 0.8987,
+      "step": 25489
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.15588214730978e-06,
+      "loss": 0.7605,
+      "step": 25490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.155136130377414e-06,
+      "loss": 1.1564,
+      "step": 25491
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.154390148677437e-06,
+      "loss": 0.9406,
+      "step": 25492
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.153644202215272e-06,
+      "loss": 0.6211,
+      "step": 25493
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.152898290996344e-06,
+      "loss": 0.6271,
+      "step": 25494
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1521524150260796e-06,
+      "loss": 0.8619,
+      "step": 25495
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1514065743099e-06,
+      "loss": 1.2951,
+      "step": 25496
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.150660768853227e-06,
+      "loss": 0.835,
+      "step": 25497
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149914998661491e-06,
+      "loss": 0.7069,
+      "step": 25498
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149169263740117e-06,
+      "loss": 0.8671,
+      "step": 25499
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.148423564094517e-06,
+      "loss": 0.7535,
+      "step": 25500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.147677899730119e-06,
+      "loss": 0.6778,
+      "step": 25501
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.146932270652345e-06,
+      "loss": 0.8412,
+      "step": 25502
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.14618667686662e-06,
+      "loss": 0.6384,
+      "step": 25503
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.145441118378365e-06,
+      "loss": 0.9581,
+      "step": 25504
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.144695595193001e-06,
+      "loss": 0.9155,
+      "step": 25505
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1439501073159515e-06,
+      "loss": 1.1127,
+      "step": 25506
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.143204654752634e-06,
+      "loss": 0.8361,
+      "step": 25507
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.142459237508473e-06,
+      "loss": 0.993,
+      "step": 25508
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.141713855588888e-06,
+      "loss": 0.7612,
+      "step": 25509
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.140968508999297e-06,
+      "loss": 0.8369,
+      "step": 25510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.140223197745124e-06,
+      "loss": 0.767,
+      "step": 25511
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.139477921831788e-06,
+      "loss": 0.6541,
+      "step": 25512
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.138732681264709e-06,
+      "loss": 0.8356,
+      "step": 25513
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137987476049305e-06,
+      "loss": 1.2788,
+      "step": 25514
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137242306190991e-06,
+      "loss": 0.9115,
+      "step": 25515
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.136497171695196e-06,
+      "loss": 0.8051,
+      "step": 25516
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.135752072567334e-06,
+      "loss": 0.9752,
+      "step": 25517
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.135007008812822e-06,
+      "loss": 0.9768,
+      "step": 25518
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.134261980437084e-06,
+      "loss": 1.06,
+      "step": 25519
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.133516987445526e-06,
+      "loss": 0.9325,
+      "step": 25520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132772029843576e-06,
+      "loss": 0.875,
+      "step": 25521
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132027107636648e-06,
+      "loss": 0.991,
+      "step": 25522
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.131282220830159e-06,
+      "loss": 0.995,
+      "step": 25523
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.130537369429527e-06,
+      "loss": 0.6245,
+      "step": 25524
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.129792553440168e-06,
+      "loss": 0.7443,
+      "step": 25525
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.129047772867499e-06,
+      "loss": 1.0514,
+      "step": 25526
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12830302771693e-06,
+      "loss": 0.9844,
+      "step": 25527
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.127558317993888e-06,
+      "loss": 0.6683,
+      "step": 25528
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1268136437037855e-06,
+      "loss": 0.8794,
+      "step": 25529
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.126069004852033e-06,
+      "loss": 0.7838,
+      "step": 25530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.125324401444046e-06,
+      "loss": 0.7428,
+      "step": 25531
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.124579833485242e-06,
+      "loss": 0.8554,
+      "step": 25532
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1238353009810316e-06,
+      "loss": 0.9077,
+      "step": 25533
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.123090803936836e-06,
+      "loss": 0.9967,
+      "step": 25534
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.122346342358065e-06,
+      "loss": 0.671,
+      "step": 25535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.121601916250134e-06,
+      "loss": 0.8811,
+      "step": 25536
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.120857525618455e-06,
+      "loss": 0.7554,
+      "step": 25537
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.120113170468443e-06,
+      "loss": 1.0544,
+      "step": 25538
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.119368850805509e-06,
+      "loss": 0.9244,
+      "step": 25539
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.118624566635066e-06,
+      "loss": 0.8239,
+      "step": 25540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117880317962528e-06,
+      "loss": 0.8608,
+      "step": 25541
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117136104793307e-06,
+      "loss": 0.7921,
+      "step": 25542
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.116391927132813e-06,
+      "loss": 0.8314,
+      "step": 25543
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.115647784986459e-06,
+      "loss": 0.7369,
+      "step": 25544
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114903678359655e-06,
+      "loss": 0.7517,
+      "step": 25545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114159607257817e-06,
+      "loss": 0.6471,
+      "step": 25546
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.113415571686353e-06,
+      "loss": 0.9658,
+      "step": 25547
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.112671571650673e-06,
+      "loss": 0.8236,
+      "step": 25548
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.111927607156192e-06,
+      "loss": 0.8191,
+      "step": 25549
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.111183678208312e-06,
+      "loss": 0.7107,
+      "step": 25550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.110439784812444e-06,
+      "loss": 1.0006,
+      "step": 25551
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.109695926974005e-06,
+      "loss": 0.8502,
+      "step": 25552
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1089521046983994e-06,
+      "loss": 0.8974,
+      "step": 25553
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.108208317991037e-06,
+      "loss": 0.6998,
+      "step": 25554
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.107464566857327e-06,
+      "loss": 0.6947,
+      "step": 25555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.10672085130268e-06,
+      "loss": 0.8466,
+      "step": 25556
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.105977171332501e-06,
+      "loss": 0.8558,
+      "step": 25557
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.105233526952196e-06,
+      "loss": 0.8714,
+      "step": 25558
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.104489918167184e-06,
+      "loss": 0.8022,
+      "step": 25559
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.103746344982859e-06,
+      "loss": 0.8914,
+      "step": 25560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.103002807404637e-06,
+      "loss": 0.5887,
+      "step": 25561
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.102259305437921e-06,
+      "loss": 1.0595,
+      "step": 25562
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.101515839088116e-06,
+      "loss": 1.0463,
+      "step": 25563
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.100772408360635e-06,
+      "loss": 0.6825,
+      "step": 25564
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.100029013260881e-06,
+      "loss": 0.9228,
+      "step": 25565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.099285653794261e-06,
+      "loss": 0.8317,
+      "step": 25566
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.098542329966179e-06,
+      "loss": 0.8455,
+      "step": 25567
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0977990417820434e-06,
+      "loss": 1.0921,
+      "step": 25568
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.097055789247255e-06,
+      "loss": 1.0851,
+      "step": 25569
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.096312572367224e-06,
+      "loss": 0.9998,
+      "step": 25570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.09556939114735e-06,
+      "loss": 0.6098,
+      "step": 25571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.094826245593041e-06,
+      "loss": 0.7298,
+      "step": 25572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0940831357097e-06,
+      "loss": 0.7741,
+      "step": 25573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.093340061502731e-06,
+      "loss": 0.8958,
+      "step": 25574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.092597022977539e-06,
+      "loss": 0.8687,
+      "step": 25575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0918540201395215e-06,
+      "loss": 0.911,
+      "step": 25576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.091111052994091e-06,
+      "loss": 1.0882,
+      "step": 25577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.090368121546647e-06,
+      "loss": 0.971,
+      "step": 25578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.089625225802593e-06,
+      "loss": 0.911,
+      "step": 25579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.088882365767327e-06,
+      "loss": 0.7235,
+      "step": 25580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.088139541446251e-06,
+      "loss": 0.8088,
+      "step": 25581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.087396752844771e-06,
+      "loss": 1.1715,
+      "step": 25582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0866539999682896e-06,
+      "loss": 0.9115,
+      "step": 25583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.085911282822205e-06,
+      "loss": 0.7623,
+      "step": 25584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.085168601411918e-06,
+      "loss": 1.1443,
+      "step": 25585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.084425955742833e-06,
+      "loss": 0.8124,
+      "step": 25586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.083683345820347e-06,
+      "loss": 0.9553,
+      "step": 25587
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.082940771649858e-06,
+      "loss": 0.7529,
+      "step": 25588
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.082198233236778e-06,
+      "loss": 0.6548,
+      "step": 25589
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.081455730586495e-06,
+      "loss": 0.5256,
+      "step": 25590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.080713263704411e-06,
+      "loss": 1.2536,
+      "step": 25591
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.079970832595926e-06,
+      "loss": 0.8431,
+      "step": 25592
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.079228437266441e-06,
+      "loss": 0.7256,
+      "step": 25593
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0784860777213485e-06,
+      "loss": 0.9726,
+      "step": 25594
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077743753966056e-06,
+      "loss": 0.8945,
+      "step": 25595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077001466005957e-06,
+      "loss": 0.4924,
+      "step": 25596
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.07625921384645e-06,
+      "loss": 0.6089,
+      "step": 25597
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075516997492933e-06,
+      "loss": 0.6982,
+      "step": 25598
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074774816950804e-06,
+      "loss": 0.7853,
+      "step": 25599
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074032672225458e-06,
+      "loss": 0.9768,
+      "step": 25600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.073290563322295e-06,
+      "loss": 0.7898,
+      "step": 25601
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.072548490246709e-06,
+      "loss": 0.8541,
+      "step": 25602
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071806453004098e-06,
+      "loss": 0.9721,
+      "step": 25603
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071064451599857e-06,
+      "loss": 0.7694,
+      "step": 25604
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.070322486039383e-06,
+      "loss": 0.8107,
+      "step": 25605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0695805563280685e-06,
+      "loss": 0.7336,
+      "step": 25606
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.068838662471315e-06,
+      "loss": 0.9294,
+      "step": 25607
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0680968044745134e-06,
+      "loss": 1.0676,
+      "step": 25608
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.067354982343065e-06,
+      "loss": 0.5685,
+      "step": 25609
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0666131960823535e-06,
+      "loss": 1.0152,
+      "step": 25610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0658714456977795e-06,
+      "loss": 0.8214,
+      "step": 25611
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.065129731194732e-06,
+      "loss": 0.8267,
+      "step": 25612
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.064388052578614e-06,
+      "loss": 0.9789,
+      "step": 25613
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.063646409854812e-06,
+      "loss": 0.8065,
+      "step": 25614
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.062904803028723e-06,
+      "loss": 0.8512,
+      "step": 25615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.062163232105739e-06,
+      "loss": 0.9036,
+      "step": 25616
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0614216970912515e-06,
+      "loss": 0.8904,
+      "step": 25617
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.060680197990655e-06,
+      "loss": 0.9531,
+      "step": 25618
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.05993873480934e-06,
+      "loss": 0.7539,
+      "step": 25619
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.059197307552698e-06,
+      "loss": 0.9397,
+      "step": 25620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.058455916226123e-06,
+      "loss": 1.0323,
+      "step": 25621
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.057714560835004e-06,
+      "loss": 0.8492,
+      "step": 25622
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056973241384735e-06,
+      "loss": 0.8537,
+      "step": 25623
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0562319578807006e-06,
+      "loss": 0.9034,
+      "step": 25624
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.055490710328301e-06,
+      "loss": 0.7214,
+      "step": 25625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.054749498732921e-06,
+      "loss": 0.9249,
+      "step": 25626
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0540083230999525e-06,
+      "loss": 0.9804,
+      "step": 25627
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.053267183434786e-06,
+      "loss": 0.7998,
+      "step": 25628
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.052526079742812e-06,
+      "loss": 0.8286,
+      "step": 25629
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051785012029411e-06,
+      "loss": 0.5938,
+      "step": 25630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051043980299981e-06,
+      "loss": 0.8901,
+      "step": 25631
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.050302984559912e-06,
+      "loss": 0.979,
+      "step": 25632
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.049562024814586e-06,
+      "loss": 0.7152,
+      "step": 25633
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.048821101069397e-06,
+      "loss": 1.0499,
+      "step": 25634
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.048080213329729e-06,
+      "loss": 1.0643,
+      "step": 25635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.047339361600972e-06,
+      "loss": 0.8729,
+      "step": 25636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.046598545888511e-06,
+      "loss": 0.7883,
+      "step": 25637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045857766197737e-06,
+      "loss": 1.1677,
+      "step": 25638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045117022534041e-06,
+      "loss": 0.8292,
+      "step": 25639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.044376314902799e-06,
+      "loss": 0.7591,
+      "step": 25640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0436356433094034e-06,
+      "loss": 0.6675,
+      "step": 25641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.042895007759236e-06,
+      "loss": 0.766,
+      "step": 25642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0421544082576915e-06,
+      "loss": 0.4189,
+      "step": 25643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.041413844810148e-06,
+      "loss": 1.0391,
+      "step": 25644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.040673317421996e-06,
+      "loss": 0.8201,
+      "step": 25645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.039932826098617e-06,
+      "loss": 0.8503,
+      "step": 25646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.039192370845398e-06,
+      "loss": 0.9622,
+      "step": 25647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.038451951667724e-06,
+      "loss": 0.8977,
+      "step": 25648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.037711568570977e-06,
+      "loss": 0.7479,
+      "step": 25649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.036971221560543e-06,
+      "loss": 0.6915,
+      "step": 25650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.036230910641804e-06,
+      "loss": 0.899,
+      "step": 25651
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.035490635820146e-06,
+      "loss": 0.8519,
+      "step": 25652
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.034750397100953e-06,
+      "loss": 0.9163,
+      "step": 25653
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0340101944896044e-06,
+      "loss": 0.7644,
+      "step": 25654
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.033270027991481e-06,
+      "loss": 0.7647,
+      "step": 25655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.032529897611975e-06,
+      "loss": 0.6177,
+      "step": 25656
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.031789803356464e-06,
+      "loss": 0.7319,
+      "step": 25657
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.031049745230328e-06,
+      "loss": 0.543,
+      "step": 25658
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030309723238953e-06,
+      "loss": 0.7187,
+      "step": 25659
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.02956973738771e-06,
+      "loss": 0.729,
+      "step": 25660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.028829787681993e-06,
+      "loss": 0.7782,
+      "step": 25661
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.028089874127176e-06,
+      "loss": 0.9683,
+      "step": 25662
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.027349996728643e-06,
+      "loss": 1.0395,
+      "step": 25663
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.026610155491772e-06,
+      "loss": 0.8682,
+      "step": 25664
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025870350421945e-06,
+      "loss": 1.0772,
+      "step": 25665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025130581524542e-06,
+      "loss": 0.8349,
+      "step": 25666
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.024390848804937e-06,
+      "loss": 0.9141,
+      "step": 25667
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0236511522685185e-06,
+      "loss": 1.1195,
+      "step": 25668
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.022911491920663e-06,
+      "loss": 0.8666,
+      "step": 25669
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0221718677667474e-06,
+      "loss": 0.8942,
+      "step": 25670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0214322798121485e-06,
+      "loss": 0.951,
+      "step": 25671
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.020692728062248e-06,
+      "loss": 0.7747,
+      "step": 25672
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019953212522419e-06,
+      "loss": 1.1778,
+      "step": 25673
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019213733198047e-06,
+      "loss": 0.7776,
+      "step": 25674
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.018474290094506e-06,
+      "loss": 1.0905,
+      "step": 25675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0177348832171715e-06,
+      "loss": 0.6143,
+      "step": 25676
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.016995512571424e-06,
+      "loss": 0.9035,
+      "step": 25677
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.016256178162638e-06,
+      "loss": 0.8297,
+      "step": 25678
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.015516879996191e-06,
+      "loss": 0.8384,
+      "step": 25679
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0147776180774575e-06,
+      "loss": 0.8712,
+      "step": 25680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.014038392411815e-06,
+      "loss": 0.7526,
+      "step": 25681
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.01329920300464e-06,
+      "loss": 0.7874,
+      "step": 25682
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.012560049861307e-06,
+      "loss": 0.7152,
+      "step": 25683
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.01182093298719e-06,
+      "loss": 1.0112,
+      "step": 25684
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.011081852387662e-06,
+      "loss": 0.916,
+      "step": 25685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.010342808068105e-06,
+      "loss": 0.9753,
+      "step": 25686
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.009603800033889e-06,
+      "loss": 1.0414,
+      "step": 25687
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.008864828290387e-06,
+      "loss": 0.9411,
+      "step": 25688
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00812589284298e-06,
+      "loss": 0.9722,
+      "step": 25689
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.007386993697031e-06,
+      "loss": 0.6921,
+      "step": 25690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006648130857914e-06,
+      "loss": 0.6891,
+      "step": 25691
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.005909304331012e-06,
+      "loss": 0.939,
+      "step": 25692
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00517051412169e-06,
+      "loss": 0.6011,
+      "step": 25693
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0044317602353234e-06,
+      "loss": 0.7758,
+      "step": 25694
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.003693042677283e-06,
+      "loss": 0.8988,
+      "step": 25695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.002954361452943e-06,
+      "loss": 1.0657,
+      "step": 25696
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.002215716567673e-06,
+      "loss": 1.0293,
+      "step": 25697
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.001477108026842e-06,
+      "loss": 0.9434,
+      "step": 25698
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.000738535835832e-06,
+      "loss": 1.0842,
+      "step": 25699
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 0.9568,
+      "step": 25700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.999261500524728e-06,
+      "loss": 0.671,
+      "step": 25701
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.998523037415379e-06,
+      "loss": 1.021,
+      "step": 25702
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.997784610677328e-06,
+      "loss": 0.7919,
+      "step": 25703
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.997046220315937e-06,
+      "loss": 1.0903,
+      "step": 25704
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996307866336585e-06,
+      "loss": 0.7305,
+      "step": 25705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.995569548744639e-06,
+      "loss": 0.781,
+      "step": 25706
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9948312675454655e-06,
+      "loss": 0.9113,
+      "step": 25707
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9940930227444366e-06,
+      "loss": 1.0618,
+      "step": 25708
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.993354814346918e-06,
+      "loss": 0.9269,
+      "step": 25709
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.992616642358279e-06,
+      "loss": 0.9687,
+      "step": 25710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9918785067838885e-06,
+      "loss": 1.1167,
+      "step": 25711
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9911404076291135e-06,
+      "loss": 0.7972,
+      "step": 25712
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.990402344899321e-06,
+      "loss": 0.8857,
+      "step": 25713
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.989664318599881e-06,
+      "loss": 0.9214,
+      "step": 25714
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.988926328736157e-06,
+      "loss": 0.5298,
+      "step": 25715
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.988188375313513e-06,
+      "loss": 0.7955,
+      "step": 25716
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.987450458337323e-06,
+      "loss": 0.8374,
+      "step": 25717
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986712577812951e-06,
+      "loss": 0.6898,
+      "step": 25718
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.985974733745765e-06,
+      "loss": 0.91,
+      "step": 25719
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.985236926141125e-06,
+      "loss": 0.6123,
+      "step": 25720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984499155004398e-06,
+      "loss": 0.8885,
+      "step": 25721
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.983761420340947e-06,
+      "loss": 0.8625,
+      "step": 25722
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.983023722156143e-06,
+      "loss": 1.0528,
+      "step": 25723
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.982286060455347e-06,
+      "loss": 0.7652,
+      "step": 25724
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9815484352439244e-06,
+      "loss": 0.8376,
+      "step": 25725
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9808108465272395e-06,
+      "loss": 1.0245,
+      "step": 25726
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.980073294310656e-06,
+      "loss": 0.668,
+      "step": 25727
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979335778599536e-06,
+      "loss": 0.8958,
+      "step": 25728
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.978598299399245e-06,
+      "loss": 0.5758,
+      "step": 25729
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.977860856715143e-06,
+      "loss": 0.8235,
+      "step": 25730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9771234505525945e-06,
+      "loss": 0.7914,
+      "step": 25731
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.976386080916963e-06,
+      "loss": 0.5212,
+      "step": 25732
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.975648747813609e-06,
+      "loss": 0.6319,
+      "step": 25733
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974911451247893e-06,
+      "loss": 0.7055,
+      "step": 25734
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974174191225182e-06,
+      "loss": 0.9382,
+      "step": 25735
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.973436967750834e-06,
+      "loss": 0.8101,
+      "step": 25736
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.97269978083021e-06,
+      "loss": 0.9997,
+      "step": 25737
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971962630468673e-06,
+      "loss": 0.5347,
+      "step": 25738
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971225516671585e-06,
+      "loss": 0.8281,
+      "step": 25739
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.970488439444296e-06,
+      "loss": 1.0216,
+      "step": 25740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.969751398792178e-06,
+      "loss": 0.7821,
+      "step": 25741
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.969014394720585e-06,
+      "loss": 0.9664,
+      "step": 25742
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.968277427234879e-06,
+      "loss": 0.4604,
+      "step": 25743
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.967540496340418e-06,
+      "loss": 0.7445,
+      "step": 25744
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9668036020425615e-06,
+      "loss": 0.6297,
+      "step": 25745
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.966066744346668e-06,
+      "loss": 0.8012,
+      "step": 25746
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.965329923258092e-06,
+      "loss": 0.7384,
+      "step": 25747
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.964593138782201e-06,
+      "loss": 0.5811,
+      "step": 25748
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9638563909243506e-06,
+      "loss": 0.8577,
+      "step": 25749
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.963119679689892e-06,
+      "loss": 0.7335,
+      "step": 25750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.962383005084187e-06,
+      "loss": 0.7703,
+      "step": 25751
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.961646367112588e-06,
+      "loss": 1.0689,
+      "step": 25752
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9609097657804595e-06,
+      "loss": 0.7767,
+      "step": 25753
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.960173201093155e-06,
+      "loss": 0.6805,
+      "step": 25754
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95943667305603e-06,
+      "loss": 0.5401,
+      "step": 25755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9587001816744425e-06,
+      "loss": 0.8565,
+      "step": 25756
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957963726953747e-06,
+      "loss": 0.7693,
+      "step": 25757
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957227308899298e-06,
+      "loss": 0.5706,
+      "step": 25758
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.956490927516454e-06,
+      "loss": 0.7896,
+      "step": 25759
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.955754582810567e-06,
+      "loss": 0.9627,
+      "step": 25760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.955018274786993e-06,
+      "loss": 0.8151,
+      "step": 25761
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.954282003451085e-06,
+      "loss": 0.8427,
+      "step": 25762
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9535457688082e-06,
+      "loss": 0.6613,
+      "step": 25763
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95280957086369e-06,
+      "loss": 0.6252,
+      "step": 25764
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.952073409622906e-06,
+      "loss": 0.9126,
+      "step": 25765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.951337285091208e-06,
+      "loss": 0.9387,
+      "step": 25766
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.950601197273946e-06,
+      "loss": 1.0384,
+      "step": 25767
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949865146176473e-06,
+      "loss": 0.8725,
+      "step": 25768
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949129131804146e-06,
+      "loss": 1.097,
+      "step": 25769
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.948393154162303e-06,
+      "loss": 0.5394,
+      "step": 25770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.947657213256312e-06,
+      "loss": 0.6915,
+      "step": 25771
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.946921309091518e-06,
+      "loss": 0.9752,
+      "step": 25772
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.946185441673273e-06,
+      "loss": 1.0038,
+      "step": 25773
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.945449611006929e-06,
+      "loss": 0.6434,
+      "step": 25774
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.944713817097837e-06,
+      "loss": 0.7645,
+      "step": 25775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9439780599513465e-06,
+      "loss": 1.1048,
+      "step": 25776
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.943242339572806e-06,
+      "loss": 0.8031,
+      "step": 25777
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.942506655967573e-06,
+      "loss": 1.0234,
+      "step": 25778
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.941771009140996e-06,
+      "loss": 0.9985,
+      "step": 25779
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9410353990984184e-06,
+      "loss": 0.8501,
+      "step": 25780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.940299825845194e-06,
+      "loss": 0.7089,
+      "step": 25781
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.93956428938667e-06,
+      "loss": 0.7662,
+      "step": 25782
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9388287897281936e-06,
+      "loss": 1.061,
+      "step": 25783
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.93809332687512e-06,
+      "loss": 0.8518,
+      "step": 25784
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.937357900832793e-06,
+      "loss": 0.8585,
+      "step": 25785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936622511606562e-06,
+      "loss": 0.9077,
+      "step": 25786
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.935887159201774e-06,
+      "loss": 0.9067,
+      "step": 25787
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.935151843623778e-06,
+      "loss": 1.0134,
+      "step": 25788
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.934416564877919e-06,
+      "loss": 0.9986,
+      "step": 25789
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.933681322969546e-06,
+      "loss": 0.7301,
+      "step": 25790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.932946117904005e-06,
+      "loss": 0.8314,
+      "step": 25791
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.932210949686642e-06,
+      "loss": 0.6628,
+      "step": 25792
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.931475818322804e-06,
+      "loss": 0.7251,
+      "step": 25793
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.930740723817837e-06,
+      "loss": 1.4027,
+      "step": 25794
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.930005666177083e-06,
+      "loss": 1.1063,
+      "step": 25795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.929270645405894e-06,
+      "loss": 1.0957,
+      "step": 25796
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9285356615096114e-06,
+      "loss": 0.9595,
+      "step": 25797
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.927800714493582e-06,
+      "loss": 0.9056,
+      "step": 25798
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.927065804363152e-06,
+      "loss": 1.1274,
+      "step": 25799
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.926330931123659e-06,
+      "loss": 0.9384,
+      "step": 25800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.925596094780446e-06,
+      "loss": 0.7302,
+      "step": 25801
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9248612953388665e-06,
+      "loss": 0.8108,
+      "step": 25802
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.92412653280426e-06,
+      "loss": 0.8607,
+      "step": 25803
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.923391807181968e-06,
+      "loss": 0.8807,
+      "step": 25804
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.922657118477334e-06,
+      "loss": 0.7046,
+      "step": 25805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.921922466695703e-06,
+      "loss": 0.8122,
+      "step": 25806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9211878518424146e-06,
+      "loss": 0.9069,
+      "step": 25807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9204532739228085e-06,
+      "loss": 0.9019,
+      "step": 25808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.919718732942238e-06,
+      "loss": 0.9169,
+      "step": 25809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.918984228906033e-06,
+      "loss": 0.7618,
+      "step": 25810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.918249761819539e-06,
+      "loss": 0.9574,
+      "step": 25811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.917515331688097e-06,
+      "loss": 1.008,
+      "step": 25812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.916780938517045e-06,
+      "loss": 0.8899,
+      "step": 25813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.91604658231173e-06,
+      "loss": 0.5946,
+      "step": 25814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.915312263077488e-06,
+      "loss": 0.6198,
+      "step": 25815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.91457798081966e-06,
+      "loss": 0.9851,
+      "step": 25816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.913843735543586e-06,
+      "loss": 0.7012,
+      "step": 25817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9131095272546045e-06,
+      "loss": 0.8309,
+      "step": 25818
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.912375355958056e-06,
+      "loss": 0.7795,
+      "step": 25819
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.911641221659279e-06,
+      "loss": 0.939,
+      "step": 25820
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910907124363612e-06,
+      "loss": 0.7693,
+      "step": 25821
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910173064076394e-06,
+      "loss": 0.8304,
+      "step": 25822
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.909439040802961e-06,
+      "loss": 0.7224,
+      "step": 25823
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908705054548655e-06,
+      "loss": 0.7539,
+      "step": 25824
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90797110531881e-06,
+      "loss": 0.9624,
+      "step": 25825
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.90723719311876e-06,
+      "loss": 0.4975,
+      "step": 25826
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.906503317953851e-06,
+      "loss": 0.9298,
+      "step": 25827
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.905769479829415e-06,
+      "loss": 0.8427,
+      "step": 25828
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.905035678750794e-06,
+      "loss": 0.5375,
+      "step": 25829
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.904301914723315e-06,
+      "loss": 0.46,
+      "step": 25830
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.903568187752314e-06,
+      "loss": 0.8286,
+      "step": 25831
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9028344978431366e-06,
+      "loss": 0.7591,
+      "step": 25832
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.902100845001111e-06,
+      "loss": 0.8104,
+      "step": 25833
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901367229231575e-06,
+      "loss": 0.9764,
+      "step": 25834
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.900633650539862e-06,
+      "loss": 0.8459,
+      "step": 25835
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899900108931308e-06,
+      "loss": 0.6694,
+      "step": 25836
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899166604411247e-06,
+      "loss": 0.7345,
+      "step": 25837
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.898433136985009e-06,
+      "loss": 1.1336,
+      "step": 25838
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.897699706657939e-06,
+      "loss": 0.5544,
+      "step": 25839
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.89696631343536e-06,
+      "loss": 0.7091,
+      "step": 25840
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8962329573226076e-06,
+      "loss": 0.7316,
+      "step": 25841
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.895499638325016e-06,
+      "loss": 1.1803,
+      "step": 25842
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.894766356447918e-06,
+      "loss": 0.8515,
+      "step": 25843
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8940331116966435e-06,
+      "loss": 0.9587,
+      "step": 25844
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.89329990407653e-06,
+      "loss": 0.9554,
+      "step": 25845
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.892566733592907e-06,
+      "loss": 0.6604,
+      "step": 25846
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.891833600251106e-06,
+      "loss": 0.4373,
+      "step": 25847
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.891100504056457e-06,
+      "loss": 0.744,
+      "step": 25848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890367445014294e-06,
+      "loss": 0.7434,
+      "step": 25849
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.889634423129946e-06,
+      "loss": 0.5091,
+      "step": 25850
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.888901438408743e-06,
+      "loss": 0.718,
+      "step": 25851
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8881684908560165e-06,
+      "loss": 0.7163,
+      "step": 25852
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.887435580477098e-06,
+      "loss": 0.6882,
+      "step": 25853
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.886702707277314e-06,
+      "loss": 0.5199,
+      "step": 25854
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.885969871261997e-06,
+      "loss": 0.9456,
+      "step": 25855
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.88523707243647e-06,
+      "loss": 0.8863,
+      "step": 25856
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.884504310806072e-06,
+      "loss": 0.5821,
+      "step": 25857
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.883771586376126e-06,
+      "loss": 0.9175,
+      "step": 25858
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.883038899151965e-06,
+      "loss": 0.9673,
+      "step": 25859
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.882306249138909e-06,
+      "loss": 0.8017,
+      "step": 25860
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881573636342292e-06,
+      "loss": 0.7884,
+      "step": 25861
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.880841060767434e-06,
+      "loss": 1.0745,
+      "step": 25862
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.880108522419673e-06,
+      "loss": 0.7328,
+      "step": 25863
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.879376021304332e-06,
+      "loss": 0.8215,
+      "step": 25864
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.878643557426737e-06,
+      "loss": 1.0281,
+      "step": 25865
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.877911130792214e-06,
+      "loss": 0.68,
+      "step": 25866
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87717874140609e-06,
+      "loss": 0.9084,
+      "step": 25867
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.876446389273692e-06,
+      "loss": 0.7401,
+      "step": 25868
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.875714074400344e-06,
+      "loss": 0.6519,
+      "step": 25869
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874981796791372e-06,
+      "loss": 0.7878,
+      "step": 25870
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874249556452102e-06,
+      "loss": 0.6536,
+      "step": 25871
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.873517353387859e-06,
+      "loss": 0.7731,
+      "step": 25872
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.872785187603967e-06,
+      "loss": 0.8747,
+      "step": 25873
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.872053059105746e-06,
+      "loss": 0.779,
+      "step": 25874
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.871320967898528e-06,
+      "loss": 0.8026,
+      "step": 25875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.870588913987634e-06,
+      "loss": 0.5574,
+      "step": 25876
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.869856897378388e-06,
+      "loss": 1.1037,
+      "step": 25877
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8691249180761115e-06,
+      "loss": 0.6956,
+      "step": 25878
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.868392976086131e-06,
+      "loss": 0.8357,
+      "step": 25879
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.86766107141376e-06,
+      "loss": 0.7268,
+      "step": 25880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8669292040643326e-06,
+      "loss": 0.8079,
+      "step": 25881
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.866197374043165e-06,
+      "loss": 0.6186,
+      "step": 25882
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.865465581355581e-06,
+      "loss": 0.8033,
+      "step": 25883
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8647338260069e-06,
+      "loss": 0.8421,
+      "step": 25884
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.864002108002446e-06,
+      "loss": 0.7428,
+      "step": 25885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8632704273475385e-06,
+      "loss": 0.6707,
+      "step": 25886
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8625387840474945e-06,
+      "loss": 0.5855,
+      "step": 25887
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.861807178107644e-06,
+      "loss": 0.8653,
+      "step": 25888
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.861075609533305e-06,
+      "loss": 0.7085,
+      "step": 25889
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.860344078329791e-06,
+      "loss": 0.8833,
+      "step": 25890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.859612584502427e-06,
+      "loss": 0.8102,
+      "step": 25891
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.858881128056526e-06,
+      "loss": 0.7366,
+      "step": 25892
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.858149708997417e-06,
+      "loss": 0.4838,
+      "step": 25893
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8574183273304135e-06,
+      "loss": 1.1257,
+      "step": 25894
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.856686983060835e-06,
+      "loss": 0.8883,
+      "step": 25895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8559556761940005e-06,
+      "loss": 0.8575,
+      "step": 25896
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8552244067352275e-06,
+      "loss": 0.5298,
+      "step": 25897
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.854493174689834e-06,
+      "loss": 0.7528,
+      "step": 25898
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8537619800631376e-06,
+      "loss": 0.9391,
+      "step": 25899
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.853030822860455e-06,
+      "loss": 0.6826,
+      "step": 25900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.852299703087104e-06,
+      "loss": 1.1167,
+      "step": 25901
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.851568620748401e-06,
+      "loss": 1.0295,
+      "step": 25902
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.850837575849663e-06,
+      "loss": 0.7654,
+      "step": 25903
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.850106568396207e-06,
+      "loss": 1.0169,
+      "step": 25904
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849375598393342e-06,
+      "loss": 0.9318,
+      "step": 25905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8486446658463945e-06,
+      "loss": 0.8658,
+      "step": 25906
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.847913770760675e-06,
+      "loss": 1.2885,
+      "step": 25907
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8471829131414995e-06,
+      "loss": 0.7918,
+      "step": 25908
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.846452092994185e-06,
+      "loss": 1.0346,
+      "step": 25909
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8457213103240355e-06,
+      "loss": 0.9897,
+      "step": 25910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.844990565136377e-06,
+      "loss": 0.8505,
+      "step": 25911
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.84425985743652e-06,
+      "loss": 1.1902,
+      "step": 25912
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.843529187229776e-06,
+      "loss": 0.8377,
+      "step": 25913
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8427985545214615e-06,
+      "loss": 0.7601,
+      "step": 25914
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8420679593168895e-06,
+      "loss": 0.9955,
+      "step": 25915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.841337401621371e-06,
+      "loss": 0.7009,
+      "step": 25916
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840606881440216e-06,
+      "loss": 0.7737,
+      "step": 25917
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.839876398778745e-06,
+      "loss": 0.9382,
+      "step": 25918
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8391459536422705e-06,
+      "loss": 0.7207,
+      "step": 25919
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838415546036095e-06,
+      "loss": 0.7175,
+      "step": 25920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.837685175965534e-06,
+      "loss": 0.7223,
+      "step": 25921
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8369548434359e-06,
+      "loss": 0.8421,
+      "step": 25922
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8362245484525e-06,
+      "loss": 0.8953,
+      "step": 25923
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8354942910206535e-06,
+      "loss": 0.8402,
+      "step": 25924
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.834764071145666e-06,
+      "loss": 1.0377,
+      "step": 25925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.834033888832847e-06,
+      "loss": 0.8998,
+      "step": 25926
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.833303744087506e-06,
+      "loss": 0.8034,
+      "step": 25927
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8325736369149554e-06,
+      "loss": 0.8432,
+      "step": 25928
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.831843567320503e-06,
+      "loss": 0.992,
+      "step": 25929
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8311135353094574e-06,
+      "loss": 0.9153,
+      "step": 25930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.830383540887129e-06,
+      "loss": 0.5498,
+      "step": 25931
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8296535840588255e-06,
+      "loss": 1.1192,
+      "step": 25932
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.828923664829854e-06,
+      "loss": 1.0304,
+      "step": 25933
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8281937832055256e-06,
+      "loss": 1.1088,
+      "step": 25934
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.827463939191141e-06,
+      "loss": 0.9673,
+      "step": 25935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826734132792018e-06,
+      "loss": 0.9557,
+      "step": 25936
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826004364013459e-06,
+      "loss": 0.9795,
+      "step": 25937
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.825274632860771e-06,
+      "loss": 1.1354,
+      "step": 25938
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.824544939339263e-06,
+      "loss": 1.0189,
+      "step": 25939
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.823815283454235e-06,
+      "loss": 0.8262,
+      "step": 25940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.823085665210995e-06,
+      "loss": 0.7449,
+      "step": 25941
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.822356084614853e-06,
+      "loss": 0.8997,
+      "step": 25942
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.821626541671114e-06,
+      "loss": 0.9738,
+      "step": 25943
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.82089703638508e-06,
+      "loss": 0.8959,
+      "step": 25944
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8201675687620596e-06,
+      "loss": 0.7399,
+      "step": 25945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.819438138807355e-06,
+      "loss": 0.8741,
+      "step": 25946
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.818708746526272e-06,
+      "loss": 0.6758,
+      "step": 25947
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.81797939192411e-06,
+      "loss": 0.7296,
+      "step": 25948
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.817250075006186e-06,
+      "loss": 1.0669,
+      "step": 25949
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.816520795777789e-06,
+      "loss": 0.6882,
+      "step": 25950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.81579155424423e-06,
+      "loss": 0.9688,
+      "step": 25951
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.815062350410811e-06,
+      "loss": 0.8392,
+      "step": 25952
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.81433318428283e-06,
+      "loss": 0.8824,
+      "step": 25953
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.813604055865597e-06,
+      "loss": 0.9875,
+      "step": 25954
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812874965164412e-06,
+      "loss": 1.0801,
+      "step": 25955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812145912184576e-06,
+      "loss": 0.8558,
+      "step": 25956
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.811416896931391e-06,
+      "loss": 0.8082,
+      "step": 25957
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.810687919410158e-06,
+      "loss": 1.0116,
+      "step": 25958
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.80995897962618e-06,
+      "loss": 0.91,
+      "step": 25959
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.809230077584755e-06,
+      "loss": 0.9745,
+      "step": 25960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.808501213291188e-06,
+      "loss": 0.7455,
+      "step": 25961
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.807772386750774e-06,
+      "loss": 0.9811,
+      "step": 25962
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.807043597968817e-06,
+      "loss": 0.9651,
+      "step": 25963
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.806314846950615e-06,
+      "loss": 0.5958,
+      "step": 25964
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.805586133701468e-06,
+      "loss": 0.8993,
+      "step": 25965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.804857458226673e-06,
+      "loss": 0.6984,
+      "step": 25966
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8041288205315354e-06,
+      "loss": 0.7739,
+      "step": 25967
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.803400220621348e-06,
+      "loss": 0.76,
+      "step": 25968
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8026716585014165e-06,
+      "loss": 0.7797,
+      "step": 25969
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.80194313417703e-06,
+      "loss": 0.8219,
+      "step": 25970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801214647653486e-06,
+      "loss": 0.6722,
+      "step": 25971
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.800486198936091e-06,
+      "loss": 1.1309,
+      "step": 25972
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799757788030136e-06,
+      "loss": 0.9706,
+      "step": 25973
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799029414940921e-06,
+      "loss": 1.1024,
+      "step": 25974
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7983010796737415e-06,
+      "loss": 0.5856,
+      "step": 25975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797572782233895e-06,
+      "loss": 0.755,
+      "step": 25976
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7968445226266755e-06,
+      "loss": 0.7972,
+      "step": 25977
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.796116300857378e-06,
+      "loss": 0.8916,
+      "step": 25978
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795388116931308e-06,
+      "loss": 0.9459,
+      "step": 25979
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7946599708537485e-06,
+      "loss": 0.8861,
+      "step": 25980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.79393186263e-06,
+      "loss": 0.8965,
+      "step": 25981
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.793203792265359e-06,
+      "loss": 0.9075,
+      "step": 25982
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7924757597651165e-06,
+      "loss": 0.9354,
+      "step": 25983
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7917477651345645e-06,
+      "loss": 1.0616,
+      "step": 25984
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.791019808379006e-06,
+      "loss": 0.8306,
+      "step": 25985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.79029188950373e-06,
+      "loss": 0.6646,
+      "step": 25986
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.789564008514029e-06,
+      "loss": 0.8458,
+      "step": 25987
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.788836165415197e-06,
+      "loss": 0.8815,
+      "step": 25988
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.788108360212528e-06,
+      "loss": 0.7865,
+      "step": 25989
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7873805929113134e-06,
+      "loss": 0.6744,
+      "step": 25990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.786652863516846e-06,
+      "loss": 0.7504,
+      "step": 25991
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.785925172034418e-06,
+      "loss": 0.7184,
+      "step": 25992
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7851975184693215e-06,
+      "loss": 0.6249,
+      "step": 25993
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.784469902826847e-06,
+      "loss": 0.8022,
+      "step": 25994
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783742325112286e-06,
+      "loss": 0.8999,
+      "step": 25995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783014785330927e-06,
+      "loss": 0.7174,
+      "step": 25996
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7822872834880685e-06,
+      "loss": 0.5234,
+      "step": 25997
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7815598195889955e-06,
+      "loss": 0.8778,
+      "step": 25998
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.780832393639003e-06,
+      "loss": 0.9708,
+      "step": 25999
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.780105005643372e-06,
+      "loss": 0.8299,
+      "step": 26000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7793776556073975e-06,
+      "loss": 1.0347,
+      "step": 26001
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778650343536366e-06,
+      "loss": 0.7918,
+      "step": 26002
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.777923069435572e-06,
+      "loss": 0.7127,
+      "step": 26003
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7771958333103006e-06,
+      "loss": 0.887,
+      "step": 26004
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.776468635165842e-06,
+      "loss": 1.0257,
+      "step": 26005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775741475007482e-06,
+      "loss": 0.7738,
+      "step": 26006
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775014352840512e-06,
+      "loss": 0.7064,
+      "step": 26007
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.774287268670218e-06,
+      "loss": 0.7885,
+      "step": 26008
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.773560222501886e-06,
+      "loss": 0.7669,
+      "step": 26009
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772833214340805e-06,
+      "loss": 0.7347,
+      "step": 26010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772106244192262e-06,
+      "loss": 0.7762,
+      "step": 26011
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.771379312061541e-06,
+      "loss": 0.7834,
+      "step": 26012
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7706524179539315e-06,
+      "loss": 1.0165,
+      "step": 26013
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769925561874715e-06,
+      "loss": 0.8129,
+      "step": 26014
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7691987438291845e-06,
+      "loss": 0.9559,
+      "step": 26015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.768471963822621e-06,
+      "loss": 1.0761,
+      "step": 26016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.767745221860309e-06,
+      "loss": 0.8812,
+      "step": 26017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.767018517947541e-06,
+      "loss": 0.8408,
+      "step": 26018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.76629185208959e-06,
+      "loss": 0.757,
+      "step": 26019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7655652242917425e-06,
+      "loss": 1.0131,
+      "step": 26020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.76483863455929e-06,
+      "loss": 0.7706,
+      "step": 26021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.764112082897512e-06,
+      "loss": 0.9883,
+      "step": 26022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7633855693116925e-06,
+      "loss": 0.7445,
+      "step": 26023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762659093807115e-06,
+      "loss": 0.8526,
+      "step": 26024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.761932656389061e-06,
+      "loss": 0.7822,
+      "step": 26025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7612062570628155e-06,
+      "loss": 0.9451,
+      "step": 26026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760479895833657e-06,
+      "loss": 0.9994,
+      "step": 26027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7597535727068755e-06,
+      "loss": 0.8058,
+      "step": 26028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.759027287687746e-06,
+      "loss": 1.1737,
+      "step": 26029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758301040781551e-06,
+      "loss": 0.9735,
+      "step": 26030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7575748319935746e-06,
+      "loss": 0.7213,
+      "step": 26031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.756848661329091e-06,
+      "loss": 1.085,
+      "step": 26032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75612252879339e-06,
+      "loss": 0.6502,
+      "step": 26033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.755396434391748e-06,
+      "loss": 0.943,
+      "step": 26034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7546703781294464e-06,
+      "loss": 0.7054,
+      "step": 26035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753944360011764e-06,
+      "loss": 0.7922,
+      "step": 26036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753218380043981e-06,
+      "loss": 0.9639,
+      "step": 26037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.752492438231376e-06,
+      "loss": 0.8765,
+      "step": 26038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75176653457923e-06,
+      "loss": 1.0699,
+      "step": 26039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751040669092819e-06,
+      "loss": 1.095,
+      "step": 26040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.750314841777424e-06,
+      "loss": 0.5131,
+      "step": 26041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749589052638321e-06,
+      "loss": 0.8874,
+      "step": 26042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.748863301680792e-06,
+      "loss": 0.7363,
+      "step": 26043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.74813758891011e-06,
+      "loss": 0.7705,
+      "step": 26044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.747411914331552e-06,
+      "loss": 0.9021,
+      "step": 26045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7466862779504005e-06,
+      "loss": 0.8275,
+      "step": 26046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.745960679771932e-06,
+      "loss": 0.9028,
+      "step": 26047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.745235119801423e-06,
+      "loss": 0.8137,
+      "step": 26048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7445095980441435e-06,
+      "loss": 1.2035,
+      "step": 26049
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.743784114505371e-06,
+      "loss": 0.7838,
+      "step": 26050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.743058669190388e-06,
+      "loss": 0.7975,
+      "step": 26051
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.742333262104465e-06,
+      "loss": 0.7215,
+      "step": 26052
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.741607893252879e-06,
+      "loss": 0.9043,
+      "step": 26053
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740882562640905e-06,
+      "loss": 0.6471,
+      "step": 26054
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740157270273816e-06,
+      "loss": 0.8415,
+      "step": 26055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.739432016156888e-06,
+      "loss": 0.9572,
+      "step": 26056
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.738706800295391e-06,
+      "loss": 0.7607,
+      "step": 26057
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737981622694609e-06,
+      "loss": 0.5981,
+      "step": 26058
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737256483359807e-06,
+      "loss": 0.5963,
+      "step": 26059
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736531382296257e-06,
+      "loss": 0.8474,
+      "step": 26060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.735806319509239e-06,
+      "loss": 0.7637,
+      "step": 26061
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.73508129500402e-06,
+      "loss": 0.8419,
+      "step": 26062
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7343563087858715e-06,
+      "loss": 0.7668,
+      "step": 26063
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7336313608600715e-06,
+      "loss": 0.6114,
+      "step": 26064
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.73290645123189e-06,
+      "loss": 0.6915,
+      "step": 26065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7321815799065975e-06,
+      "loss": 0.7677,
+      "step": 26066
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.731456746889464e-06,
+      "loss": 0.9413,
+      "step": 26067
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.730731952185764e-06,
+      "loss": 0.5038,
+      "step": 26068
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.730007195800765e-06,
+      "loss": 0.7305,
+      "step": 26069
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729282477739741e-06,
+      "loss": 0.7319,
+      "step": 26070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.728557798007957e-06,
+      "loss": 0.7409,
+      "step": 26071
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.727833156610689e-06,
+      "loss": 0.682,
+      "step": 26072
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.727108553553202e-06,
+      "loss": 0.6699,
+      "step": 26073
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726383988840768e-06,
+      "loss": 0.7703,
+      "step": 26074
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.725659462478651e-06,
+      "loss": 1.0045,
+      "step": 26075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.724934974472129e-06,
+      "loss": 0.8802,
+      "step": 26076
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7242105248264645e-06,
+      "loss": 0.9086,
+      "step": 26077
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.723486113546931e-06,
+      "loss": 0.4651,
+      "step": 26078
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722761740638788e-06,
+      "loss": 0.708,
+      "step": 26079
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722037406107308e-06,
+      "loss": 0.5654,
+      "step": 26080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.721313109957755e-06,
+      "loss": 1.1637,
+      "step": 26081
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.720588852195402e-06,
+      "loss": 0.5882,
+      "step": 26082
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.719864632825514e-06,
+      "loss": 0.7672,
+      "step": 26083
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.719140451853357e-06,
+      "loss": 1.0454,
+      "step": 26084
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.718416309284196e-06,
+      "loss": 0.7985,
+      "step": 26085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7176922051232985e-06,
+      "loss": 0.64,
+      "step": 26086
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716968139375931e-06,
+      "loss": 1.1376,
+      "step": 26087
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716244112047356e-06,
+      "loss": 0.8265,
+      "step": 26088
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.715520123142841e-06,
+      "loss": 0.6369,
+      "step": 26089
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.714796172667652e-06,
+      "loss": 0.6164,
+      "step": 26090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.714072260627051e-06,
+      "loss": 0.6135,
+      "step": 26091
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.713348387026303e-06,
+      "loss": 1.0993,
+      "step": 26092
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.71262455187067e-06,
+      "loss": 0.8259,
+      "step": 26093
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7119007551654205e-06,
+      "loss": 0.8093,
+      "step": 26094
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.711176996915817e-06,
+      "loss": 0.9114,
+      "step": 26095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.710453277127122e-06,
+      "loss": 0.5489,
+      "step": 26096
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.709729595804597e-06,
+      "loss": 0.8151,
+      "step": 26097
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7090059529535115e-06,
+      "loss": 0.8966,
+      "step": 26098
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.708282348579113e-06,
+      "loss": 0.6997,
+      "step": 26099
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.707558782686677e-06,
+      "loss": 0.9957,
+      "step": 26100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.706835255281462e-06,
+      "loss": 0.728,
+      "step": 26101
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.706111766368728e-06,
+      "loss": 0.8071,
+      "step": 26102
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.705388315953739e-06,
+      "loss": 0.9133,
+      "step": 26103
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704664904041752e-06,
+      "loss": 0.7512,
+      "step": 26104
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.70394153063803e-06,
+      "loss": 0.9277,
+      "step": 26105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7032181957478295e-06,
+      "loss": 1.0279,
+      "step": 26106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.702494899376419e-06,
+      "loss": 0.6119,
+      "step": 26107
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.701771641529057e-06,
+      "loss": 0.8657,
+      "step": 26108
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.701048422210997e-06,
+      "loss": 1.2009,
+      "step": 26109
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.700325241427501e-06,
+      "loss": 0.7805,
+      "step": 26110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.699602099183824e-06,
+      "loss": 0.5677,
+      "step": 26111
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.698878995485233e-06,
+      "loss": 1.1736,
+      "step": 26112
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.698155930336984e-06,
+      "loss": 0.8465,
+      "step": 26113
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6974329037443325e-06,
+      "loss": 0.8629,
+      "step": 26114
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6967099157125384e-06,
+      "loss": 1.0886,
+      "step": 26115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.695986966246858e-06,
+      "loss": 1.0248,
+      "step": 26116
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69526405535255e-06,
+      "loss": 0.9644,
+      "step": 26117
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.694541183034871e-06,
+      "loss": 1.0608,
+      "step": 26118
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6938183492990766e-06,
+      "loss": 0.9581,
+      "step": 26119
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.693095554150425e-06,
+      "loss": 0.8856,
+      "step": 26120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692372797594172e-06,
+      "loss": 0.8387,
+      "step": 26121
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.691650079635574e-06,
+      "loss": 0.7028,
+      "step": 26122
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690927400279886e-06,
+      "loss": 0.7865,
+      "step": 26123
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.690204759532359e-06,
+      "loss": 0.9014,
+      "step": 26124
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.689482157398256e-06,
+      "loss": 0.5082,
+      "step": 26125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.688759593882829e-06,
+      "loss": 0.8454,
+      "step": 26126
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.688037068991333e-06,
+      "loss": 1.0464,
+      "step": 26127
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.687314582729023e-06,
+      "loss": 0.9568,
+      "step": 26128
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.686592135101146e-06,
+      "loss": 1.1071,
+      "step": 26129
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.685869726112963e-06,
+      "loss": 0.7213,
+      "step": 26130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.685147355769726e-06,
+      "loss": 0.9349,
+      "step": 26131
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6844250240766875e-06,
+      "loss": 1.1438,
+      "step": 26132
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683702731039101e-06,
+      "loss": 0.8803,
+      "step": 26133
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682980476662219e-06,
+      "loss": 0.841,
+      "step": 26134
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682258260951292e-06,
+      "loss": 0.9169,
+      "step": 26135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.681536083911571e-06,
+      "loss": 0.9445,
+      "step": 26136
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6808139455483135e-06,
+      "loss": 0.796,
+      "step": 26137
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6800918458667725e-06,
+      "loss": 0.9369,
+      "step": 26138
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.679369784872188e-06,
+      "loss": 0.8007,
+      "step": 26139
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.67864776256982e-06,
+      "loss": 0.6285,
+      "step": 26140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.677925778964914e-06,
+      "loss": 0.6399,
+      "step": 26141
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.677203834062721e-06,
+      "loss": 0.855,
+      "step": 26142
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.676481927868496e-06,
+      "loss": 0.6731,
+      "step": 26143
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675760060387485e-06,
+      "loss": 0.7964,
+      "step": 26144
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675038231624939e-06,
+      "loss": 1.1698,
+      "step": 26145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.674316441586105e-06,
+      "loss": 0.651,
+      "step": 26146
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.673594690276234e-06,
+      "loss": 0.9998,
+      "step": 26147
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6728729777005745e-06,
+      "loss": 0.7617,
+      "step": 26148
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6721513038643725e-06,
+      "loss": 0.8328,
+      "step": 26149
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.67142966877288e-06,
+      "loss": 0.6261,
+      "step": 26150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.670708072431341e-06,
+      "loss": 1.0071,
+      "step": 26151
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.669986514845006e-06,
+      "loss": 0.7743,
+      "step": 26152
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.669264996019121e-06,
+      "loss": 0.768,
+      "step": 26153
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6685435159589275e-06,
+      "loss": 0.8273,
+      "step": 26154
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.667822074669682e-06,
+      "loss": 0.8235,
+      "step": 26155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.667100672156627e-06,
+      "loss": 0.7288,
+      "step": 26156
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.666379308425009e-06,
+      "loss": 0.6898,
+      "step": 26157
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.665657983480075e-06,
+      "loss": 0.8492,
+      "step": 26158
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.664936697327065e-06,
+      "loss": 0.746,
+      "step": 26159
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.664215449971225e-06,
+      "loss": 0.9983,
+      "step": 26160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.663494241417805e-06,
+      "loss": 0.8655,
+      "step": 26161
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6627730716720485e-06,
+      "loss": 0.8513,
+      "step": 26162
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.662051940739199e-06,
+      "loss": 0.6037,
+      "step": 26163
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.661330848624501e-06,
+      "loss": 0.7932,
+      "step": 26164
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.660609795333197e-06,
+      "loss": 0.892,
+      "step": 26165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.659888780870532e-06,
+      "loss": 0.8452,
+      "step": 26166
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.659167805241744e-06,
+      "loss": 0.8534,
+      "step": 26167
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65844686845209e-06,
+      "loss": 0.4815,
+      "step": 26168
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.657725970506799e-06,
+      "loss": 0.7818,
+      "step": 26169
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6570051114111185e-06,
+      "loss": 0.8731,
+      "step": 26170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.656284291170289e-06,
+      "loss": 0.7155,
+      "step": 26171
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65556350978955e-06,
+      "loss": 0.6949,
+      "step": 26172
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.654842767274151e-06,
+      "loss": 1.1969,
+      "step": 26173
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.654122063629329e-06,
+      "loss": 0.8521,
+      "step": 26174
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.653401398860324e-06,
+      "loss": 0.819,
+      "step": 26175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.652680772972379e-06,
+      "loss": 0.9667,
+      "step": 26176
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.651960185970733e-06,
+      "loss": 0.8409,
+      "step": 26177
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.651239637860626e-06,
+      "loss": 0.6307,
+      "step": 26178
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.650519128647298e-06,
+      "loss": 0.6412,
+      "step": 26179
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.64979865833599e-06,
+      "loss": 0.9125,
+      "step": 26180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.649078226931939e-06,
+      "loss": 0.5917,
+      "step": 26181
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.648357834440386e-06,
+      "loss": 0.7725,
+      "step": 26182
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6476374808665695e-06,
+      "loss": 0.9197,
+      "step": 26183
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.646917166215728e-06,
+      "loss": 1.0098,
+      "step": 26184
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.646196890493094e-06,
+      "loss": 0.7868,
+      "step": 26185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6454766537039154e-06,
+      "loss": 0.6673,
+      "step": 26186
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.644756455853426e-06,
+      "loss": 0.8363,
+      "step": 26187
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6440362969468645e-06,
+      "loss": 0.7237,
+      "step": 26188
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.643316176989462e-06,
+      "loss": 0.7865,
+      "step": 26189
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6425960959864556e-06,
+      "loss": 0.6261,
+      "step": 26190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6418760539430905e-06,
+      "loss": 1.058,
+      "step": 26191
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.641156050864597e-06,
+      "loss": 0.6468,
+      "step": 26192
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.640436086756212e-06,
+      "loss": 0.8666,
+      "step": 26193
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63971616162317e-06,
+      "loss": 1.037,
+      "step": 26194
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638996275470707e-06,
+      "loss": 0.6707,
+      "step": 26195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638276428304059e-06,
+      "loss": 1.0165,
+      "step": 26196
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.637556620128457e-06,
+      "loss": 0.8405,
+      "step": 26197
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.636836850949146e-06,
+      "loss": 0.5832,
+      "step": 26198
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6361171207713495e-06,
+      "loss": 1.0111,
+      "step": 26199
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6353974296003045e-06,
+      "loss": 0.8129,
+      "step": 26200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.634677777441246e-06,
+      "loss": 0.9063,
+      "step": 26201
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.633958164299404e-06,
+      "loss": 0.6758,
+      "step": 26202
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6332385901800135e-06,
+      "loss": 0.9802,
+      "step": 26203
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63251905508831e-06,
+      "loss": 0.6278,
+      "step": 26204
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.631799559029524e-06,
+      "loss": 0.8759,
+      "step": 26205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.631080102008888e-06,
+      "loss": 0.5489,
+      "step": 26206
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.630360684031634e-06,
+      "loss": 1.0018,
+      "step": 26207
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.629641305102993e-06,
+      "loss": 0.8477,
+      "step": 26208
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628921965228197e-06,
+      "loss": 0.7594,
+      "step": 26209
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628202664412477e-06,
+      "loss": 0.9991,
+      "step": 26210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6274834026610625e-06,
+      "loss": 0.7991,
+      "step": 26211
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626764179979187e-06,
+      "loss": 0.8544,
+      "step": 26212
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626044996372079e-06,
+      "loss": 0.7166,
+      "step": 26213
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.625325851844967e-06,
+      "loss": 0.8054,
+      "step": 26214
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.624606746403084e-06,
+      "loss": 0.8277,
+      "step": 26215
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.623887680051653e-06,
+      "loss": 0.7122,
+      "step": 26216
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.623168652795912e-06,
+      "loss": 0.6305,
+      "step": 26217
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6224496646410885e-06,
+      "loss": 0.9397,
+      "step": 26218
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621730715592405e-06,
+      "loss": 1.0464,
+      "step": 26219
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621011805655093e-06,
+      "loss": 0.634,
+      "step": 26220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.620292934834376e-06,
+      "loss": 0.5961,
+      "step": 26221
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.619574103135491e-06,
+      "loss": 1.0784,
+      "step": 26222
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6188553105636604e-06,
+      "loss": 0.9117,
+      "step": 26223
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.618136557124111e-06,
+      "loss": 0.8114,
+      "step": 26224
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61741784282207e-06,
+      "loss": 0.8566,
+      "step": 26225
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.616699167662766e-06,
+      "loss": 0.7646,
+      "step": 26226
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.615980531651422e-06,
+      "loss": 0.8232,
+      "step": 26227
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.615261934793266e-06,
+      "loss": 0.7322,
+      "step": 26228
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.614543377093522e-06,
+      "loss": 0.7093,
+      "step": 26229
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.613824858557419e-06,
+      "loss": 0.8183,
+      "step": 26230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.613106379190178e-06,
+      "loss": 0.7445,
+      "step": 26231
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.612387938997025e-06,
+      "loss": 0.8323,
+      "step": 26232
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.611669537983185e-06,
+      "loss": 0.6792,
+      "step": 26233
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.610951176153881e-06,
+      "loss": 0.8937,
+      "step": 26234
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61023285351434e-06,
+      "loss": 0.7609,
+      "step": 26235
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.609514570069786e-06,
+      "loss": 1.0673,
+      "step": 26236
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6087963258254395e-06,
+      "loss": 0.7429,
+      "step": 26237
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.608078120786529e-06,
+      "loss": 0.7141,
+      "step": 26238
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607359954958266e-06,
+      "loss": 0.745,
+      "step": 26239
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.606641828345884e-06,
+      "loss": 1.0291,
+      "step": 26240
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6059237409546e-06,
+      "loss": 1.0075,
+      "step": 26241
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.605205692789638e-06,
+      "loss": 0.8199,
+      "step": 26242
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.60448768385622e-06,
+      "loss": 0.924,
+      "step": 26243
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.603769714159567e-06,
+      "loss": 0.9217,
+      "step": 26244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.603051783704899e-06,
+      "loss": 0.4179,
+      "step": 26245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.602333892497435e-06,
+      "loss": 1.0146,
+      "step": 26246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.601616040542402e-06,
+      "loss": 0.8746,
+      "step": 26247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6008982278450194e-06,
+      "loss": 0.8348,
+      "step": 26248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.600180454410501e-06,
+      "loss": 0.8555,
+      "step": 26249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.599462720244071e-06,
+      "loss": 0.7198,
+      "step": 26250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.598745025350947e-06,
+      "loss": 0.8532,
+      "step": 26251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5980273697363456e-06,
+      "loss": 0.9832,
+      "step": 26252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.597309753405493e-06,
+      "loss": 0.928,
+      "step": 26253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.596592176363602e-06,
+      "loss": 1.0532,
+      "step": 26254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595874638615894e-06,
+      "loss": 0.7441,
+      "step": 26255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595157140167585e-06,
+      "loss": 1.3092,
+      "step": 26256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5944396810238955e-06,
+      "loss": 0.8534,
+      "step": 26257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5937222611900385e-06,
+      "loss": 1.0171,
+      "step": 26258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593004880671235e-06,
+      "loss": 0.903,
+      "step": 26259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.592287539472701e-06,
+      "loss": 0.8934,
+      "step": 26260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.591570237599651e-06,
+      "loss": 0.7297,
+      "step": 26261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.590852975057304e-06,
+      "loss": 0.7888,
+      "step": 26262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5901357518508735e-06,
+      "loss": 0.9268,
+      "step": 26263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.589418567985574e-06,
+      "loss": 0.7965,
+      "step": 26264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5887014234666275e-06,
+      "loss": 0.7737,
+      "step": 26265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587984318299245e-06,
+      "loss": 0.8795,
+      "step": 26266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.58726725248864e-06,
+      "loss": 0.8443,
+      "step": 26267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.586550226040034e-06,
+      "loss": 0.7841,
+      "step": 26268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.585833238958631e-06,
+      "loss": 0.8746,
+      "step": 26269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.585116291249647e-06,
+      "loss": 1.0442,
+      "step": 26270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.584399382918302e-06,
+      "loss": 0.8561,
+      "step": 26271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.583682513969806e-06,
+      "loss": 0.6559,
+      "step": 26272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582965684409373e-06,
+      "loss": 1.011,
+      "step": 26273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582248894242214e-06,
+      "loss": 0.9853,
+      "step": 26274
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.581532143473543e-06,
+      "loss": 0.8549,
+      "step": 26275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580815432108573e-06,
+      "loss": 0.7289,
+      "step": 26276
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.58009876015251e-06,
+      "loss": 0.8724,
+      "step": 26277
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.579382127610578e-06,
+      "loss": 0.9033,
+      "step": 26278
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.578665534487979e-06,
+      "loss": 0.7682,
+      "step": 26279
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.577948980789924e-06,
+      "loss": 0.7426,
+      "step": 26280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.577232466521628e-06,
+      "loss": 0.6495,
+      "step": 26281
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.576515991688295e-06,
+      "loss": 1.2296,
+      "step": 26282
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575799556295144e-06,
+      "loss": 0.9208,
+      "step": 26283
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575083160347381e-06,
+      "loss": 0.8758,
+      "step": 26284
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.574366803850215e-06,
+      "loss": 0.9911,
+      "step": 26285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.573650486808857e-06,
+      "loss": 0.9554,
+      "step": 26286
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.572934209228516e-06,
+      "loss": 0.8929,
+      "step": 26287
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.572217971114399e-06,
+      "loss": 0.6418,
+      "step": 26288
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.571501772471715e-06,
+      "loss": 0.7378,
+      "step": 26289
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.570785613305674e-06,
+      "loss": 0.7616,
+      "step": 26290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5700694936214815e-06,
+      "loss": 0.684,
+      "step": 26291
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5693534134243475e-06,
+      "loss": 0.6257,
+      "step": 26292
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.56863737271948e-06,
+      "loss": 0.7561,
+      "step": 26293
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567921371512083e-06,
+      "loss": 0.9406,
+      "step": 26294
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567205409807362e-06,
+      "loss": 0.8462,
+      "step": 26295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.566489487610529e-06,
+      "loss": 0.9189,
+      "step": 26296
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5657736049267895e-06,
+      "loss": 1.01,
+      "step": 26297
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.56505776176135e-06,
+      "loss": 0.8552,
+      "step": 26298
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.564341958119412e-06,
+      "loss": 1.129,
+      "step": 26299
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.563626194006178e-06,
+      "loss": 0.8335,
+      "step": 26300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562910469426862e-06,
+      "loss": 0.8301,
+      "step": 26301
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5621947843866645e-06,
+      "loss": 0.8836,
+      "step": 26302
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5614791388907905e-06,
+      "loss": 0.8672,
+      "step": 26303
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.560763532944444e-06,
+      "loss": 0.9007,
+      "step": 26304
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.56004796655283e-06,
+      "loss": 0.8534,
+      "step": 26305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5593324397211505e-06,
+      "loss": 0.8773,
+      "step": 26306
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.558616952454606e-06,
+      "loss": 0.7952,
+      "step": 26307
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.557901504758411e-06,
+      "loss": 0.6443,
+      "step": 26308
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.557186096637757e-06,
+      "loss": 0.6431,
+      "step": 26309
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.556470728097849e-06,
+      "loss": 0.8645,
+      "step": 26310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555755399143892e-06,
+      "loss": 0.7796,
+      "step": 26311
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555040109781086e-06,
+      "loss": 0.8679,
+      "step": 26312
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.554324860014628e-06,
+      "loss": 0.9083,
+      "step": 26313
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5536096498497295e-06,
+      "loss": 0.9443,
+      "step": 26314
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.552894479291585e-06,
+      "loss": 0.7646,
+      "step": 26315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.552179348345398e-06,
+      "loss": 0.9808,
+      "step": 26316
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.551464257016367e-06,
+      "loss": 0.6515,
+      "step": 26317
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.550749205309694e-06,
+      "loss": 0.9283,
+      "step": 26318
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.550034193230576e-06,
+      "loss": 1.0193,
+      "step": 26319
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549319220784217e-06,
+      "loss": 0.7306,
+      "step": 26320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5486042879758115e-06,
+      "loss": 0.9896,
+      "step": 26321
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.547889394810562e-06,
+      "loss": 0.8947,
+      "step": 26322
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5471745412936665e-06,
+      "loss": 0.851,
+      "step": 26323
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.546459727430324e-06,
+      "loss": 0.4755,
+      "step": 26324
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.545744953225726e-06,
+      "loss": 0.8932,
+      "step": 26325
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.545030218685082e-06,
+      "loss": 0.9876,
+      "step": 26326
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544315523813583e-06,
+      "loss": 0.7746,
+      "step": 26327
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.543600868616431e-06,
+      "loss": 1.1453,
+      "step": 26328
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5428862530988176e-06,
+      "loss": 1.1077,
+      "step": 26329
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5421716772659395e-06,
+      "loss": 0.7977,
+      "step": 26330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.541457141122992e-06,
+      "loss": 0.8397,
+      "step": 26331
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540742644675176e-06,
+      "loss": 0.7599,
+      "step": 26332
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540028187927689e-06,
+      "loss": 0.8732,
+      "step": 26333
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.539313770885722e-06,
+      "loss": 0.9408,
+      "step": 26334
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.538599393554471e-06,
+      "loss": 0.8315,
+      "step": 26335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5378850559391316e-06,
+      "loss": 0.8902,
+      "step": 26336
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5371707580449e-06,
+      "loss": 0.595,
+      "step": 26337
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.536456499876969e-06,
+      "loss": 0.9965,
+      "step": 26338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.535742281440532e-06,
+      "loss": 1.0058,
+      "step": 26339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.535028102740785e-06,
+      "loss": 0.7758,
+      "step": 26340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.534313963782919e-06,
+      "loss": 0.8669,
+      "step": 26341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5335998645721305e-06,
+      "loss": 0.9847,
+      "step": 26342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532885805113606e-06,
+      "loss": 0.8891,
+      "step": 26343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532171785412547e-06,
+      "loss": 0.6966,
+      "step": 26344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.531457805474142e-06,
+      "loss": 0.9552,
+      "step": 26345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.530743865303584e-06,
+      "loss": 1.1664,
+      "step": 26346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.530029964906063e-06,
+      "loss": 0.9997,
+      "step": 26347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529316104286775e-06,
+      "loss": 0.7734,
+      "step": 26348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.528602283450901e-06,
+      "loss": 1.1673,
+      "step": 26349
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527888502403642e-06,
+      "loss": 1.0585,
+      "step": 26350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527174761150186e-06,
+      "loss": 0.8351,
+      "step": 26351
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5264610596957224e-06,
+      "loss": 0.8134,
+      "step": 26352
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525747398045442e-06,
+      "loss": 1.2315,
+      "step": 26353
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525033776204535e-06,
+      "loss": 1.0702,
+      "step": 26354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.524320194178189e-06,
+      "loss": 0.7038,
+      "step": 26355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5236066519715915e-06,
+      "loss": 0.8726,
+      "step": 26356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522893149589939e-06,
+      "loss": 0.5761,
+      "step": 26357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522179687038419e-06,
+      "loss": 0.8215,
+      "step": 26358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.521466264322213e-06,
+      "loss": 0.7739,
+      "step": 26359
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520752881446512e-06,
+      "loss": 1.0188,
+      "step": 26360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520039538416502e-06,
+      "loss": 1.0422,
+      "step": 26361
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5193262352373755e-06,
+      "loss": 0.916,
+      "step": 26362
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.518612971914318e-06,
+      "loss": 1.097,
+      "step": 26363
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.517899748452515e-06,
+      "loss": 0.7157,
+      "step": 26364
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.517186564857155e-06,
+      "loss": 0.6604,
+      "step": 26365
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.516473421133421e-06,
+      "loss": 0.981,
+      "step": 26366
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515760317286504e-06,
+      "loss": 0.9339,
+      "step": 26367
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515047253321585e-06,
+      "loss": 0.6501,
+      "step": 26368
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.514334229243852e-06,
+      "loss": 0.5852,
+      "step": 26369
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5136212450584895e-06,
+      "loss": 1.0437,
+      "step": 26370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.512908300770683e-06,
+      "loss": 0.8056,
+      "step": 26371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.512195396385617e-06,
+      "loss": 0.6985,
+      "step": 26372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511482531908474e-06,
+      "loss": 0.781,
+      "step": 26373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.510769707344437e-06,
+      "loss": 0.7998,
+      "step": 26374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.510056922698696e-06,
+      "loss": 0.876,
+      "step": 26375
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.509344177976431e-06,
+      "loss": 0.9872,
+      "step": 26376
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.508631473182823e-06,
+      "loss": 0.5358,
+      "step": 26377
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.507918808323062e-06,
+      "loss": 0.8353,
+      "step": 26378
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.507206183402318e-06,
+      "loss": 1.0746,
+      "step": 26379
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.506493598425783e-06,
+      "loss": 0.5767,
+      "step": 26380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.505781053398638e-06,
+      "loss": 0.9473,
+      "step": 26381
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.505068548326063e-06,
+      "loss": 0.572,
+      "step": 26382
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.504356083213239e-06,
+      "loss": 0.8464,
+      "step": 26383
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503643658065349e-06,
+      "loss": 0.9975,
+      "step": 26384
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502931272887572e-06,
+      "loss": 1.0506,
+      "step": 26385
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502218927685085e-06,
+      "loss": 0.9313,
+      "step": 26386
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.501506622463077e-06,
+      "loss": 1.0677,
+      "step": 26387
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.500794357226727e-06,
+      "loss": 0.9073,
+      "step": 26388
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.500082131981206e-06,
+      "loss": 1.0252,
+      "step": 26389
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4993699467317e-06,
+      "loss": 0.9224,
+      "step": 26390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498657801483386e-06,
+      "loss": 1.1181,
+      "step": 26391
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.49794569624144e-06,
+      "loss": 0.6314,
+      "step": 26392
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4972336310110475e-06,
+      "loss": 1.0653,
+      "step": 26393
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.496521605797384e-06,
+      "loss": 0.7524,
+      "step": 26394
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.495809620605626e-06,
+      "loss": 0.9752,
+      "step": 26395
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.495097675440951e-06,
+      "loss": 0.8598,
+      "step": 26396
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.494385770308537e-06,
+      "loss": 0.872,
+      "step": 26397
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493673905213561e-06,
+      "loss": 0.7142,
+      "step": 26398
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4929620801612e-06,
+      "loss": 0.7055,
+      "step": 26399
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.492250295156632e-06,
+      "loss": 0.8744,
+      "step": 26400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.491538550205029e-06,
+      "loss": 0.7337,
+      "step": 26401
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.49082684531157e-06,
+      "loss": 0.7337,
+      "step": 26402
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.49011518048143e-06,
+      "loss": 0.6629,
+      "step": 26403
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4894035557197815e-06,
+      "loss": 0.9043,
+      "step": 26404
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.488691971031806e-06,
+      "loss": 0.8856,
+      "step": 26405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.487980426422674e-06,
+      "loss": 0.5923,
+      "step": 26406
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.48726892189756e-06,
+      "loss": 0.8422,
+      "step": 26407
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.486557457461642e-06,
+      "loss": 0.7924,
+      "step": 26408
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485846033120087e-06,
+      "loss": 0.6993,
+      "step": 26409
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485134648878069e-06,
+      "loss": 0.8216,
+      "step": 26410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.484423304740768e-06,
+      "loss": 0.6897,
+      "step": 26411
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.483712000713353e-06,
+      "loss": 0.9754,
+      "step": 26412
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.483000736800996e-06,
+      "loss": 0.8402,
+      "step": 26413
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482289513008871e-06,
+      "loss": 0.7058,
+      "step": 26414
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.481578329342149e-06,
+      "loss": 1.0202,
+      "step": 26415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480867185806002e-06,
+      "loss": 0.8369,
+      "step": 26416
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480156082405599e-06,
+      "loss": 0.7157,
+      "step": 26417
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4794450191461214e-06,
+      "loss": 0.8999,
+      "step": 26418
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.478733996032728e-06,
+      "loss": 0.819,
+      "step": 26419
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.478023013070595e-06,
+      "loss": 0.8154,
+      "step": 26420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.477312070264891e-06,
+      "loss": 0.8278,
+      "step": 26421
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.476601167620783e-06,
+      "loss": 0.7676,
+      "step": 26422
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.47589030514345e-06,
+      "loss": 0.6919,
+      "step": 26423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.475179482838054e-06,
+      "loss": 1.0691,
+      "step": 26424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.474468700709768e-06,
+      "loss": 0.6805,
+      "step": 26425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.473757958763759e-06,
+      "loss": 0.6789,
+      "step": 26426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.473047257005195e-06,
+      "loss": 0.7921,
+      "step": 26427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.472336595439246e-06,
+      "loss": 0.9841,
+      "step": 26428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.471625974071079e-06,
+      "loss": 0.7613,
+      "step": 26429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.470915392905862e-06,
+      "loss": 0.8885,
+      "step": 26430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.470204851948761e-06,
+      "loss": 0.742,
+      "step": 26431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.469494351204945e-06,
+      "loss": 0.7608,
+      "step": 26432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46878389067958e-06,
+      "loss": 0.5851,
+      "step": 26433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.468073470377833e-06,
+      "loss": 0.964,
+      "step": 26434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.467363090304866e-06,
+      "loss": 0.9539,
+      "step": 26435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.466652750465853e-06,
+      "loss": 0.9664,
+      "step": 26436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.465942450865957e-06,
+      "loss": 0.9481,
+      "step": 26437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.465232191510344e-06,
+      "loss": 0.8837,
+      "step": 26438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4645219724041735e-06,
+      "loss": 1.1225,
+      "step": 26439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4638117935526105e-06,
+      "loss": 0.9864,
+      "step": 26440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4631016549608275e-06,
+      "loss": 0.7704,
+      "step": 26441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.462391556633982e-06,
+      "loss": 0.8203,
+      "step": 26442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.461681498577242e-06,
+      "loss": 0.9908,
+      "step": 26443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46097148079577e-06,
+      "loss": 0.8272,
+      "step": 26444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460261503294726e-06,
+      "loss": 0.7581,
+      "step": 26445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.459551566079277e-06,
+      "loss": 0.8064,
+      "step": 26446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.45884166915458e-06,
+      "loss": 0.728,
+      "step": 26447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.458131812525809e-06,
+      "loss": 0.8879,
+      "step": 26448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4574219961981155e-06,
+      "loss": 0.6476,
+      "step": 26449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.456712220176665e-06,
+      "loss": 1.039,
+      "step": 26450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.456002484466618e-06,
+      "loss": 1.1221,
+      "step": 26451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.455292789073138e-06,
+      "loss": 1.2104,
+      "step": 26452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.454583134001379e-06,
+      "loss": 0.9123,
+      "step": 26453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453873519256512e-06,
+      "loss": 0.6233,
+      "step": 26454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453163944843693e-06,
+      "loss": 1.0655,
+      "step": 26455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.452454410768082e-06,
+      "loss": 0.9253,
+      "step": 26456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451744917034837e-06,
+      "loss": 0.9232,
+      "step": 26457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451035463649119e-06,
+      "loss": 0.9639,
+      "step": 26458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.450326050616089e-06,
+      "loss": 1.0105,
+      "step": 26459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.449616677940904e-06,
+      "loss": 0.9871,
+      "step": 26460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.448907345628721e-06,
+      "loss": 0.754,
+      "step": 26461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.448198053684702e-06,
+      "loss": 1.061,
+      "step": 26462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.447488802114003e-06,
+      "loss": 1.0623,
+      "step": 26463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.446779590921783e-06,
+      "loss": 0.5926,
+      "step": 26464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.446070420113194e-06,
+      "loss": 0.9534,
+      "step": 26465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.445361289693399e-06,
+      "loss": 0.933,
+      "step": 26466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444652199667556e-06,
+      "loss": 0.7369,
+      "step": 26467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.443943150040823e-06,
+      "loss": 0.9321,
+      "step": 26468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.443234140818348e-06,
+      "loss": 0.6103,
+      "step": 26469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.442525172005291e-06,
+      "loss": 0.7184,
+      "step": 26470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.441816243606804e-06,
+      "loss": 1.1081,
+      "step": 26471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4411073556280515e-06,
+      "loss": 0.8595,
+      "step": 26472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440398508074183e-06,
+      "loss": 0.8827,
+      "step": 26473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.439689700950354e-06,
+      "loss": 0.8664,
+      "step": 26474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4389809342617195e-06,
+      "loss": 0.6767,
+      "step": 26475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.438272208013433e-06,
+      "loss": 1.0028,
+      "step": 26476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.437563522210649e-06,
+      "loss": 0.8239,
+      "step": 26477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.43685487685852e-06,
+      "loss": 0.9998,
+      "step": 26478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.436146271962199e-06,
+      "loss": 1.0297,
+      "step": 26479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.435437707526841e-06,
+      "loss": 0.7012,
+      "step": 26480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.434729183557599e-06,
+      "loss": 0.81,
+      "step": 26481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.434020700059624e-06,
+      "loss": 1.0309,
+      "step": 26482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.433312257038063e-06,
+      "loss": 0.8002,
+      "step": 26483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4326038544980794e-06,
+      "loss": 0.5181,
+      "step": 26484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4318954924448185e-06,
+      "loss": 0.4311,
+      "step": 26485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.431187170883431e-06,
+      "loss": 0.9103,
+      "step": 26486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.43047888981907e-06,
+      "loss": 0.6415,
+      "step": 26487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429770649256887e-06,
+      "loss": 1.0197,
+      "step": 26488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429062449202024e-06,
+      "loss": 0.8542,
+      "step": 26489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.428354289659641e-06,
+      "loss": 0.97,
+      "step": 26490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.427646170634885e-06,
+      "loss": 0.9322,
+      "step": 26491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.426938092132904e-06,
+      "loss": 0.8332,
+      "step": 26492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.426230054158848e-06,
+      "loss": 0.8901,
+      "step": 26493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.425522056717867e-06,
+      "loss": 1.0668,
+      "step": 26494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424814099815108e-06,
+      "loss": 0.9463,
+      "step": 26495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424106183455716e-06,
+      "loss": 0.8894,
+      "step": 26496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.423398307644848e-06,
+      "loss": 0.5909,
+      "step": 26497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.422690472387648e-06,
+      "loss": 0.9187,
+      "step": 26498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42198267768926e-06,
+      "loss": 1.05,
+      "step": 26499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.421274923554835e-06,
+      "loss": 0.8532,
+      "step": 26500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420567209989513e-06,
+      "loss": 0.8327,
+      "step": 26501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.419859536998449e-06,
+      "loss": 0.9321,
+      "step": 26502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.419151904586786e-06,
+      "loss": 0.7211,
+      "step": 26503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.418444312759671e-06,
+      "loss": 0.7552,
+      "step": 26504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.417736761522249e-06,
+      "loss": 0.604,
+      "step": 26505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.417029250879664e-06,
+      "loss": 0.7271,
+      "step": 26506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.416321780837063e-06,
+      "loss": 0.8325,
+      "step": 26507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41561435139959e-06,
+      "loss": 0.7526,
+      "step": 26508
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.41490696257239e-06,
+      "loss": 0.9075,
+      "step": 26509
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.414199614360606e-06,
+      "loss": 0.95,
+      "step": 26510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.413492306769382e-06,
+      "loss": 0.7818,
+      "step": 26511
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.412785039803864e-06,
+      "loss": 0.4547,
+      "step": 26512
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.412077813469193e-06,
+      "loss": 1.3262,
+      "step": 26513
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4113706277705095e-06,
+      "loss": 0.8138,
+      "step": 26514
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4106634827129615e-06,
+      "loss": 1.0459,
+      "step": 26515
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.409956378301691e-06,
+      "loss": 0.9021,
+      "step": 26516
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.409249314541838e-06,
+      "loss": 0.8912,
+      "step": 26517
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4085422914385475e-06,
+      "loss": 0.7111,
+      "step": 26518
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.407835308996952e-06,
+      "loss": 0.8205,
+      "step": 26519
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.407128367222203e-06,
+      "loss": 0.7855,
+      "step": 26520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.406421466119437e-06,
+      "loss": 0.7919,
+      "step": 26521
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.405714605693795e-06,
+      "loss": 0.4172,
+      "step": 26522
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.405007785950418e-06,
+      "loss": 0.8373,
+      "step": 26523
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.404301006894446e-06,
+      "loss": 0.9121,
+      "step": 26524
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.403594268531018e-06,
+      "loss": 0.9504,
+      "step": 26525
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402887570865271e-06,
+      "loss": 0.7859,
+      "step": 26526
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402180913902351e-06,
+      "loss": 0.9518,
+      "step": 26527
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.401474297647395e-06,
+      "loss": 0.7255,
+      "step": 26528
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.400767722105538e-06,
+      "loss": 0.6789,
+      "step": 26529
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.400061187281919e-06,
+      "loss": 0.6898,
+      "step": 26530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3993546931816755e-06,
+      "loss": 1.1093,
+      "step": 26531
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.398648239809945e-06,
+      "loss": 1.0022,
+      "step": 26532
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.397941827171869e-06,
+      "loss": 0.9848,
+      "step": 26533
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.397235455272582e-06,
+      "loss": 0.9373,
+      "step": 26534
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.396529124117223e-06,
+      "loss": 0.8686,
+      "step": 26535
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3958228337109245e-06,
+      "loss": 0.798,
+      "step": 26536
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.395116584058825e-06,
+      "loss": 0.6704,
+      "step": 26537
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.39441037516606e-06,
+      "loss": 0.8782,
+      "step": 26538
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3937042070377655e-06,
+      "loss": 0.6677,
+      "step": 26539
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.392998079679076e-06,
+      "loss": 1.0253,
+      "step": 26540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.392291993095128e-06,
+      "loss": 0.7154,
+      "step": 26541
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.391585947291055e-06,
+      "loss": 1.1737,
+      "step": 26542
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.390879942271992e-06,
+      "loss": 0.6244,
+      "step": 26543
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.390173978043068e-06,
+      "loss": 1.0528,
+      "step": 26544
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.389468054609427e-06,
+      "loss": 0.9328,
+      "step": 26545
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3887621719761965e-06,
+      "loss": 0.4987,
+      "step": 26546
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.388056330148511e-06,
+      "loss": 0.9035,
+      "step": 26547
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.387350529131506e-06,
+      "loss": 0.8369,
+      "step": 26548
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.386644768930309e-06,
+      "loss": 0.7541,
+      "step": 26549
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.38593904955005e-06,
+      "loss": 0.7831,
+      "step": 26550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.385233370995869e-06,
+      "loss": 1.082,
+      "step": 26551
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.384527733272895e-06,
+      "loss": 0.8928,
+      "step": 26552
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.38382213638626e-06,
+      "loss": 0.9365,
+      "step": 26553
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3831165803410926e-06,
+      "loss": 0.8234,
+      "step": 26554
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3824110651425255e-06,
+      "loss": 0.9128,
+      "step": 26555
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.381705590795688e-06,
+      "loss": 0.8138,
+      "step": 26556
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.381000157305709e-06,
+      "loss": 1.0495,
+      "step": 26557
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.380294764677728e-06,
+      "loss": 0.7146,
+      "step": 26558
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.379589412916864e-06,
+      "loss": 1.0584,
+      "step": 26559
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.378884102028249e-06,
+      "loss": 1.0269,
+      "step": 26560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.378178832017014e-06,
+      "loss": 0.6565,
+      "step": 26561
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.377473602888282e-06,
+      "loss": 0.9406,
+      "step": 26562
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.376768414647191e-06,
+      "loss": 0.6782,
+      "step": 26563
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.376063267298865e-06,
+      "loss": 0.8164,
+      "step": 26564
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.37535816084843e-06,
+      "loss": 1.1154,
+      "step": 26565
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.374653095301016e-06,
+      "loss": 0.5298,
+      "step": 26566
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3739480706617486e-06,
+      "loss": 0.8851,
+      "step": 26567
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.373243086935756e-06,
+      "loss": 0.8451,
+      "step": 26568
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.372538144128164e-06,
+      "loss": 0.5673,
+      "step": 26569
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.371833242244101e-06,
+      "loss": 0.9634,
+      "step": 26570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.371128381288691e-06,
+      "loss": 0.5875,
+      "step": 26571
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.37042356126706e-06,
+      "loss": 0.6705,
+      "step": 26572
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.369718782184334e-06,
+      "loss": 0.8616,
+      "step": 26573
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.369014044045639e-06,
+      "loss": 0.895,
+      "step": 26574
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3683093468560945e-06,
+      "loss": 0.6267,
+      "step": 26575
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367604690620834e-06,
+      "loss": 1.1439,
+      "step": 26576
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.366900075344978e-06,
+      "loss": 0.9279,
+      "step": 26577
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.366195501033653e-06,
+      "loss": 0.8641,
+      "step": 26578
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.365490967691977e-06,
+      "loss": 0.7989,
+      "step": 26579
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.364786475325072e-06,
+      "loss": 0.7474,
+      "step": 26580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3640820239380695e-06,
+      "loss": 0.3799,
+      "step": 26581
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.363377613536089e-06,
+      "loss": 0.88,
+      "step": 26582
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.362673244124253e-06,
+      "loss": 0.8666,
+      "step": 26583
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3619689157076815e-06,
+      "loss": 1.0223,
+      "step": 26584
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.361264628291498e-06,
+      "loss": 0.6516,
+      "step": 26585
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3605603818808266e-06,
+      "loss": 0.9018,
+      "step": 26586
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.35985617648078e-06,
+      "loss": 0.6337,
+      "step": 26587
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3591520120964945e-06,
+      "loss": 0.9029,
+      "step": 26588
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.358447888733078e-06,
+      "loss": 0.6055,
+      "step": 26589
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.357743806395655e-06,
+      "loss": 0.9677,
+      "step": 26590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.357039765089345e-06,
+      "loss": 1.0225,
+      "step": 26591
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.35633576481927e-06,
+      "loss": 0.8128,
+      "step": 26592
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.355631805590542e-06,
+      "loss": 1.073,
+      "step": 26593
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.354927887408292e-06,
+      "loss": 0.6908,
+      "step": 26594
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.354224010277632e-06,
+      "loss": 0.6763,
+      "step": 26595
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.353520174203682e-06,
+      "loss": 0.7843,
+      "step": 26596
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.352816379191561e-06,
+      "loss": 0.6735,
+      "step": 26597
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.352112625246386e-06,
+      "loss": 0.7583,
+      "step": 26598
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.351408912373276e-06,
+      "loss": 0.5969,
+      "step": 26599
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.350705240577348e-06,
+      "loss": 0.7623,
+      "step": 26600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.350001609863716e-06,
+      "loss": 0.9485,
+      "step": 26601
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.349298020237502e-06,
+      "loss": 1.1624,
+      "step": 26602
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.348594471703821e-06,
+      "loss": 0.7379,
+      "step": 26603
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347890964267787e-06,
+      "loss": 0.9444,
+      "step": 26604
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347187497934514e-06,
+      "loss": 1.0582,
+      "step": 26605
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3464840727091254e-06,
+      "loss": 0.6527,
+      "step": 26606
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.345780688596732e-06,
+      "loss": 0.821,
+      "step": 26607
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3450773456024535e-06,
+      "loss": 0.5777,
+      "step": 26608
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.344374043731396e-06,
+      "loss": 0.7497,
+      "step": 26609
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.343670782988679e-06,
+      "loss": 1.039,
+      "step": 26610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.342967563379413e-06,
+      "loss": 0.7783,
+      "step": 26611
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3422643849087184e-06,
+      "loss": 1.2059,
+      "step": 26612
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.341561247581706e-06,
+      "loss": 0.8405,
+      "step": 26613
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.340858151403487e-06,
+      "loss": 0.9629,
+      "step": 26614
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.340155096379177e-06,
+      "loss": 1.0281,
+      "step": 26615
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.339452082513887e-06,
+      "loss": 0.7496,
+      "step": 26616
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.338749109812732e-06,
+      "loss": 0.7877,
+      "step": 26617
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3380461782808205e-06,
+      "loss": 0.6392,
+      "step": 26618
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.337343287923266e-06,
+      "loss": 0.7972,
+      "step": 26619
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.33664043874518e-06,
+      "loss": 0.8749,
+      "step": 26620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.335937630751675e-06,
+      "loss": 0.7209,
+      "step": 26621
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.335234863947858e-06,
+      "loss": 0.6868,
+      "step": 26622
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3345321383388395e-06,
+      "loss": 0.8776,
+      "step": 26623
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.333829453929736e-06,
+      "loss": 1.0424,
+      "step": 26624
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.333126810725655e-06,
+      "loss": 0.9331,
+      "step": 26625
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.332424208731705e-06,
+      "loss": 0.6257,
+      "step": 26626
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.331721647952994e-06,
+      "loss": 1.0277,
+      "step": 26627
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.331019128394638e-06,
+      "loss": 1.1246,
+      "step": 26628
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3303166500617325e-06,
+      "loss": 0.8679,
+      "step": 26629
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3296142129593964e-06,
+      "loss": 0.5728,
+      "step": 26630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.328911817092737e-06,
+      "loss": 0.6384,
+      "step": 26631
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.32820946246686e-06,
+      "loss": 0.7559,
+      "step": 26632
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3275071490868735e-06,
+      "loss": 0.5541,
+      "step": 26633
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.326804876957885e-06,
+      "loss": 0.9203,
+      "step": 26634
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.326102646085002e-06,
+      "loss": 0.7976,
+      "step": 26635
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.325400456473326e-06,
+      "loss": 0.7581,
+      "step": 26636
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.324698308127973e-06,
+      "loss": 0.9386,
+      "step": 26637
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.323996201054046e-06,
+      "loss": 0.6481,
+      "step": 26638
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.323294135256646e-06,
+      "loss": 0.8604,
+      "step": 26639
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.322592110740882e-06,
+      "loss": 0.6796,
+      "step": 26640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321890127511856e-06,
+      "loss": 0.9598,
+      "step": 26641
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321188185574676e-06,
+      "loss": 0.9182,
+      "step": 26642
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320486284934449e-06,
+      "loss": 0.8597,
+      "step": 26643
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.319784425596275e-06,
+      "loss": 0.8588,
+      "step": 26644
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.319082607565261e-06,
+      "loss": 0.7945,
+      "step": 26645
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.318380830846507e-06,
+      "loss": 0.9027,
+      "step": 26646
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3176790954451195e-06,
+      "loss": 0.8972,
+      "step": 26647
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3169774013662e-06,
+      "loss": 1.0343,
+      "step": 26648
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.316275748614853e-06,
+      "loss": 1.1893,
+      "step": 26649
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.315574137196179e-06,
+      "loss": 1.0295,
+      "step": 26650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.314872567115282e-06,
+      "loss": 0.7049,
+      "step": 26651
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.314171038377263e-06,
+      "loss": 0.7729,
+      "step": 26652
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.313469550987223e-06,
+      "loss": 0.6798,
+      "step": 26653
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.31276810495026e-06,
+      "loss": 0.7651,
+      "step": 26654
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.312066700271483e-06,
+      "loss": 0.6751,
+      "step": 26655
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.311365336955988e-06,
+      "loss": 0.894,
+      "step": 26656
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.310664015008877e-06,
+      "loss": 0.5801,
+      "step": 26657
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.309962734435252e-06,
+      "loss": 0.7178,
+      "step": 26658
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3092614952402035e-06,
+      "loss": 0.5583,
+      "step": 26659
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.30856029742884e-06,
+      "loss": 1.0137,
+      "step": 26660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3078591410062585e-06,
+      "loss": 0.9706,
+      "step": 26661
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.307158025977557e-06,
+      "loss": 0.9474,
+      "step": 26662
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.306456952347835e-06,
+      "loss": 1.0534,
+      "step": 26663
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.30575592012219e-06,
+      "loss": 0.6815,
+      "step": 26664
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.305054929305721e-06,
+      "loss": 0.6467,
+      "step": 26665
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3043539799035206e-06,
+      "loss": 0.8379,
+      "step": 26666
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.303653071920694e-06,
+      "loss": 0.8545,
+      "step": 26667
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.302952205362338e-06,
+      "loss": 0.772,
+      "step": 26668
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3022513802335445e-06,
+      "loss": 0.9371,
+      "step": 26669
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.30155059653941e-06,
+      "loss": 0.8977,
+      "step": 26670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.300849854285033e-06,
+      "loss": 1.0293,
+      "step": 26671
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.300149153475504e-06,
+      "loss": 0.9087,
+      "step": 26672
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.299448494115929e-06,
+      "loss": 1.0052,
+      "step": 26673
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.298747876211395e-06,
+      "loss": 0.7044,
+      "step": 26674
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.298047299767002e-06,
+      "loss": 1.1521,
+      "step": 26675
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.297346764787841e-06,
+      "loss": 0.9193,
+      "step": 26676
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.296646271279008e-06,
+      "loss": 1.0542,
+      "step": 26677
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2959458192455965e-06,
+      "loss": 0.9156,
+      "step": 26678
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2952454086927005e-06,
+      "loss": 0.8287,
+      "step": 26679
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.294545039625413e-06,
+      "loss": 0.7613,
+      "step": 26680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.293844712048828e-06,
+      "loss": 0.9167,
+      "step": 26681
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.293144425968038e-06,
+      "loss": 0.737,
+      "step": 26682
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.292444181388136e-06,
+      "loss": 1.118,
+      "step": 26683
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2917439783142086e-06,
+      "loss": 0.8182,
+      "step": 26684
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.291043816751357e-06,
+      "loss": 0.8479,
+      "step": 26685
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.290343696704668e-06,
+      "loss": 0.7474,
+      "step": 26686
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.289643618179236e-06,
+      "loss": 1.0502,
+      "step": 26687
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.288943581180151e-06,
+      "loss": 0.9368,
+      "step": 26688
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2882435857125e-06,
+      "loss": 0.8431,
+      "step": 26689
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.287543631781372e-06,
+      "loss": 0.7249,
+      "step": 26690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2868437193918655e-06,
+      "loss": 1.06,
+      "step": 26691
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.286143848549065e-06,
+      "loss": 0.6135,
+      "step": 26692
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2854440192580615e-06,
+      "loss": 0.7235,
+      "step": 26693
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.284744231523943e-06,
+      "loss": 0.6846,
+      "step": 26694
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.284044485351799e-06,
+      "loss": 1.0192,
+      "step": 26695
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.283344780746718e-06,
+      "loss": 0.9682,
+      "step": 26696
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.282645117713787e-06,
+      "loss": 0.8995,
+      "step": 26697
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.281945496258102e-06,
+      "loss": 0.911,
+      "step": 26698
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.28124591638474e-06,
+      "loss": 0.923,
+      "step": 26699
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.280546378098792e-06,
+      "loss": 0.637,
+      "step": 26700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.279846881405347e-06,
+      "loss": 0.6377,
+      "step": 26701
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2791474263094865e-06,
+      "loss": 0.8634,
+      "step": 26702
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2784480128163065e-06,
+      "loss": 0.6452,
+      "step": 26703
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.277748640930887e-06,
+      "loss": 0.9076,
+      "step": 26704
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.277049310658313e-06,
+      "loss": 0.7505,
+      "step": 26705
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.276350022003673e-06,
+      "loss": 0.9198,
+      "step": 26706
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.275650774972052e-06,
+      "loss": 0.8191,
+      "step": 26707
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.274951569568534e-06,
+      "loss": 0.7613,
+      "step": 26708
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.274252405798204e-06,
+      "loss": 1.3006,
+      "step": 26709
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.273553283666147e-06,
+      "loss": 0.5933,
+      "step": 26710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272854203177444e-06,
+      "loss": 1.0019,
+      "step": 26711
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272155164337184e-06,
+      "loss": 0.642,
+      "step": 26712
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.271456167150446e-06,
+      "loss": 0.8406,
+      "step": 26713
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.270757211622315e-06,
+      "loss": 0.5398,
+      "step": 26714
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.270058297757871e-06,
+      "loss": 0.6494,
+      "step": 26715
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.269359425562203e-06,
+      "loss": 0.8658,
+      "step": 26716
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.268660595040388e-06,
+      "loss": 1.1369,
+      "step": 26717
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.267961806197515e-06,
+      "loss": 0.5086,
+      "step": 26718
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.267263059038656e-06,
+      "loss": 0.7647,
+      "step": 26719
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2665643535688924e-06,
+      "loss": 0.6541,
+      "step": 26720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.265865689793315e-06,
+      "loss": 0.8359,
+      "step": 26721
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.265167067716998e-06,
+      "loss": 0.8139,
+      "step": 26722
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.264468487345021e-06,
+      "loss": 0.7305,
+      "step": 26723
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.263769948682469e-06,
+      "loss": 0.8429,
+      "step": 26724
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.263071451734417e-06,
+      "loss": 0.7754,
+      "step": 26725
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.262372996505948e-06,
+      "loss": 0.8586,
+      "step": 26726
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.261674583002139e-06,
+      "loss": 0.6616,
+      "step": 26727
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.260976211228069e-06,
+      "loss": 1.113,
+      "step": 26728
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.260277881188818e-06,
+      "loss": 0.8976,
+      "step": 26729
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.259579592889464e-06,
+      "loss": 0.6911,
+      "step": 26730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.258881346335084e-06,
+      "loss": 0.9659,
+      "step": 26731
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.258183141530756e-06,
+      "loss": 0.9967,
+      "step": 26732
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.257484978481555e-06,
+      "loss": 0.825,
+      "step": 26733
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.256786857192564e-06,
+      "loss": 0.5587,
+      "step": 26734
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.256088777668858e-06,
+      "loss": 0.8222,
+      "step": 26735
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.25539073991551e-06,
+      "loss": 0.9129,
+      "step": 26736
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.254692743937601e-06,
+      "loss": 0.7337,
+      "step": 26737
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.253994789740206e-06,
+      "loss": 0.7228,
+      "step": 26738
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.253296877328392e-06,
+      "loss": 0.7849,
+      "step": 26739
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.252599006707245e-06,
+      "loss": 0.7293,
+      "step": 26740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251901177881838e-06,
+      "loss": 0.6216,
+      "step": 26741
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251203390857243e-06,
+      "loss": 0.8788,
+      "step": 26742
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.250505645638534e-06,
+      "loss": 1.2029,
+      "step": 26743
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249807942230787e-06,
+      "loss": 0.9943,
+      "step": 26744
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249110280639076e-06,
+      "loss": 0.6712,
+      "step": 26745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.248412660868469e-06,
+      "loss": 0.8194,
+      "step": 26746
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247715082924048e-06,
+      "loss": 1.1495,
+      "step": 26747
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247017546810885e-06,
+      "loss": 0.9,
+      "step": 26748
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2463200525340464e-06,
+      "loss": 0.8674,
+      "step": 26749
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.245622600098607e-06,
+      "loss": 0.7467,
+      "step": 26750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.244925189509636e-06,
+      "loss": 0.7124,
+      "step": 26751
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.24422782077221e-06,
+      "loss": 0.886,
+      "step": 26752
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.243530493891399e-06,
+      "loss": 0.7994,
+      "step": 26753
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.242833208872275e-06,
+      "loss": 0.7934,
+      "step": 26754
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2421359657199055e-06,
+      "loss": 0.9895,
+      "step": 26755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.241438764439364e-06,
+      "loss": 1.0445,
+      "step": 26756
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240741605035718e-06,
+      "loss": 0.9071,
+      "step": 26757
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.240044487514039e-06,
+      "loss": 1.0064,
+      "step": 26758
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.239347411879397e-06,
+      "loss": 1.0657,
+      "step": 26759
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.238650378136859e-06,
+      "loss": 0.9192,
+      "step": 26760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.237953386291495e-06,
+      "loss": 1.0313,
+      "step": 26761
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.237256436348376e-06,
+      "loss": 0.7888,
+      "step": 26762
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.236559528312567e-06,
+      "loss": 1.0777,
+      "step": 26763
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.235862662189133e-06,
+      "loss": 0.725,
+      "step": 26764
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.23516583798315e-06,
+      "loss": 0.9147,
+      "step": 26765
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2344690556996805e-06,
+      "loss": 0.7243,
+      "step": 26766
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2337723153437925e-06,
+      "loss": 0.6934,
+      "step": 26767
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.233075616920557e-06,
+      "loss": 0.7815,
+      "step": 26768
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.232378960435029e-06,
+      "loss": 1.0412,
+      "step": 26769
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.231682345892284e-06,
+      "loss": 0.6843,
+      "step": 26770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.230985773297385e-06,
+      "loss": 0.8877,
+      "step": 26771
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2302892426554e-06,
+      "loss": 0.6686,
+      "step": 26772
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229592753971391e-06,
+      "loss": 0.819,
+      "step": 26773
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.228896307250425e-06,
+      "loss": 0.8936,
+      "step": 26774
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.228199902497565e-06,
+      "loss": 0.9876,
+      "step": 26775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.227503539717873e-06,
+      "loss": 0.7826,
+      "step": 26776
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2268072189164215e-06,
+      "loss": 1.0183,
+      "step": 26777
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.226110940098271e-06,
+      "loss": 0.6877,
+      "step": 26778
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.225414703268479e-06,
+      "loss": 1.0555,
+      "step": 26779
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.224718508432113e-06,
+      "loss": 0.7799,
+      "step": 26780
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.224022355594236e-06,
+      "loss": 1.0085,
+      "step": 26781
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.223326244759905e-06,
+      "loss": 0.985,
+      "step": 26782
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.222630175934192e-06,
+      "loss": 0.8661,
+      "step": 26783
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2219341491221534e-06,
+      "loss": 0.5067,
+      "step": 26784
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2212381643288515e-06,
+      "loss": 1.1144,
+      "step": 26785
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.220542221559346e-06,
+      "loss": 0.9076,
+      "step": 26786
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2198463208187015e-06,
+      "loss": 0.9038,
+      "step": 26787
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.219150462111975e-06,
+      "loss": 0.8122,
+      "step": 26788
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.218454645444227e-06,
+      "loss": 0.9063,
+      "step": 26789
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.217758870820522e-06,
+      "loss": 1.2425,
+      "step": 26790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2170631382459146e-06,
+      "loss": 0.9153,
+      "step": 26791
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.216367447725467e-06,
+      "loss": 0.7648,
+      "step": 26792
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.215671799264237e-06,
+      "loss": 0.7646,
+      "step": 26793
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.214976192867281e-06,
+      "loss": 0.8553,
+      "step": 26794
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.214280628539663e-06,
+      "loss": 0.9913,
+      "step": 26795
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.21358510628644e-06,
+      "loss": 0.7302,
+      "step": 26796
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2128896261126674e-06,
+      "loss": 0.836,
+      "step": 26797
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.212194188023408e-06,
+      "loss": 1.0507,
+      "step": 26798
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.211498792023713e-06,
+      "loss": 0.9372,
+      "step": 26799
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210803438118637e-06,
+      "loss": 0.6623,
+      "step": 26800
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.210108126313245e-06,
+      "loss": 1.0416,
+      "step": 26801
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.209412856612589e-06,
+      "loss": 0.6015,
+      "step": 26802
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.208717629021727e-06,
+      "loss": 0.5019,
+      "step": 26803
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2080224435457115e-06,
+      "loss": 0.8597,
+      "step": 26804
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.207327300189602e-06,
+      "loss": 1.2905,
+      "step": 26805
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2066321989584515e-06,
+      "loss": 0.4202,
+      "step": 26806
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20593713985731e-06,
+      "loss": 0.8797,
+      "step": 26807
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.205242122891246e-06,
+      "loss": 0.8727,
+      "step": 26808
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.204547148065299e-06,
+      "loss": 1.1214,
+      "step": 26809
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.203852215384531e-06,
+      "loss": 0.7942,
+      "step": 26810
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.203157324853992e-06,
+      "loss": 0.9451,
+      "step": 26811
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.202462476478733e-06,
+      "loss": 0.9718,
+      "step": 26812
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.201767670263814e-06,
+      "loss": 0.8238,
+      "step": 26813
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.201072906214285e-06,
+      "loss": 1.0816,
+      "step": 26814
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.200378184335197e-06,
+      "loss": 0.8006,
+      "step": 26815
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.199683504631605e-06,
+      "loss": 0.9031,
+      "step": 26816
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198988867108557e-06,
+      "loss": 1.0095,
+      "step": 26817
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198294271771106e-06,
+      "loss": 0.8656,
+      "step": 26818
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.197599718624304e-06,
+      "loss": 0.6568,
+      "step": 26819
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.196905207673201e-06,
+      "loss": 1.026,
+      "step": 26820
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.196210738922848e-06,
+      "loss": 0.7502,
+      "step": 26821
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.195516312378294e-06,
+      "loss": 1.0136,
+      "step": 26822
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.194821928044591e-06,
+      "loss": 0.8934,
+      "step": 26823
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.194127585926787e-06,
+      "loss": 0.725,
+      "step": 26824
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.193433286029928e-06,
+      "loss": 0.9646,
+      "step": 26825
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.192739028359072e-06,
+      "loss": 0.9089,
+      "step": 26826
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.192044812919261e-06,
+      "loss": 0.8735,
+      "step": 26827
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.19135063971555e-06,
+      "loss": 0.6543,
+      "step": 26828
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.190656508752978e-06,
+      "loss": 0.8604,
+      "step": 26829
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.189962420036593e-06,
+      "loss": 0.9294,
+      "step": 26830
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.189268373571451e-06,
+      "loss": 0.8781,
+      "step": 26831
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.188574369362593e-06,
+      "loss": 0.9854,
+      "step": 26832
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.187880407415069e-06,
+      "loss": 0.6505,
+      "step": 26833
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1871864877339245e-06,
+      "loss": 0.5965,
+      "step": 26834
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.186492610324204e-06,
+      "loss": 1.0107,
+      "step": 26835
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185798775190956e-06,
+      "loss": 0.6892,
+      "step": 26836
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185104982339221e-06,
+      "loss": 0.8687,
+      "step": 26837
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.184411231774055e-06,
+      "loss": 0.8015,
+      "step": 26838
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.183717523500493e-06,
+      "loss": 0.8208,
+      "step": 26839
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1830238575235825e-06,
+      "loss": 1.01,
+      "step": 26840
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.182330233848369e-06,
+      "loss": 0.8581,
+      "step": 26841
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.181636652479897e-06,
+      "loss": 0.8549,
+      "step": 26842
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.180943113423205e-06,
+      "loss": 0.9107,
+      "step": 26843
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.180249616683344e-06,
+      "loss": 0.9692,
+      "step": 26844
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.179556162265353e-06,
+      "loss": 0.5598,
+      "step": 26845
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.178862750174278e-06,
+      "loss": 0.8981,
+      "step": 26846
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.178169380415159e-06,
+      "loss": 0.8281,
+      "step": 26847
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.177476052993039e-06,
+      "loss": 0.8451,
+      "step": 26848
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1767827679129595e-06,
+      "loss": 1.0461,
+      "step": 26849
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.176089525179961e-06,
+      "loss": 0.7488,
+      "step": 26850
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.175396324799088e-06,
+      "loss": 0.7506,
+      "step": 26851
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.174703166775379e-06,
+      "loss": 1.0297,
+      "step": 26852
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.174010051113876e-06,
+      "loss": 0.9177,
+      "step": 26853
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.173316977819618e-06,
+      "loss": 0.6135,
+      "step": 26854
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.172623946897642e-06,
+      "loss": 0.8977,
+      "step": 26855
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171930958352997e-06,
+      "loss": 0.5881,
+      "step": 26856
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171238012190717e-06,
+      "loss": 0.8,
+      "step": 26857
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.170545108415845e-06,
+      "loss": 0.6742,
+      "step": 26858
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.169852247033411e-06,
+      "loss": 0.7221,
+      "step": 26859
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.16915942804846e-06,
+      "loss": 0.5729,
+      "step": 26860
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.168466651466027e-06,
+      "loss": 0.8308,
+      "step": 26861
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1677739172911535e-06,
+      "loss": 0.8173,
+      "step": 26862
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.167081225528876e-06,
+      "loss": 0.7645,
+      "step": 26863
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.166388576184233e-06,
+      "loss": 0.7598,
+      "step": 26864
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.165695969262259e-06,
+      "loss": 0.7521,
+      "step": 26865
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.165003404767992e-06,
+      "loss": 0.7935,
+      "step": 26866
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.164310882706468e-06,
+      "loss": 0.6747,
+      "step": 26867
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.163618403082723e-06,
+      "loss": 0.659,
+      "step": 26868
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.162925965901794e-06,
+      "loss": 0.709,
+      "step": 26869
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.162233571168716e-06,
+      "loss": 0.7309,
+      "step": 26870
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.161541218888522e-06,
+      "loss": 0.8257,
+      "step": 26871
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160848909066249e-06,
+      "loss": 0.9654,
+      "step": 26872
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160156641706928e-06,
+      "loss": 1.2123,
+      "step": 26873
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.159464416815601e-06,
+      "loss": 0.8112,
+      "step": 26874
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1587722343972955e-06,
+      "loss": 1.1129,
+      "step": 26875
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.158080094457049e-06,
+      "loss": 0.6482,
+      "step": 26876
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.157387996999891e-06,
+      "loss": 0.8799,
+      "step": 26877
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.15669594203086e-06,
+      "loss": 1.2546,
+      "step": 26878
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.156003929554979e-06,
+      "loss": 0.6195,
+      "step": 26879
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.155311959577289e-06,
+      "loss": 0.7719,
+      "step": 26880
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1546200321028195e-06,
+      "loss": 0.9232,
+      "step": 26881
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.153928147136603e-06,
+      "loss": 0.7866,
+      "step": 26882
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.15323630468367e-06,
+      "loss": 1.0286,
+      "step": 26883
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.152544504749051e-06,
+      "loss": 0.8019,
+      "step": 26884
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.151852747337779e-06,
+      "loss": 0.9504,
+      "step": 26885
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.151161032454878e-06,
+      "loss": 0.9988,
+      "step": 26886
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.150469360105388e-06,
+      "loss": 0.9849,
+      "step": 26887
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.149777730294339e-06,
+      "loss": 0.6277,
+      "step": 26888
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.14908614302675e-06,
+      "loss": 0.9077,
+      "step": 26889
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.148394598307658e-06,
+      "loss": 1.0655,
+      "step": 26890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1477030961420876e-06,
+      "loss": 0.9337,
+      "step": 26891
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.147011636535072e-06,
+      "loss": 0.8895,
+      "step": 26892
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.146320219491638e-06,
+      "loss": 0.7764,
+      "step": 26893
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.145628845016814e-06,
+      "loss": 0.722,
+      "step": 26894
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.144937513115627e-06,
+      "loss": 0.859,
+      "step": 26895
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.144246223793104e-06,
+      "loss": 0.6197,
+      "step": 26896
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.143554977054273e-06,
+      "loss": 0.5562,
+      "step": 26897
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.142863772904161e-06,
+      "loss": 0.8108,
+      "step": 26898
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.142172611347793e-06,
+      "loss": 0.8247,
+      "step": 26899
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.141481492390197e-06,
+      "loss": 0.837,
+      "step": 26900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140790416036399e-06,
+      "loss": 0.7768,
+      "step": 26901
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140099382291423e-06,
+      "loss": 0.8753,
+      "step": 26902
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.139408391160294e-06,
+      "loss": 0.7823,
+      "step": 26903
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1387174426480355e-06,
+      "loss": 0.8281,
+      "step": 26904
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.138026536759678e-06,
+      "loss": 0.9078,
+      "step": 26905
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.137335673500243e-06,
+      "loss": 0.9243,
+      "step": 26906
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.136644852874754e-06,
+      "loss": 0.6826,
+      "step": 26907
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.135954074888238e-06,
+      "loss": 0.495,
+      "step": 26908
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.135263339545708e-06,
+      "loss": 0.8513,
+      "step": 26909
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134572646852196e-06,
+      "loss": 0.5374,
+      "step": 26910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.133881996812725e-06,
+      "loss": 0.6435,
+      "step": 26911
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.133191389432316e-06,
+      "loss": 1.1245,
+      "step": 26912
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1325008247159904e-06,
+      "loss": 0.9704,
+      "step": 26913
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.131810302668769e-06,
+      "loss": 0.8967,
+      "step": 26914
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.131119823295676e-06,
+      "loss": 0.9005,
+      "step": 26915
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.130429386601728e-06,
+      "loss": 0.7147,
+      "step": 26916
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.129738992591952e-06,
+      "loss": 0.6079,
+      "step": 26917
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.129048641271369e-06,
+      "loss": 0.8743,
+      "step": 26918
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.128358332644994e-06,
+      "loss": 0.6479,
+      "step": 26919
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.127668066717849e-06,
+      "loss": 0.9514,
+      "step": 26920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.126977843494953e-06,
+      "loss": 0.8455,
+      "step": 26921
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.126287662981324e-06,
+      "loss": 0.8185,
+      "step": 26922
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.125597525181987e-06,
+      "loss": 0.992,
+      "step": 26923
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.124907430101957e-06,
+      "loss": 0.7085,
+      "step": 26924
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.124217377746251e-06,
+      "loss": 1.0471,
+      "step": 26925
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12352736811989e-06,
+      "loss": 0.684,
+      "step": 26926
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12283740122789e-06,
+      "loss": 0.8497,
+      "step": 26927
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12214747707527e-06,
+      "loss": 0.87,
+      "step": 26928
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.121457595667046e-06,
+      "loss": 1.0274,
+      "step": 26929
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.120767757008234e-06,
+      "loss": 0.8155,
+      "step": 26930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.120077961103854e-06,
+      "loss": 0.83,
+      "step": 26931
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1193882079589175e-06,
+      "loss": 0.8877,
+      "step": 26932
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.118698497578444e-06,
+      "loss": 0.9108,
+      "step": 26933
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1180088299674446e-06,
+      "loss": 0.966,
+      "step": 26934
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.117319205130942e-06,
+      "loss": 0.847,
+      "step": 26935
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.116629623073947e-06,
+      "loss": 0.7171,
+      "step": 26936
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.115940083801478e-06,
+      "loss": 0.7332,
+      "step": 26937
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1152505873185435e-06,
+      "loss": 0.6692,
+      "step": 26938
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.114561133630159e-06,
+      "loss": 0.8604,
+      "step": 26939
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.113871722741337e-06,
+      "loss": 0.9176,
+      "step": 26940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.113182354657097e-06,
+      "loss": 0.8248,
+      "step": 26941
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.112493029382448e-06,
+      "loss": 0.8485,
+      "step": 26942
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.111803746922404e-06,
+      "loss": 1.0636,
+      "step": 26943
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.111114507281977e-06,
+      "loss": 0.8673,
+      "step": 26944
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.110425310466179e-06,
+      "loss": 1.1051,
+      "step": 26945
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.109736156480022e-06,
+      "loss": 0.7392,
+      "step": 26946
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.109047045328518e-06,
+      "loss": 0.9511,
+      "step": 26947
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.108357977016678e-06,
+      "loss": 0.7627,
+      "step": 26948
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.107668951549514e-06,
+      "loss": 1.0854,
+      "step": 26949
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.106979968932034e-06,
+      "loss": 0.7242,
+      "step": 26950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.106291029169251e-06,
+      "loss": 0.5884,
+      "step": 26951
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.105602132266171e-06,
+      "loss": 0.8992,
+      "step": 26952
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.10491327822781e-06,
+      "loss": 0.9171,
+      "step": 26953
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.104224467059173e-06,
+      "loss": 0.8274,
+      "step": 26954
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.103535698765272e-06,
+      "loss": 1.2264,
+      "step": 26955
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.102846973351115e-06,
+      "loss": 0.8943,
+      "step": 26956
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1021582908217105e-06,
+      "loss": 0.9719,
+      "step": 26957
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.101469651182061e-06,
+      "loss": 0.7592,
+      "step": 26958
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1007810544371815e-06,
+      "loss": 0.9956,
+      "step": 26959
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.100092500592078e-06,
+      "loss": 1.0719,
+      "step": 26960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.099403989651757e-06,
+      "loss": 0.6526,
+      "step": 26961
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.098715521621225e-06,
+      "loss": 1.0469,
+      "step": 26962
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.098027096505489e-06,
+      "loss": 1.2387,
+      "step": 26963
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.097338714309555e-06,
+      "loss": 0.8578,
+      "step": 26964
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.096650375038427e-06,
+      "loss": 0.7851,
+      "step": 26965
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.095962078697116e-06,
+      "loss": 0.9348,
+      "step": 26966
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.095273825290628e-06,
+      "loss": 0.9943,
+      "step": 26967
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.09458561482396e-06,
+      "loss": 0.929,
+      "step": 26968
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.093897447302122e-06,
+      "loss": 0.8746,
+      "step": 26969
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.093209322730114e-06,
+      "loss": 0.8354,
+      "step": 26970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.092521241112948e-06,
+      "loss": 0.7824,
+      "step": 26971
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.091833202455623e-06,
+      "loss": 0.7444,
+      "step": 26972
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0911452067631415e-06,
+      "loss": 0.5307,
+      "step": 26973
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.09045725404051e-06,
+      "loss": 0.7057,
+      "step": 26974
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.08976934429273e-06,
+      "loss": 0.7354,
+      "step": 26975
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.089081477524802e-06,
+      "loss": 0.9087,
+      "step": 26976
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.088393653741731e-06,
+      "loss": 0.6154,
+      "step": 26977
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.087705872948518e-06,
+      "loss": 0.7716,
+      "step": 26978
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.087018135150164e-06,
+      "loss": 0.8355,
+      "step": 26979
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.08633044035167e-06,
+      "loss": 0.7136,
+      "step": 26980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.085642788558038e-06,
+      "loss": 0.7584,
+      "step": 26981
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0849551797742694e-06,
+      "loss": 0.9583,
+      "step": 26982
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.08426761400536e-06,
+      "loss": 0.9792,
+      "step": 26983
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0835800912563165e-06,
+      "loss": 0.8872,
+      "step": 26984
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.082892611532136e-06,
+      "loss": 1.0196,
+      "step": 26985
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.082205174837816e-06,
+      "loss": 0.9305,
+      "step": 26986
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.081517781178362e-06,
+      "loss": 0.7184,
+      "step": 26987
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.080830430558761e-06,
+      "loss": 0.6643,
+      "step": 26988
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0801431229840215e-06,
+      "loss": 1.0033,
+      "step": 26989
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.079455858459138e-06,
+      "loss": 0.8238,
+      "step": 26990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.07876863698911e-06,
+      "loss": 0.6385,
+      "step": 26991
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.078081458578934e-06,
+      "loss": 1.0812,
+      "step": 26992
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0773943232336066e-06,
+      "loss": 1.0035,
+      "step": 26993
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.076707230958126e-06,
+      "loss": 0.8636,
+      "step": 26994
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.076020181757484e-06,
+      "loss": 0.862,
+      "step": 26995
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0753331756366845e-06,
+      "loss": 0.8788,
+      "step": 26996
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0746462126007245e-06,
+      "loss": 0.8575,
+      "step": 26997
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.073959292654591e-06,
+      "loss": 0.8285,
+      "step": 26998
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.073272415803283e-06,
+      "loss": 0.888,
+      "step": 26999
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.072585582051798e-06,
+      "loss": 0.9946,
+      "step": 27000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.071898791405124e-06,
+      "loss": 0.8441,
+      "step": 27001
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0712120438682645e-06,
+      "loss": 0.7687,
+      "step": 27002
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0705253394462084e-06,
+      "loss": 0.9053,
+      "step": 27003
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.069838678143951e-06,
+      "loss": 1.0682,
+      "step": 27004
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0691520599664855e-06,
+      "loss": 1.0496,
+      "step": 27005
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.068465484918804e-06,
+      "loss": 0.8408,
+      "step": 27006
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.067778953005902e-06,
+      "loss": 0.7611,
+      "step": 27007
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.067092464232768e-06,
+      "loss": 0.9335,
+      "step": 27008
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.066406018604399e-06,
+      "loss": 0.9043,
+      "step": 27009
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.065719616125782e-06,
+      "loss": 0.7157,
+      "step": 27010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.065033256801911e-06,
+      "loss": 0.7381,
+      "step": 27011
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.064346940637779e-06,
+      "loss": 0.9306,
+      "step": 27012
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.063660667638371e-06,
+      "loss": 0.8483,
+      "step": 27013
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.062974437808685e-06,
+      "loss": 1.043,
+      "step": 27014
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0622882511537076e-06,
+      "loss": 0.829,
+      "step": 27015
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.06160210767843e-06,
+      "loss": 0.8162,
+      "step": 27016
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.060916007387845e-06,
+      "loss": 0.7355,
+      "step": 27017
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.060229950286935e-06,
+      "loss": 0.8585,
+      "step": 27018
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.059543936380688e-06,
+      "loss": 0.6972,
+      "step": 27019
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.058857965674101e-06,
+      "loss": 0.6923,
+      "step": 27020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.058172038172159e-06,
+      "loss": 0.8526,
+      "step": 27021
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.057486153879849e-06,
+      "loss": 0.5639,
+      "step": 27022
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.056800312802161e-06,
+      "loss": 1.0288,
+      "step": 27023
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.056114514944079e-06,
+      "loss": 0.9502,
+      "step": 27024
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.055428760310595e-06,
+      "loss": 0.8654,
+      "step": 27025
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0547430489066876e-06,
+      "loss": 0.8165,
+      "step": 27026
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.054057380737356e-06,
+      "loss": 0.8346,
+      "step": 27027
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.053371755807577e-06,
+      "loss": 0.8719,
+      "step": 27028
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.052686174122339e-06,
+      "loss": 1.217,
+      "step": 27029
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.052000635686627e-06,
+      "loss": 1.017,
+      "step": 27030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.051315140505423e-06,
+      "loss": 0.7576,
+      "step": 27031
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.050629688583719e-06,
+      "loss": 0.7953,
+      "step": 27032
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0499442799264975e-06,
+      "loss": 0.8597,
+      "step": 27033
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.049258914538743e-06,
+      "loss": 0.8849,
+      "step": 27034
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.048573592425436e-06,
+      "loss": 0.6213,
+      "step": 27035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0478883135915645e-06,
+      "loss": 0.8261,
+      "step": 27036
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.04720307804211e-06,
+      "loss": 0.9049,
+      "step": 27037
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.046517885782055e-06,
+      "loss": 0.7469,
+      "step": 27038
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045832736816384e-06,
+      "loss": 0.664,
+      "step": 27039
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045147631150078e-06,
+      "loss": 1.1772,
+      "step": 27040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.044462568788119e-06,
+      "loss": 0.5696,
+      "step": 27041
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0437775497354905e-06,
+      "loss": 0.9453,
+      "step": 27042
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0430925739971724e-06,
+      "loss": 0.859,
+      "step": 27043
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.042407641578145e-06,
+      "loss": 0.7729,
+      "step": 27044
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0417227524833925e-06,
+      "loss": 0.7995,
+      "step": 27045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.041037906717894e-06,
+      "loss": 1.0512,
+      "step": 27046
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.040353104286633e-06,
+      "loss": 0.7078,
+      "step": 27047
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.039668345194583e-06,
+      "loss": 0.754,
+      "step": 27048
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.038983629446722e-06,
+      "loss": 1.0071,
+      "step": 27049
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.038298957048039e-06,
+      "loss": 0.8414,
+      "step": 27050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.037614328003507e-06,
+      "loss": 0.6222,
+      "step": 27051
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0369297423181066e-06,
+      "loss": 0.9251,
+      "step": 27052
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.036245199996814e-06,
+      "loss": 0.6275,
+      "step": 27053
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0355607010446094e-06,
+      "loss": 0.9215,
+      "step": 27054
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.034876245466471e-06,
+      "loss": 1.0137,
+      "step": 27055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.034191833267369e-06,
+      "loss": 0.7852,
+      "step": 27056
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.033507464452296e-06,
+      "loss": 0.9087,
+      "step": 27057
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.032823139026214e-06,
+      "loss": 1.0255,
+      "step": 27058
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.032138856994104e-06,
+      "loss": 0.8339,
+      "step": 27059
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.031454618360945e-06,
+      "loss": 0.8307,
+      "step": 27060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.03077042313171e-06,
+      "loss": 0.7465,
+      "step": 27061
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.030086271311372e-06,
+      "loss": 0.6471,
+      "step": 27062
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.029402162904913e-06,
+      "loss": 0.8453,
+      "step": 27063
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.028718097917304e-06,
+      "loss": 0.5214,
+      "step": 27064
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.028034076353521e-06,
+      "loss": 1.0192,
+      "step": 27065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.027350098218537e-06,
+      "loss": 0.9249,
+      "step": 27066
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0266661635173255e-06,
+      "loss": 0.7474,
+      "step": 27067
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.025982272254861e-06,
+      "loss": 0.7129,
+      "step": 27068
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.025298424436118e-06,
+      "loss": 0.5496,
+      "step": 27069
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.024614620066067e-06,
+      "loss": 0.6767,
+      "step": 27070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.023930859149684e-06,
+      "loss": 0.8745,
+      "step": 27071
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.023247141691938e-06,
+      "loss": 1.0086,
+      "step": 27072
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0225634676978035e-06,
+      "loss": 0.8091,
+      "step": 27073
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.021879837172246e-06,
+      "loss": 0.5716,
+      "step": 27074
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.021196250120248e-06,
+      "loss": 0.719,
+      "step": 27075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.020512706546772e-06,
+      "loss": 0.76,
+      "step": 27076
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.019829206456797e-06,
+      "loss": 0.853,
+      "step": 27077
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.019145749855285e-06,
+      "loss": 0.7422,
+      "step": 27078
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.018462336747208e-06,
+      "loss": 0.7979,
+      "step": 27079
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017778967137533e-06,
+      "loss": 0.7659,
+      "step": 27080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017095641031239e-06,
+      "loss": 1.0207,
+      "step": 27081
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.016412358433288e-06,
+      "loss": 0.902,
+      "step": 27082
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.015729119348652e-06,
+      "loss": 0.8542,
+      "step": 27083
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.015045923782299e-06,
+      "loss": 1.0514,
+      "step": 27084
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014362771739195e-06,
+      "loss": 1.0806,
+      "step": 27085
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.013679663224311e-06,
+      "loss": 0.8099,
+      "step": 27086
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.012996598242613e-06,
+      "loss": 0.7299,
+      "step": 27087
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.012313576799069e-06,
+      "loss": 0.9558,
+      "step": 27088
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0116305988986444e-06,
+      "loss": 0.6452,
+      "step": 27089
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0109476645463076e-06,
+      "loss": 0.9994,
+      "step": 27090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.010264773747025e-06,
+      "loss": 0.8539,
+      "step": 27091
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.009581926505758e-06,
+      "loss": 1.0565,
+      "step": 27092
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.00889912282748e-06,
+      "loss": 0.9381,
+      "step": 27093
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.008216362717152e-06,
+      "loss": 0.7366,
+      "step": 27094
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007533646179741e-06,
+      "loss": 0.9775,
+      "step": 27095
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.006850973220211e-06,
+      "loss": 0.8567,
+      "step": 27096
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0061683438435295e-06,
+      "loss": 0.7128,
+      "step": 27097
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.00548575805465e-06,
+      "loss": 0.7678,
+      "step": 27098
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.004803215858548e-06,
+      "loss": 0.812,
+      "step": 27099
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.004120717260182e-06,
+      "loss": 0.6724,
+      "step": 27100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.003438262264516e-06,
+      "loss": 0.8794,
+      "step": 27101
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.002755850876514e-06,
+      "loss": 0.7957,
+      "step": 27102
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.002073483101136e-06,
+      "loss": 0.9016,
+      "step": 27103
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.001391158943346e-06,
+      "loss": 0.7905,
+      "step": 27104
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000708878408103e-06,
+      "loss": 0.9804,
+      "step": 27105
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000026641500374e-06,
+      "loss": 0.9193,
+      "step": 27106
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.999344448225121e-06,
+      "loss": 0.7076,
+      "step": 27107
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.998662298587298e-06,
+      "loss": 0.7494,
+      "step": 27108
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997980192591869e-06,
+      "loss": 1.0047,
+      "step": 27109
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.997298130243792e-06,
+      "loss": 0.8325,
+      "step": 27110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9966161115480325e-06,
+      "loss": 0.9527,
+      "step": 27111
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.995934136509547e-06,
+      "loss": 0.734,
+      "step": 27112
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9952522051332966e-06,
+      "loss": 0.7724,
+      "step": 27113
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994570317424238e-06,
+      "loss": 1.0534,
+      "step": 27114
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.99388847338733e-06,
+      "loss": 1.1258,
+      "step": 27115
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.993206673027533e-06,
+      "loss": 0.9041,
+      "step": 27116
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.992524916349804e-06,
+      "loss": 0.9126,
+      "step": 27117
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9918432033591e-06,
+      "loss": 0.9979,
+      "step": 27118
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.991161534060381e-06,
+      "loss": 0.743,
+      "step": 27119
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.990479908458602e-06,
+      "loss": 0.7032,
+      "step": 27120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98979832655872e-06,
+      "loss": 1.0008,
+      "step": 27121
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.989116788365692e-06,
+      "loss": 0.6889,
+      "step": 27122
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.988435293884472e-06,
+      "loss": 0.9031,
+      "step": 27123
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98775384312002e-06,
+      "loss": 0.7491,
+      "step": 27124
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.987072436077291e-06,
+      "loss": 0.8976,
+      "step": 27125
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.986391072761237e-06,
+      "loss": 1.0408,
+      "step": 27126
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.985709753176821e-06,
+      "loss": 0.925,
+      "step": 27127
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.985028477328982e-06,
+      "loss": 0.6745,
+      "step": 27128
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9843472452226895e-06,
+      "loss": 1.1087,
+      "step": 27129
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9836660568628904e-06,
+      "loss": 0.6331,
+      "step": 27130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.98298491225454e-06,
+      "loss": 0.8285,
+      "step": 27131
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9823038114025926e-06,
+      "loss": 0.6004,
+      "step": 27132
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.981622754311999e-06,
+      "loss": 1.1211,
+      "step": 27133
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980941740987714e-06,
+      "loss": 0.7736,
+      "step": 27134
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.980260771434685e-06,
+      "loss": 0.6276,
+      "step": 27135
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9795798456578715e-06,
+      "loss": 0.7228,
+      "step": 27136
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9788989636622255e-06,
+      "loss": 0.9832,
+      "step": 27137
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9782181254526915e-06,
+      "loss": 0.7495,
+      "step": 27138
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.977537331034224e-06,
+      "loss": 1.0402,
+      "step": 27139
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9768565804117745e-06,
+      "loss": 0.8696,
+      "step": 27140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.976175873590288e-06,
+      "loss": 1.0782,
+      "step": 27141
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.975495210574723e-06,
+      "loss": 1.0854,
+      "step": 27142
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974814591370026e-06,
+      "loss": 0.8536,
+      "step": 27143
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974134015981147e-06,
+      "loss": 1.0631,
+      "step": 27144
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.973453484413035e-06,
+      "loss": 0.8006,
+      "step": 27145
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972772996670638e-06,
+      "loss": 0.8944,
+      "step": 27146
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972092552758905e-06,
+      "loss": 0.7186,
+      "step": 27147
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.971412152682785e-06,
+      "loss": 0.8974,
+      "step": 27148
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970731796447224e-06,
+      "loss": 1.1312,
+      "step": 27149
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970051484057171e-06,
+      "loss": 1.2219,
+      "step": 27150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9693712155175735e-06,
+      "loss": 1.071,
+      "step": 27151
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.968690990833378e-06,
+      "loss": 0.7513,
+      "step": 27152
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.968010810009528e-06,
+      "loss": 0.6119,
+      "step": 27153
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.967330673050978e-06,
+      "loss": 0.8204,
+      "step": 27154
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.966650579962667e-06,
+      "loss": 0.9752,
+      "step": 27155
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.965970530749545e-06,
+      "loss": 0.8963,
+      "step": 27156
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9652905254165575e-06,
+      "loss": 0.7459,
+      "step": 27157
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.964610563968645e-06,
+      "loss": 0.7761,
+      "step": 27158
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963930646410751e-06,
+      "loss": 0.813,
+      "step": 27159
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963250772747826e-06,
+      "loss": 0.7416,
+      "step": 27160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.962570942984813e-06,
+      "loss": 0.888,
+      "step": 27161
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.961891157126654e-06,
+      "loss": 0.5109,
+      "step": 27162
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9612114151782935e-06,
+      "loss": 0.8008,
+      "step": 27163
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9605317171446734e-06,
+      "loss": 0.7532,
+      "step": 27164
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.959852063030738e-06,
+      "loss": 0.7584,
+      "step": 27165
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.959172452841425e-06,
+      "loss": 0.4826,
+      "step": 27166
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.958492886581688e-06,
+      "loss": 0.9429,
+      "step": 27167
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957813364256459e-06,
+      "loss": 0.9543,
+      "step": 27168
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957133885870681e-06,
+      "loss": 1.0895,
+      "step": 27169
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.956454451429297e-06,
+      "loss": 0.6629,
+      "step": 27170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.955775060937244e-06,
+      "loss": 0.9149,
+      "step": 27171
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.955095714399468e-06,
+      "loss": 0.8684,
+      "step": 27172
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.95441641182091e-06,
+      "loss": 0.7796,
+      "step": 27173
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.953737153206505e-06,
+      "loss": 0.9185,
+      "step": 27174
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.953057938561194e-06,
+      "loss": 0.7342,
+      "step": 27175
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.952378767889919e-06,
+      "loss": 0.9497,
+      "step": 27176
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9516996411976175e-06,
+      "loss": 0.7515,
+      "step": 27177
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.951020558489227e-06,
+      "loss": 0.9874,
+      "step": 27178
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9503415197696875e-06,
+      "loss": 0.6509,
+      "step": 27179
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.949662525043935e-06,
+      "loss": 1.0114,
+      "step": 27180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9489835743169095e-06,
+      "loss": 0.7906,
+      "step": 27181
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9483046675935465e-06,
+      "loss": 0.6715,
+      "step": 27182
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.947625804878784e-06,
+      "loss": 0.5545,
+      "step": 27183
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946946986177555e-06,
+      "loss": 0.7495,
+      "step": 27184
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.946268211494805e-06,
+      "loss": 1.1704,
+      "step": 27185
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.945589480835463e-06,
+      "loss": 0.6458,
+      "step": 27186
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.944910794204472e-06,
+      "loss": 1.0894,
+      "step": 27187
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.944232151606757e-06,
+      "loss": 0.8511,
+      "step": 27188
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.943553553047255e-06,
+      "loss": 0.8561,
+      "step": 27189
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.942874998530908e-06,
+      "loss": 0.898,
+      "step": 27190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.942196488062646e-06,
+      "loss": 0.5166,
+      "step": 27191
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9415180216474046e-06,
+      "loss": 1.0268,
+      "step": 27192
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.940839599290117e-06,
+      "loss": 1.0875,
+      "step": 27193
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.940161220995716e-06,
+      "loss": 0.9073,
+      "step": 27194
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.939482886769136e-06,
+      "loss": 0.8275,
+      "step": 27195
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9388045966153055e-06,
+      "loss": 1.1442,
+      "step": 27196
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.938126350539168e-06,
+      "loss": 0.8639,
+      "step": 27197
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.937448148545646e-06,
+      "loss": 0.7082,
+      "step": 27198
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.936769990639675e-06,
+      "loss": 0.8117,
+      "step": 27199
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.936091876826185e-06,
+      "loss": 0.7181,
+      "step": 27200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.935413807110108e-06,
+      "loss": 1.0549,
+      "step": 27201
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.934735781496371e-06,
+      "loss": 0.8136,
+      "step": 27202
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.934057799989914e-06,
+      "loss": 1.1113,
+      "step": 27203
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.933379862595662e-06,
+      "loss": 0.6536,
+      "step": 27204
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.932701969318544e-06,
+      "loss": 0.9293,
+      "step": 27205
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.932024120163493e-06,
+      "loss": 0.7471,
+      "step": 27206
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.931346315135435e-06,
+      "loss": 0.7856,
+      "step": 27207
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.930668554239301e-06,
+      "loss": 0.7163,
+      "step": 27208
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929990837480019e-06,
+      "loss": 0.5885,
+      "step": 27209
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.929313164862518e-06,
+      "loss": 0.9659,
+      "step": 27210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.928635536391726e-06,
+      "loss": 0.753,
+      "step": 27211
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.92795795207257e-06,
+      "loss": 0.8167,
+      "step": 27212
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.927280411909978e-06,
+      "loss": 1.1335,
+      "step": 27213
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.926602915908875e-06,
+      "loss": 0.682,
+      "step": 27214
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925925464074192e-06,
+      "loss": 0.8388,
+      "step": 27215
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.925248056410854e-06,
+      "loss": 1.0783,
+      "step": 27216
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924570692923791e-06,
+      "loss": 0.8207,
+      "step": 27217
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923893373617919e-06,
+      "loss": 0.4819,
+      "step": 27218
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923216098498171e-06,
+      "loss": 0.8257,
+      "step": 27219
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.922538867569467e-06,
+      "loss": 0.6676,
+      "step": 27220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9218616808367385e-06,
+      "loss": 0.5505,
+      "step": 27221
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.921184538304907e-06,
+      "loss": 0.8516,
+      "step": 27222
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.920507439978894e-06,
+      "loss": 0.3774,
+      "step": 27223
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.919830385863629e-06,
+      "loss": 0.6695,
+      "step": 27224
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.919153375964032e-06,
+      "loss": 0.748,
+      "step": 27225
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.918476410285026e-06,
+      "loss": 1.0636,
+      "step": 27226
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9177994888315355e-06,
+      "loss": 0.9251,
+      "step": 27227
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.917122611608482e-06,
+      "loss": 0.733,
+      "step": 27228
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.91644577862079e-06,
+      "loss": 0.9404,
+      "step": 27229
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.915768989873377e-06,
+      "loss": 0.8128,
+      "step": 27230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.91509224537117e-06,
+      "loss": 0.9849,
+      "step": 27231
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.914415545119084e-06,
+      "loss": 0.8737,
+      "step": 27232
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9137388891220475e-06,
+      "loss": 0.8454,
+      "step": 27233
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.913062277384977e-06,
+      "loss": 0.9077,
+      "step": 27234
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.912385709912794e-06,
+      "loss": 0.7163,
+      "step": 27235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.911709186710417e-06,
+      "loss": 0.8534,
+      "step": 27236
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.911032707782772e-06,
+      "loss": 0.9401,
+      "step": 27237
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9103562731347654e-06,
+      "loss": 0.8687,
+      "step": 27238
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.909679882771328e-06,
+      "loss": 0.9878,
+      "step": 27239
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.909003536697374e-06,
+      "loss": 1.0497,
+      "step": 27240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9083272349178235e-06,
+      "loss": 0.7452,
+      "step": 27241
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9076509774375936e-06,
+      "loss": 0.8132,
+      "step": 27242
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.906974764261603e-06,
+      "loss": 0.7597,
+      "step": 27243
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9062985953947685e-06,
+      "loss": 1.0104,
+      "step": 27244
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.905622470842003e-06,
+      "loss": 0.9589,
+      "step": 27245
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.904946390608231e-06,
+      "loss": 0.7936,
+      "step": 27246
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.904270354698369e-06,
+      "loss": 0.9394,
+      "step": 27247
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.903594363117328e-06,
+      "loss": 0.695,
+      "step": 27248
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902918415870025e-06,
+      "loss": 0.7239,
+      "step": 27249
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902242512961376e-06,
+      "loss": 0.5981,
+      "step": 27250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.901566654396292e-06,
+      "loss": 1.0225,
+      "step": 27251
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.900890840179699e-06,
+      "loss": 0.9061,
+      "step": 27252
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.900215070316503e-06,
+      "loss": 0.7835,
+      "step": 27253
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.89953934481162e-06,
+      "loss": 0.7153,
+      "step": 27254
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.898863663669965e-06,
+      "loss": 0.8278,
+      "step": 27255
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.898188026896451e-06,
+      "loss": 0.9682,
+      "step": 27256
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8975124344959915e-06,
+      "loss": 0.9862,
+      "step": 27257
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8968368864735e-06,
+      "loss": 0.8354,
+      "step": 27258
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.896161382833886e-06,
+      "loss": 1.0563,
+      "step": 27259
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.895485923582067e-06,
+      "loss": 0.8324,
+      "step": 27260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.89481050872295e-06,
+      "loss": 0.8779,
+      "step": 27261
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.89413513826145e-06,
+      "loss": 0.5201,
+      "step": 27262
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8934598122024735e-06,
+      "loss": 0.6844,
+      "step": 27263
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.89278453055094e-06,
+      "loss": 1.044,
+      "step": 27264
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.892109293311753e-06,
+      "loss": 1.051,
+      "step": 27265
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.891434100489828e-06,
+      "loss": 1.0509,
+      "step": 27266
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.890758952090075e-06,
+      "loss": 0.6908,
+      "step": 27267
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.890083848117397e-06,
+      "loss": 0.7471,
+      "step": 27268
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.889408788576706e-06,
+      "loss": 1.0325,
+      "step": 27269
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.888733773472916e-06,
+      "loss": 0.9954,
+      "step": 27270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.888058802810932e-06,
+      "loss": 0.9805,
+      "step": 27271
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.887383876595663e-06,
+      "loss": 0.7266,
+      "step": 27272
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886708994832018e-06,
+      "loss": 0.5039,
+      "step": 27273
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886034157524904e-06,
+      "loss": 0.8228,
+      "step": 27274
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.885359364679228e-06,
+      "loss": 0.6517,
+      "step": 27275
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.884684616299894e-06,
+      "loss": 0.7802,
+      "step": 27276
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.88400991239182e-06,
+      "loss": 0.5547,
+      "step": 27277
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.883335252959901e-06,
+      "loss": 0.8809,
+      "step": 27278
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.882660638009048e-06,
+      "loss": 0.6138,
+      "step": 27279
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.881986067544166e-06,
+      "loss": 1.2604,
+      "step": 27280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.881311541570157e-06,
+      "loss": 0.7089,
+      "step": 27281
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.880637060091934e-06,
+      "loss": 0.6816,
+      "step": 27282
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8799626231143975e-06,
+      "loss": 0.7991,
+      "step": 27283
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.879288230642452e-06,
+      "loss": 0.7475,
+      "step": 27284
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.878613882681002e-06,
+      "loss": 0.9267,
+      "step": 27285
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.877939579234953e-06,
+      "loss": 1.0107,
+      "step": 27286
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.877265320309206e-06,
+      "loss": 0.8187,
+      "step": 27287
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.876591105908666e-06,
+      "loss": 0.4704,
+      "step": 27288
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.875916936038235e-06,
+      "loss": 0.6426,
+      "step": 27289
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.875242810702816e-06,
+      "loss": 0.8566,
+      "step": 27290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.874568729907312e-06,
+      "loss": 0.7982,
+      "step": 27291
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.873894693656625e-06,
+      "loss": 0.8542,
+      "step": 27292
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.873220701955655e-06,
+      "loss": 0.8072,
+      "step": 27293
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.872546754809301e-06,
+      "loss": 0.561,
+      "step": 27294
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.871872852222471e-06,
+      "loss": 0.4801,
+      "step": 27295
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.871198994200063e-06,
+      "loss": 0.8441,
+      "step": 27296
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.870525180746979e-06,
+      "loss": 0.8972,
+      "step": 27297
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.869851411868113e-06,
+      "loss": 0.6816,
+      "step": 27298
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.869177687568365e-06,
+      "loss": 0.7651,
+      "step": 27299
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.868504007852641e-06,
+      "loss": 1.0118,
+      "step": 27300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.867830372725838e-06,
+      "loss": 0.9361,
+      "step": 27301
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.867156782192853e-06,
+      "loss": 0.7646,
+      "step": 27302
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.866483236258584e-06,
+      "loss": 0.6064,
+      "step": 27303
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.86580973492793e-06,
+      "loss": 0.8089,
+      "step": 27304
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8651362782057885e-06,
+      "loss": 1.1797,
+      "step": 27305
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.864462866097055e-06,
+      "loss": 0.6396,
+      "step": 27306
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863789498606636e-06,
+      "loss": 0.7767,
+      "step": 27307
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863116175739417e-06,
+      "loss": 0.9343,
+      "step": 27308
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.862442897500298e-06,
+      "loss": 0.8893,
+      "step": 27309
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.861769663894177e-06,
+      "loss": 0.5028,
+      "step": 27310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.861096474925948e-06,
+      "loss": 0.7201,
+      "step": 27311
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.860423330600503e-06,
+      "loss": 0.8005,
+      "step": 27312
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.859750230922747e-06,
+      "loss": 0.7933,
+      "step": 27313
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.859077175897568e-06,
+      "loss": 0.8922,
+      "step": 27314
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8584041655298606e-06,
+      "loss": 0.9482,
+      "step": 27315
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.857731199824522e-06,
+      "loss": 0.6862,
+      "step": 27316
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.857058278786442e-06,
+      "loss": 0.9463,
+      "step": 27317
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.856385402420518e-06,
+      "loss": 0.8026,
+      "step": 27318
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8557125707316415e-06,
+      "loss": 1.0216,
+      "step": 27319
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.855039783724705e-06,
+      "loss": 0.8884,
+      "step": 27320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.854367041404602e-06,
+      "loss": 0.6958,
+      "step": 27321
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853694343776223e-06,
+      "loss": 1.0065,
+      "step": 27322
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853021690844463e-06,
+      "loss": 1.0772,
+      "step": 27323
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.852349082614208e-06,
+      "loss": 0.6207,
+      "step": 27324
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851676519090356e-06,
+      "loss": 0.8259,
+      "step": 27325
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.851004000277795e-06,
+      "loss": 0.8866,
+      "step": 27326
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.85033152618142e-06,
+      "loss": 0.7354,
+      "step": 27327
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.849659096806113e-06,
+      "loss": 0.7414,
+      "step": 27328
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.848986712156768e-06,
+      "loss": 0.8445,
+      "step": 27329
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.848314372238272e-06,
+      "loss": 0.945,
+      "step": 27330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.84764207705552e-06,
+      "loss": 0.9891,
+      "step": 27331
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.846969826613398e-06,
+      "loss": 0.6856,
+      "step": 27332
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.846297620916795e-06,
+      "loss": 1.0678,
+      "step": 27333
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8456254599705975e-06,
+      "loss": 1.0335,
+      "step": 27334
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.844953343779697e-06,
+      "loss": 1.1442,
+      "step": 27335
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.844281272348979e-06,
+      "loss": 0.8842,
+      "step": 27336
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.843609245683331e-06,
+      "loss": 0.6871,
+      "step": 27337
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.842937263787639e-06,
+      "loss": 0.6718,
+      "step": 27338
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.842265326666792e-06,
+      "loss": 0.8093,
+      "step": 27339
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.841593434325675e-06,
+      "loss": 0.6574,
+      "step": 27340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8409215867691755e-06,
+      "loss": 0.7027,
+      "step": 27341
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.840249784002174e-06,
+      "loss": 0.9566,
+      "step": 27342
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.839578026029563e-06,
+      "loss": 0.8777,
+      "step": 27343
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.838906312856226e-06,
+      "loss": 0.7911,
+      "step": 27344
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.838234644487045e-06,
+      "loss": 0.8138,
+      "step": 27345
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.837563020926907e-06,
+      "loss": 0.806,
+      "step": 27346
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.836891442180697e-06,
+      "loss": 0.7858,
+      "step": 27347
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.836219908253291e-06,
+      "loss": 0.9105,
+      "step": 27348
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.835548419149582e-06,
+      "loss": 0.8989,
+      "step": 27349
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.834876974874449e-06,
+      "loss": 0.702,
+      "step": 27350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8342055754327755e-06,
+      "loss": 0.7746,
+      "step": 27351
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.833534220829443e-06,
+      "loss": 0.8655,
+      "step": 27352
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8328629110693364e-06,
+      "loss": 1.1477,
+      "step": 27353
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.832191646157334e-06,
+      "loss": 0.8126,
+      "step": 27354
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.831520426098315e-06,
+      "loss": 0.7372,
+      "step": 27355
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.830849250897168e-06,
+      "loss": 0.7684,
+      "step": 27356
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.830178120558775e-06,
+      "loss": 0.5562,
+      "step": 27357
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.829507035088007e-06,
+      "loss": 0.4879,
+      "step": 27358
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.82883599448975e-06,
+      "loss": 0.7816,
+      "step": 27359
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.828164998768879e-06,
+      "loss": 0.9493,
+      "step": 27360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.827494047930281e-06,
+      "loss": 0.899,
+      "step": 27361
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8268231419788325e-06,
+      "loss": 0.5151,
+      "step": 27362
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.826152280919411e-06,
+      "loss": 0.6581,
+      "step": 27363
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.825481464756896e-06,
+      "loss": 0.7744,
+      "step": 27364
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.824810693496165e-06,
+      "loss": 0.8013,
+      "step": 27365
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.824139967142097e-06,
+      "loss": 0.8719,
+      "step": 27366
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.823469285699568e-06,
+      "loss": 0.8051,
+      "step": 27367
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.822798649173457e-06,
+      "loss": 0.6818,
+      "step": 27368
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.82212805756864e-06,
+      "loss": 0.9035,
+      "step": 27369
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.821457510889993e-06,
+      "loss": 1.049,
+      "step": 27370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.820787009142395e-06,
+      "loss": 0.9533,
+      "step": 27371
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.820116552330718e-06,
+      "loss": 0.9449,
+      "step": 27372
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.819446140459837e-06,
+      "loss": 0.9452,
+      "step": 27373
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.818775773534634e-06,
+      "loss": 0.7942,
+      "step": 27374
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8181054515599806e-06,
+      "loss": 0.6849,
+      "step": 27375
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.817435174540749e-06,
+      "loss": 0.9433,
+      "step": 27376
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.81676494248182e-06,
+      "loss": 1.1643,
+      "step": 27377
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.816094755388056e-06,
+      "loss": 0.9938,
+      "step": 27378
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8154246132643405e-06,
+      "loss": 0.9739,
+      "step": 27379
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.814754516115544e-06,
+      "loss": 0.5845,
+      "step": 27380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8140844639465404e-06,
+      "loss": 0.8088,
+      "step": 27381
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8134144567622e-06,
+      "loss": 0.5946,
+      "step": 27382
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8127444945673975e-06,
+      "loss": 0.8544,
+      "step": 27383
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8120745773670044e-06,
+      "loss": 0.8612,
+      "step": 27384
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8114047051658885e-06,
+      "loss": 0.805,
+      "step": 27385
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.810734877968929e-06,
+      "loss": 0.9715,
+      "step": 27386
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8100650957809947e-06,
+      "loss": 0.8572,
+      "step": 27387
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8093953586069523e-06,
+      "loss": 0.8471,
+      "step": 27388
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8087256664516737e-06,
+      "loss": 0.5965,
+      "step": 27389
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8080560193200288e-06,
+      "loss": 1.1179,
+      "step": 27390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.807386417216886e-06,
+      "loss": 0.7535,
+      "step": 27391
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8067168601471195e-06,
+      "loss": 0.7873,
+      "step": 27392
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.806047348115597e-06,
+      "loss": 0.853,
+      "step": 27393
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8053778811271847e-06,
+      "loss": 0.8112,
+      "step": 27394
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.804708459186752e-06,
+      "loss": 0.9637,
+      "step": 27395
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.804039082299168e-06,
+      "loss": 0.7895,
+      "step": 27396
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8033697504693e-06,
+      "loss": 0.8579,
+      "step": 27397
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.802700463702016e-06,
+      "loss": 0.7834,
+      "step": 27398
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8020312220021817e-06,
+      "loss": 0.7863,
+      "step": 27399
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.801362025374664e-06,
+      "loss": 1.1124,
+      "step": 27400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8006928738243297e-06,
+      "loss": 0.9686,
+      "step": 27401
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.800023767356047e-06,
+      "loss": 0.9005,
+      "step": 27402
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.799354705974675e-06,
+      "loss": 0.645,
+      "step": 27403
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.798685689685089e-06,
+      "loss": 0.7436,
+      "step": 27404
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.798016718492148e-06,
+      "loss": 0.8488,
+      "step": 27405
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7973477924007184e-06,
+      "loss": 0.952,
+      "step": 27406
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7966789114156677e-06,
+      "loss": 0.6679,
+      "step": 27407
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7960100755418537e-06,
+      "loss": 0.7748,
+      "step": 27408
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.795341284784141e-06,
+      "loss": 0.8527,
+      "step": 27409
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.794672539147397e-06,
+      "loss": 0.9974,
+      "step": 27410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.794003838636484e-06,
+      "loss": 0.8142,
+      "step": 27411
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7933351832562647e-06,
+      "loss": 0.8354,
+      "step": 27412
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.792666573011601e-06,
+      "loss": 1.0253,
+      "step": 27413
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.791998007907356e-06,
+      "loss": 0.9209,
+      "step": 27414
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7913294879483897e-06,
+      "loss": 0.8435,
+      "step": 27415
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7906610131395617e-06,
+      "loss": 1.0345,
+      "step": 27416
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7899925834857433e-06,
+      "loss": 0.7441,
+      "step": 27417
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.789324198991785e-06,
+      "loss": 0.8853,
+      "step": 27418
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.788655859662551e-06,
+      "loss": 0.7802,
+      "step": 27419
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7879875655029018e-06,
+      "loss": 1.1008,
+      "step": 27420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7873193165176925e-06,
+      "loss": 0.7968,
+      "step": 27421
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7866511127117912e-06,
+      "loss": 0.5666,
+      "step": 27422
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7859829540900515e-06,
+      "loss": 0.9956,
+      "step": 27423
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7853148406573357e-06,
+      "loss": 0.9672,
+      "step": 27424
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7846467724185e-06,
+      "loss": 0.8307,
+      "step": 27425
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.783978749378402e-06,
+      "loss": 0.8658,
+      "step": 27426
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7833107715419017e-06,
+      "loss": 0.7692,
+      "step": 27427
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.782642838913856e-06,
+      "loss": 0.8072,
+      "step": 27428
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.781974951499122e-06,
+      "loss": 1.0304,
+      "step": 27429
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7813071093025567e-06,
+      "loss": 0.5846,
+      "step": 27430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7806393123290166e-06,
+      "loss": 0.9167,
+      "step": 27431
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7799715605833575e-06,
+      "loss": 0.891,
+      "step": 27432
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7793038540704365e-06,
+      "loss": 0.8197,
+      "step": 27433
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7786361927951044e-06,
+      "loss": 0.8652,
+      "step": 27434
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7779685767622255e-06,
+      "loss": 0.9086,
+      "step": 27435
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.777301005976649e-06,
+      "loss": 0.8876,
+      "step": 27436
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.776633480443235e-06,
+      "loss": 0.5437,
+      "step": 27437
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7759660001668285e-06,
+      "loss": 0.4895,
+      "step": 27438
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7752985651522854e-06,
+      "loss": 0.9372,
+      "step": 27439
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7746311754044675e-06,
+      "loss": 0.9547,
+      "step": 27440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.773963830928221e-06,
+      "loss": 1.0491,
+      "step": 27441
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.773296531728402e-06,
+      "loss": 0.9974,
+      "step": 27442
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7726292778098616e-06,
+      "loss": 0.684,
+      "step": 27443
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.771962069177453e-06,
+      "loss": 1.0764,
+      "step": 27444
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7712949058360273e-06,
+      "loss": 0.8095,
+      "step": 27445
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.770627787790434e-06,
+      "loss": 0.9396,
+      "step": 27446
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.769960715045533e-06,
+      "loss": 0.6908,
+      "step": 27447
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.769293687606167e-06,
+      "loss": 0.6822,
+      "step": 27448
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7686267054771887e-06,
+      "loss": 0.8967,
+      "step": 27449
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7679597686634495e-06,
+      "loss": 0.7962,
+      "step": 27450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.767292877169799e-06,
+      "loss": 0.8718,
+      "step": 27451
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7666260310010827e-06,
+      "loss": 0.8718,
+      "step": 27452
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7659592301621583e-06,
+      "loss": 0.7029,
+      "step": 27453
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.76529247465787e-06,
+      "loss": 1.0986,
+      "step": 27454
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7646257644930674e-06,
+      "loss": 0.9155,
+      "step": 27455
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7639590996725982e-06,
+      "loss": 0.423,
+      "step": 27456
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7632924802013116e-06,
+      "loss": 1.0635,
+      "step": 27457
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7626259060840544e-06,
+      "loss": 0.8353,
+      "step": 27458
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7619593773256745e-06,
+      "loss": 0.9688,
+      "step": 27459
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.761292893931019e-06,
+      "loss": 0.6687,
+      "step": 27460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.760626455904933e-06,
+      "loss": 1.0364,
+      "step": 27461
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.759960063252266e-06,
+      "loss": 0.7644,
+      "step": 27462
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.759293715977861e-06,
+      "loss": 0.9357,
+      "step": 27463
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.758627414086562e-06,
+      "loss": 0.8164,
+      "step": 27464
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.757961157583221e-06,
+      "loss": 0.7606,
+      "step": 27465
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7572949464726805e-06,
+      "loss": 0.7562,
+      "step": 27466
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.756628780759787e-06,
+      "loss": 0.7837,
+      "step": 27467
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.755962660449379e-06,
+      "loss": 0.7589,
+      "step": 27468
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.755296585546304e-06,
+      "loss": 0.9542,
+      "step": 27469
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7546305560554018e-06,
+      "loss": 0.7659,
+      "step": 27470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.753964571981523e-06,
+      "loss": 1.2571,
+      "step": 27471
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7532986333295087e-06,
+      "loss": 0.8727,
+      "step": 27472
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7526327401041995e-06,
+      "loss": 0.6697,
+      "step": 27473
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7519668923104393e-06,
+      "loss": 0.7932,
+      "step": 27474
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7513010899530688e-06,
+      "loss": 0.7871,
+      "step": 27475
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7506353330369318e-06,
+      "loss": 0.8637,
+      "step": 27476
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.749969621566868e-06,
+      "loss": 0.4897,
+      "step": 27477
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7493039555477185e-06,
+      "loss": 0.7638,
+      "step": 27478
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.748638334984326e-06,
+      "loss": 0.7642,
+      "step": 27479
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7479727598815287e-06,
+      "loss": 0.8362,
+      "step": 27480
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.747307230244168e-06,
+      "loss": 0.8677,
+      "step": 27481
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.746641746077081e-06,
+      "loss": 0.8555,
+      "step": 27482
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.745976307385112e-06,
+      "loss": 0.8619,
+      "step": 27483
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.745310914173098e-06,
+      "loss": 0.7681,
+      "step": 27484
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.744645566445878e-06,
+      "loss": 0.7356,
+      "step": 27485
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.74398026420829e-06,
+      "loss": 0.9533,
+      "step": 27486
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.743315007465176e-06,
+      "loss": 0.8804,
+      "step": 27487
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7426497962213625e-06,
+      "loss": 0.8514,
+      "step": 27488
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7419846304816978e-06,
+      "loss": 0.5777,
+      "step": 27489
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.741319510251016e-06,
+      "loss": 0.9109,
+      "step": 27490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7406544355341547e-06,
+      "loss": 0.7488,
+      "step": 27491
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7399894063359478e-06,
+      "loss": 0.771,
+      "step": 27492
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.739324422661235e-06,
+      "loss": 0.6777,
+      "step": 27493
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.738659484514848e-06,
+      "loss": 0.5443,
+      "step": 27494
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7379945919016225e-06,
+      "loss": 0.7854,
+      "step": 27495
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.737329744826399e-06,
+      "loss": 1.0094,
+      "step": 27496
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.736664943294013e-06,
+      "loss": 0.8187,
+      "step": 27497
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.73600018730929e-06,
+      "loss": 1.1611,
+      "step": 27498
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.73533547687707e-06,
+      "loss": 0.8966,
+      "step": 27499
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7346708120021837e-06,
+      "loss": 0.8878,
+      "step": 27500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.734006192689469e-06,
+      "loss": 0.6549,
+      "step": 27501
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7333416189437576e-06,
+      "loss": 0.8306,
+      "step": 27502
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.732677090769883e-06,
+      "loss": 0.6519,
+      "step": 27503
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.732012608172675e-06,
+      "loss": 1.0775,
+      "step": 27504
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7313481711569676e-06,
+      "loss": 0.8812,
+      "step": 27505
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.730683779727593e-06,
+      "loss": 0.6632,
+      "step": 27506
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7300194338893816e-06,
+      "loss": 0.9317,
+      "step": 27507
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.729355133647166e-06,
+      "loss": 0.8936,
+      "step": 27508
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.728690879005775e-06,
+      "loss": 0.7169,
+      "step": 27509
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7280266699700406e-06,
+      "loss": 0.7617,
+      "step": 27510
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7273625065447936e-06,
+      "loss": 0.881,
+      "step": 27511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.726698388734863e-06,
+      "loss": 0.5942,
+      "step": 27512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.726034316545074e-06,
+      "loss": 0.7395,
+      "step": 27513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7253702899802637e-06,
+      "loss": 0.838,
+      "step": 27514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7247063090452574e-06,
+      "loss": 1.0246,
+      "step": 27515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.724042373744884e-06,
+      "loss": 1.0569,
+      "step": 27516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.723378484083975e-06,
+      "loss": 0.8423,
+      "step": 27517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7227146400673477e-06,
+      "loss": 0.7748,
+      "step": 27518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7220508416998404e-06,
+      "loss": 0.7461,
+      "step": 27519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7213870889862757e-06,
+      "loss": 0.6252,
+      "step": 27520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7207233819314815e-06,
+      "loss": 0.674,
+      "step": 27521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7200597205402844e-06,
+      "loss": 0.5987,
+      "step": 27522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.719396104817511e-06,
+      "loss": 0.7811,
+      "step": 27523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.718732534767986e-06,
+      "loss": 0.5709,
+      "step": 27524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7180690103965313e-06,
+      "loss": 0.6108,
+      "step": 27525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7174055317079807e-06,
+      "loss": 0.8847,
+      "step": 27526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7167420987071577e-06,
+      "loss": 0.8517,
+      "step": 27527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7160787113988804e-06,
+      "loss": 0.8802,
+      "step": 27528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.715415369787978e-06,
+      "loss": 0.6795,
+      "step": 27529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7147520738792707e-06,
+      "loss": 1.2179,
+      "step": 27530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.714088823677583e-06,
+      "loss": 0.9064,
+      "step": 27531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7134256191877416e-06,
+      "loss": 0.6899,
+      "step": 27532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712762460414567e-06,
+      "loss": 0.8517,
+      "step": 27533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712099347362882e-06,
+      "loss": 0.7984,
+      "step": 27534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.711436280037509e-06,
+      "loss": 0.6862,
+      "step": 27535
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.71077325844327e-06,
+      "loss": 0.7323,
+      "step": 27536
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7101102825849865e-06,
+      "loss": 0.8824,
+      "step": 27537
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.709447352467479e-06,
+      "loss": 0.6655,
+      "step": 27538
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7087844680955686e-06,
+      "loss": 0.9117,
+      "step": 27539
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7081216294740773e-06,
+      "loss": 0.9939,
+      "step": 27540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7074588366078235e-06,
+      "loss": 0.8823,
+      "step": 27541
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7067960895016277e-06,
+      "loss": 0.8945,
+      "step": 27542
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7061333881603056e-06,
+      "loss": 0.9222,
+      "step": 27543
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.705470732588685e-06,
+      "loss": 0.5856,
+      "step": 27544
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.70480812279158e-06,
+      "loss": 0.9418,
+      "step": 27545
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7041455587738083e-06,
+      "loss": 0.7645,
+      "step": 27546
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7034830405401945e-06,
+      "loss": 0.8537,
+      "step": 27547
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7028205680955466e-06,
+      "loss": 0.8284,
+      "step": 27548
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.702158141444684e-06,
+      "loss": 0.8713,
+      "step": 27549
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7014957605924296e-06,
+      "loss": 0.7725,
+      "step": 27550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.700833425543597e-06,
+      "loss": 0.9508,
+      "step": 27551
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7001711363030037e-06,
+      "loss": 0.7925,
+      "step": 27552
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.699508892875466e-06,
+      "loss": 0.8594,
+      "step": 27553
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6988466952657987e-06,
+      "loss": 0.6887,
+      "step": 27554
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6981845434788188e-06,
+      "loss": 1.0405,
+      "step": 27555
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6975224375193364e-06,
+      "loss": 0.7218,
+      "step": 27556
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6968603773921774e-06,
+      "loss": 0.8601,
+      "step": 27557
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.696198363102147e-06,
+      "loss": 0.8262,
+      "step": 27558
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.695536394654061e-06,
+      "loss": 0.6266,
+      "step": 27559
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6948744720527364e-06,
+      "loss": 1.0315,
+      "step": 27560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6942125953029796e-06,
+      "loss": 0.7271,
+      "step": 27561
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6935507644096135e-06,
+      "loss": 0.6746,
+      "step": 27562
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6928889793774457e-06,
+      "loss": 0.9455,
+      "step": 27563
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6922272402112904e-06,
+      "loss": 0.8358,
+      "step": 27564
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6915655469159582e-06,
+      "loss": 1.0903,
+      "step": 27565
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.690903899496263e-06,
+      "loss": 0.7455,
+      "step": 27566
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.690242297957014e-06,
+      "loss": 1.0003,
+      "step": 27567
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.689580742303025e-06,
+      "loss": 0.9069,
+      "step": 27568
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6889192325391045e-06,
+      "loss": 0.8736,
+      "step": 27569
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.688257768670065e-06,
+      "loss": 0.8339,
+      "step": 27570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.687596350700715e-06,
+      "loss": 0.7321,
+      "step": 27571
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6869349786358666e-06,
+      "loss": 0.9417,
+      "step": 27572
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6862736524803264e-06,
+      "loss": 0.8505,
+      "step": 27573
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.685612372238904e-06,
+      "loss": 0.974,
+      "step": 27574
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6849511379164117e-06,
+      "loss": 0.8859,
+      "step": 27575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6842899495176556e-06,
+      "loss": 0.9345,
+      "step": 27576
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.683628807047448e-06,
+      "loss": 1.0907,
+      "step": 27577
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.682967710510591e-06,
+      "loss": 0.6518,
+      "step": 27578
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6823066599118894e-06,
+      "loss": 1.0436,
+      "step": 27579
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.68164565525616e-06,
+      "loss": 0.8,
+      "step": 27580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.680984696548203e-06,
+      "loss": 0.6943,
+      "step": 27581
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6803237837928285e-06,
+      "loss": 0.8896,
+      "step": 27582
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6796629169948417e-06,
+      "loss": 0.7042,
+      "step": 27583
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6790020961590468e-06,
+      "loss": 0.7922,
+      "step": 27584
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.678341321290252e-06,
+      "loss": 0.7253,
+      "step": 27585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6776805923932557e-06,
+      "loss": 0.7564,
+      "step": 27586
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6770199094728763e-06,
+      "loss": 1.0894,
+      "step": 27587
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.676359272533907e-06,
+      "loss": 0.7217,
+      "step": 27588
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.675698681581155e-06,
+      "loss": 0.6141,
+      "step": 27589
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6750381366194244e-06,
+      "loss": 0.6864,
+      "step": 27590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.674377637653519e-06,
+      "loss": 0.6376,
+      "step": 27591
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.673717184688238e-06,
+      "loss": 0.5024,
+      "step": 27592
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6730567777283923e-06,
+      "loss": 0.6981,
+      "step": 27593
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6723964167787806e-06,
+      "loss": 0.7277,
+      "step": 27594
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6717361018442033e-06,
+      "loss": 0.7337,
+      "step": 27595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.671075832929465e-06,
+      "loss": 0.9232,
+      "step": 27596
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.670415610039366e-06,
+      "loss": 0.8749,
+      "step": 27597
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.669755433178708e-06,
+      "loss": 0.7102,
+      "step": 27598
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66909530235229e-06,
+      "loss": 0.957,
+      "step": 27599
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.668435217564915e-06,
+      "loss": 0.762,
+      "step": 27600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.667775178821381e-06,
+      "loss": 0.8113,
+      "step": 27601
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.667115186126491e-06,
+      "loss": 0.8679,
+      "step": 27602
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6664552394850415e-06,
+      "loss": 0.8441,
+      "step": 27603
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6657953389018296e-06,
+      "loss": 0.8464,
+      "step": 27604
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66513548438166e-06,
+      "loss": 1.0401,
+      "step": 27605
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66447567592933e-06,
+      "loss": 0.9019,
+      "step": 27606
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.663815913549639e-06,
+      "loss": 0.9434,
+      "step": 27607
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6631561972473795e-06,
+      "loss": 0.7336,
+      "step": 27608
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6624965270273514e-06,
+      "loss": 0.7348,
+      "step": 27609
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.661836902894349e-06,
+      "loss": 0.9554,
+      "step": 27610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6611773248531757e-06,
+      "loss": 0.5917,
+      "step": 27611
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.660517792908624e-06,
+      "loss": 0.5995,
+      "step": 27612
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6598583070654925e-06,
+      "loss": 0.5753,
+      "step": 27613
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.659198867328575e-06,
+      "loss": 1.0843,
+      "step": 27614
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.658539473702667e-06,
+      "loss": 0.8517,
+      "step": 27615
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6578801261925643e-06,
+      "loss": 0.8402,
+      "step": 27616
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.657220824803062e-06,
+      "loss": 0.5994,
+      "step": 27617
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.656561569538953e-06,
+      "loss": 0.8315,
+      "step": 27618
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6559023604050347e-06,
+      "loss": 0.8566,
+      "step": 27619
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6552431974060977e-06,
+      "loss": 0.7205,
+      "step": 27620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6545840805469368e-06,
+      "loss": 0.7613,
+      "step": 27621
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.653925009832342e-06,
+      "loss": 0.631,
+      "step": 27622
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6532659852671114e-06,
+      "loss": 0.6414,
+      "step": 27623
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.652607006856037e-06,
+      "loss": 1.1397,
+      "step": 27624
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.651948074603908e-06,
+      "loss": 1.0048,
+      "step": 27625
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.651289188515519e-06,
+      "loss": 0.7304,
+      "step": 27626
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.650630348595663e-06,
+      "loss": 0.9123,
+      "step": 27627
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6499715548491198e-06,
+      "loss": 0.7074,
+      "step": 27628
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.649312807280694e-06,
+      "loss": 0.7948,
+      "step": 27629
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6486541058951696e-06,
+      "loss": 0.7287,
+      "step": 27630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.647995450697338e-06,
+      "loss": 0.898,
+      "step": 27631
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.64733684169199e-06,
+      "loss": 0.962,
+      "step": 27632
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.646678278883912e-06,
+      "loss": 1.0281,
+      "step": 27633
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.646019762277896e-06,
+      "loss": 1.0675,
+      "step": 27634
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.645361291878726e-06,
+      "loss": 0.9293,
+      "step": 27635
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6447028676911987e-06,
+      "loss": 0.8468,
+      "step": 27636
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.644044489720101e-06,
+      "loss": 1.06,
+      "step": 27637
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6433861579702146e-06,
+      "loss": 0.9284,
+      "step": 27638
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.642727872446329e-06,
+      "loss": 1.249,
+      "step": 27639
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.642069633153229e-06,
+      "loss": 1.0131,
+      "step": 27640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.641411440095708e-06,
+      "loss": 1.0095,
+      "step": 27641
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6407532932785496e-06,
+      "loss": 0.8021,
+      "step": 27642
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.64009519270654e-06,
+      "loss": 0.8062,
+      "step": 27643
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.639437138384463e-06,
+      "loss": 0.6308,
+      "step": 27644
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.638779130317106e-06,
+      "loss": 0.8123,
+      "step": 27645
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6381211685092532e-06,
+      "loss": 0.7489,
+      "step": 27646
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6374632529656904e-06,
+      "loss": 0.7979,
+      "step": 27647
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.636805383691201e-06,
+      "loss": 0.7212,
+      "step": 27648
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6361475606905707e-06,
+      "loss": 0.6936,
+      "step": 27649
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6354897839685797e-06,
+      "loss": 0.7767,
+      "step": 27650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6348320535300152e-06,
+      "loss": 0.8232,
+      "step": 27651
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6341743693796596e-06,
+      "loss": 0.5397,
+      "step": 27652
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6335167315222897e-06,
+      "loss": 0.5462,
+      "step": 27653
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.632859139962698e-06,
+      "loss": 0.5359,
+      "step": 27654
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6322015947056612e-06,
+      "loss": 0.9606,
+      "step": 27655
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.631544095755961e-06,
+      "loss": 1.0582,
+      "step": 27656
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.630886643118383e-06,
+      "loss": 0.8215,
+      "step": 27657
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6302292367976977e-06,
+      "loss": 0.7503,
+      "step": 27658
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6295718767986966e-06,
+      "loss": 0.9972,
+      "step": 27659
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.628914563126156e-06,
+      "loss": 0.6438,
+      "step": 27660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.628257295784856e-06,
+      "loss": 0.9878,
+      "step": 27661
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6276000747795768e-06,
+      "loss": 0.8453,
+      "step": 27662
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6269429001150967e-06,
+      "loss": 1.0322,
+      "step": 27663
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6262857717961962e-06,
+      "loss": 0.8033,
+      "step": 27664
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6256286898276506e-06,
+      "loss": 0.9209,
+      "step": 27665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6249716542142433e-06,
+      "loss": 0.7553,
+      "step": 27666
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6243146649607543e-06,
+      "loss": 0.9907,
+      "step": 27667
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6236577220719536e-06,
+      "loss": 0.9223,
+      "step": 27668
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.623000825552622e-06,
+      "loss": 0.7944,
+      "step": 27669
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6223439754075363e-06,
+      "loss": 1.0428,
+      "step": 27670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.621687171641469e-06,
+      "loss": 0.7042,
+      "step": 27671
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6210304142592058e-06,
+      "loss": 1.0555,
+      "step": 27672
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6203737032655174e-06,
+      "loss": 0.703,
+      "step": 27673
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6197170386651792e-06,
+      "loss": 0.8202,
+      "step": 27674
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6190604204629685e-06,
+      "loss": 0.7884,
+      "step": 27675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6184038486636574e-06,
+      "loss": 0.873,
+      "step": 27676
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.617747323272024e-06,
+      "loss": 1.0843,
+      "step": 27677
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.61709084429284e-06,
+      "loss": 0.8365,
+      "step": 27678
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.616434411730879e-06,
+      "loss": 0.5591,
+      "step": 27679
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.615778025590917e-06,
+      "loss": 0.8326,
+      "step": 27680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.615121685877726e-06,
+      "loss": 0.9613,
+      "step": 27681
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6144653925960792e-06,
+      "loss": 0.8847,
+      "step": 27682
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.613809145750746e-06,
+      "loss": 0.9241,
+      "step": 27683
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6131529453465054e-06,
+      "loss": 0.8866,
+      "step": 27684
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.612496791388125e-06,
+      "loss": 0.9817,
+      "step": 27685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.611840683880378e-06,
+      "loss": 1.0461,
+      "step": 27686
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.611184622828038e-06,
+      "loss": 0.7678,
+      "step": 27687
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6105286082358695e-06,
+      "loss": 0.6616,
+      "step": 27688
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6098726401086428e-06,
+      "loss": 0.9446,
+      "step": 27689
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6092167184511352e-06,
+      "loss": 0.7272,
+      "step": 27690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.608560843268114e-06,
+      "loss": 0.7249,
+      "step": 27691
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6079050145643467e-06,
+      "loss": 0.8413,
+      "step": 27692
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.607249232344605e-06,
+      "loss": 0.8297,
+      "step": 27693
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.606593496613657e-06,
+      "loss": 0.5928,
+      "step": 27694
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6059378073762708e-06,
+      "loss": 0.7502,
+      "step": 27695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6052821646372116e-06,
+      "loss": 0.9481,
+      "step": 27696
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.604626568401258e-06,
+      "loss": 0.7692,
+      "step": 27697
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6039710186731657e-06,
+      "loss": 0.6667,
+      "step": 27698
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.603315515457707e-06,
+      "loss": 0.7378,
+      "step": 27699
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6026600587596484e-06,
+      "loss": 0.758,
+      "step": 27700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.602004648583752e-06,
+      "loss": 0.7791,
+      "step": 27701
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.601349284934792e-06,
+      "loss": 0.8276,
+      "step": 27702
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.600693967817531e-06,
+      "loss": 0.7313,
+      "step": 27703
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.600038697236734e-06,
+      "loss": 0.9178,
+      "step": 27704
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5993834731971654e-06,
+      "loss": 0.9266,
+      "step": 27705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.598728295703592e-06,
+      "loss": 0.9478,
+      "step": 27706
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.598073164760777e-06,
+      "loss": 0.4869,
+      "step": 27707
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5974180803734848e-06,
+      "loss": 0.7948,
+      "step": 27708
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5967630425464794e-06,
+      "loss": 0.9768,
+      "step": 27709
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.596108051284525e-06,
+      "loss": 0.602,
+      "step": 27710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5954531065923827e-06,
+      "loss": 0.96,
+      "step": 27711
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5947982084748167e-06,
+      "loss": 1.0028,
+      "step": 27712
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5941433569365903e-06,
+      "loss": 0.7807,
+      "step": 27713
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5934885519824604e-06,
+      "loss": 1.0915,
+      "step": 27714
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5928337936171976e-06,
+      "loss": 1.0025,
+      "step": 27715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.592179081845558e-06,
+      "loss": 0.6763,
+      "step": 27716
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5915244166723073e-06,
+      "loss": 0.5333,
+      "step": 27717
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.590869798102199e-06,
+      "loss": 1.2823,
+      "step": 27718
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5902152261399946e-06,
+      "loss": 0.8042,
+      "step": 27719
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5895607007904597e-06,
+      "loss": 0.9839,
+      "step": 27720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5889062220583517e-06,
+      "loss": 0.9648,
+      "step": 27721
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.588251789948429e-06,
+      "loss": 0.8418,
+      "step": 27722
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5875974044654515e-06,
+      "loss": 0.7484,
+      "step": 27723
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.586943065614179e-06,
+      "loss": 0.7732,
+      "step": 27724
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.586288773399367e-06,
+      "loss": 1.1098,
+      "step": 27725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.585634527825772e-06,
+      "loss": 0.9346,
+      "step": 27726
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5849803288981634e-06,
+      "loss": 0.7032,
+      "step": 27727
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5843261766212858e-06,
+      "loss": 0.7901,
+      "step": 27728
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5836720709998997e-06,
+      "loss": 0.867,
+      "step": 27729
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5830180120387636e-06,
+      "loss": 0.8886,
+      "step": 27730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.582363999742634e-06,
+      "loss": 0.6806,
+      "step": 27731
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5817100341162613e-06,
+      "loss": 0.9549,
+      "step": 27732
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5810561151644096e-06,
+      "loss": 1.2668,
+      "step": 27733
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5804022428918316e-06,
+      "loss": 0.871,
+      "step": 27734
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5797484173032806e-06,
+      "loss": 0.871,
+      "step": 27735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.579094638403512e-06,
+      "loss": 0.8564,
+      "step": 27736
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.57844090619728e-06,
+      "loss": 0.8853,
+      "step": 27737
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5777872206893392e-06,
+      "loss": 0.8122,
+      "step": 27738
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.577133581884443e-06,
+      "loss": 0.6918,
+      "step": 27739
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5764799897873448e-06,
+      "loss": 0.7947,
+      "step": 27740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.575826444402798e-06,
+      "loss": 0.6676,
+      "step": 27741
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5751729457355533e-06,
+      "loss": 0.9101,
+      "step": 27742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.574519493790365e-06,
+      "loss": 0.9763,
+      "step": 27743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.573866088571981e-06,
+      "loss": 0.8346,
+      "step": 27744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5732127300851604e-06,
+      "loss": 0.7295,
+      "step": 27745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.572559418334649e-06,
+      "loss": 0.8557,
+      "step": 27746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.571906153325204e-06,
+      "loss": 0.5944,
+      "step": 27747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5712529350615675e-06,
+      "loss": 0.9926,
+      "step": 27748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5705997635484935e-06,
+      "loss": 0.7499,
+      "step": 27749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.569946638790729e-06,
+      "loss": 1.0551,
+      "step": 27750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.569293560793029e-06,
+      "loss": 0.6071,
+      "step": 27751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.568640529560141e-06,
+      "loss": 0.7338,
+      "step": 27752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5679875450968138e-06,
+      "loss": 0.8443,
+      "step": 27753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.567334607407795e-06,
+      "loss": 0.8301,
+      "step": 27754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.566681716497833e-06,
+      "loss": 0.876,
+      "step": 27755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5660288723716763e-06,
+      "loss": 0.8375,
+      "step": 27756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5653760750340714e-06,
+      "loss": 0.8616,
+      "step": 27757
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5647233244897674e-06,
+      "loss": 0.7527,
+      "step": 27758
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.564070620743508e-06,
+      "loss": 0.8917,
+      "step": 27759
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5634179638000443e-06,
+      "loss": 1.0112,
+      "step": 27760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5627653536641183e-06,
+      "loss": 0.9087,
+      "step": 27761
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.562112790340478e-06,
+      "loss": 0.5099,
+      "step": 27762
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5614602738338645e-06,
+      "loss": 1.0687,
+      "step": 27763
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5608078041490314e-06,
+      "loss": 0.6921,
+      "step": 27764
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5601553812907174e-06,
+      "loss": 1.055,
+      "step": 27765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5595030052636705e-06,
+      "loss": 0.6578,
+      "step": 27766
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.558850676072635e-06,
+      "loss": 0.6629,
+      "step": 27767
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5581983937223464e-06,
+      "loss": 0.7891,
+      "step": 27768
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5575461582175576e-06,
+      "loss": 0.7124,
+      "step": 27769
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.556893969563008e-06,
+      "loss": 0.7608,
+      "step": 27770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5562418277634415e-06,
+      "loss": 0.7775,
+      "step": 27771
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5555897328236e-06,
+      "loss": 0.7428,
+      "step": 27772
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5549376847482252e-06,
+      "loss": 0.8418,
+      "step": 27773
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5542856835420603e-06,
+      "loss": 0.8882,
+      "step": 27774
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.553633729209841e-06,
+      "loss": 0.8546,
+      "step": 27775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.552981821756317e-06,
+      "loss": 0.95,
+      "step": 27776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5523299611862295e-06,
+      "loss": 0.7588,
+      "step": 27777
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5516781475043093e-06,
+      "loss": 0.9147,
+      "step": 27778
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.551026380715303e-06,
+      "loss": 0.7521,
+      "step": 27779
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5503746608239487e-06,
+      "loss": 0.7831,
+      "step": 27780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5497229878349828e-06,
+      "loss": 0.9259,
+      "step": 27781
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.54907136175315e-06,
+      "loss": 1.0384,
+      "step": 27782
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.548419782583188e-06,
+      "loss": 1.0309,
+      "step": 27783
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.547768250329833e-06,
+      "loss": 0.9159,
+      "step": 27784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5471167649978244e-06,
+      "loss": 0.7155,
+      "step": 27785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.546465326591899e-06,
+      "loss": 0.7451,
+      "step": 27786
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5458139351167942e-06,
+      "loss": 1.0622,
+      "step": 27787
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.545162590577247e-06,
+      "loss": 0.9896,
+      "step": 27788
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5445112929779944e-06,
+      "loss": 0.7708,
+      "step": 27789
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.543860042323772e-06,
+      "loss": 0.9069,
+      "step": 27790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5432088386193173e-06,
+      "loss": 0.7824,
+      "step": 27791
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5425576818693653e-06,
+      "loss": 0.5862,
+      "step": 27792
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.541906572078646e-06,
+      "loss": 0.888,
+      "step": 27793
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.541255509251904e-06,
+      "loss": 0.651,
+      "step": 27794
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5406044933938688e-06,
+      "loss": 0.632,
+      "step": 27795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5399535245092743e-06,
+      "loss": 0.8597,
+      "step": 27796
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.539302602602859e-06,
+      "loss": 1.2389,
+      "step": 27797
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5386517276793497e-06,
+      "loss": 0.6706,
+      "step": 27798
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5380008997434788e-06,
+      "loss": 0.7671,
+      "step": 27799
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.537350118799986e-06,
+      "loss": 1.016,
+      "step": 27800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5366993848536015e-06,
+      "loss": 0.7405,
+      "step": 27801
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5360486979090557e-06,
+      "loss": 0.6883,
+      "step": 27802
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5353980579710823e-06,
+      "loss": 0.8869,
+      "step": 27803
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5347474650444124e-06,
+      "loss": 1.2088,
+      "step": 27804
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.534096919133776e-06,
+      "loss": 0.6686,
+      "step": 27805
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5334464202439025e-06,
+      "loss": 1.142,
+      "step": 27806
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.532795968379531e-06,
+      "loss": 0.7415,
+      "step": 27807
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5321455635453815e-06,
+      "loss": 1.0216,
+      "step": 27808
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.531495205746188e-06,
+      "loss": 1.0293,
+      "step": 27809
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5308448949866805e-06,
+      "loss": 0.7076,
+      "step": 27810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.530194631271583e-06,
+      "loss": 0.8001,
+      "step": 27811
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.529544414605632e-06,
+      "loss": 0.7784,
+      "step": 27812
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.528894244993553e-06,
+      "loss": 0.6095,
+      "step": 27813
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.528244122440073e-06,
+      "loss": 0.8063,
+      "step": 27814
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5275940469499213e-06,
+      "loss": 0.7728,
+      "step": 27815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5269440185278238e-06,
+      "loss": 0.6201,
+      "step": 27816
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5262940371785083e-06,
+      "loss": 0.8082,
+      "step": 27817
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5256441029067025e-06,
+      "loss": 0.9299,
+      "step": 27818
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5249942157171303e-06,
+      "loss": 0.9835,
+      "step": 27819
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5243443756145203e-06,
+      "loss": 0.7507,
+      "step": 27820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5236945826035964e-06,
+      "loss": 0.8095,
+      "step": 27821
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5230448366890846e-06,
+      "loss": 0.8081,
+      "step": 27822
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5223951378757092e-06,
+      "loss": 0.8894,
+      "step": 27823
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.521745486168194e-06,
+      "loss": 1.0756,
+      "step": 27824
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5210958815712672e-06,
+      "loss": 0.9507,
+      "step": 27825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5204463240896514e-06,
+      "loss": 0.8956,
+      "step": 27826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5197968137280726e-06,
+      "loss": 0.5976,
+      "step": 27827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5191473504912477e-06,
+      "loss": 0.9752,
+      "step": 27828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.518497934383899e-06,
+      "loss": 0.9121,
+      "step": 27829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5178485654107575e-06,
+      "loss": 0.9134,
+      "step": 27830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.51719924357654e-06,
+      "loss": 0.9238,
+      "step": 27831
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5165499688859695e-06,
+      "loss": 0.965,
+      "step": 27832
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5159007413437684e-06,
+      "loss": 0.8159,
+      "step": 27833
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5152515609546566e-06,
+      "loss": 0.8869,
+      "step": 27834
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5146024277233558e-06,
+      "loss": 1.0313,
+      "step": 27835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.513953341654582e-06,
+      "loss": 0.8599,
+      "step": 27836
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.513304302753068e-06,
+      "loss": 0.753,
+      "step": 27837
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5126553110235217e-06,
+      "loss": 0.7784,
+      "step": 27838
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.512006366470666e-06,
+      "loss": 0.7952,
+      "step": 27839
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5113574690992203e-06,
+      "loss": 0.6647,
+      "step": 27840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.510708618913904e-06,
+      "loss": 0.8128,
+      "step": 27841
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.510059815919432e-06,
+      "loss": 0.7471,
+      "step": 27842
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.509411060120529e-06,
+      "loss": 0.8478,
+      "step": 27843
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.508762351521909e-06,
+      "loss": 0.9175,
+      "step": 27844
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.508113690128291e-06,
+      "loss": 0.8963,
+      "step": 27845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.507465075944391e-06,
+      "loss": 0.5275,
+      "step": 27846
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.506816508974925e-06,
+      "loss": 0.9045,
+      "step": 27847
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5061679892246113e-06,
+      "loss": 0.7074,
+      "step": 27848
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.505519516698165e-06,
+      "loss": 0.6594,
+      "step": 27849
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5048710914003014e-06,
+      "loss": 0.84,
+      "step": 27850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5042227133357363e-06,
+      "loss": 0.9606,
+      "step": 27851
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.503574382509186e-06,
+      "loss": 0.7989,
+      "step": 27852
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.502926098925363e-06,
+      "loss": 1.0297,
+      "step": 27853
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.502277862588981e-06,
+      "loss": 0.9692,
+      "step": 27854
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5016296735047584e-06,
+      "loss": 0.8683,
+      "step": 27855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.50098153167741e-06,
+      "loss": 0.7781,
+      "step": 27856
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5003334371116417e-06,
+      "loss": 0.7227,
+      "step": 27857
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4996853898121707e-06,
+      "loss": 0.9737,
+      "step": 27858
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4990373897837095e-06,
+      "loss": 0.6915,
+      "step": 27859
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4983894370309666e-06,
+      "loss": 0.707,
+      "step": 27860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4977415315586614e-06,
+      "loss": 0.6943,
+      "step": 27861
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.497093673371502e-06,
+      "loss": 0.8813,
+      "step": 27862
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4964458624742e-06,
+      "loss": 0.6528,
+      "step": 27863
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4957980988714658e-06,
+      "loss": 0.748,
+      "step": 27864
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4951503825680087e-06,
+      "loss": 1.0691,
+      "step": 27865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.494502713568542e-06,
+      "loss": 1.0638,
+      "step": 27866
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4938550918777734e-06,
+      "loss": 0.7326,
+      "step": 27867
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.493207517500412e-06,
+      "loss": 0.8337,
+      "step": 27868
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.49255999044117e-06,
+      "loss": 1.1012,
+      "step": 27869
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4919125107047537e-06,
+      "loss": 0.9745,
+      "step": 27870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4912650782958712e-06,
+      "loss": 0.8979,
+      "step": 27871
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4906176932192293e-06,
+      "loss": 0.7314,
+      "step": 27872
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4899703554795415e-06,
+      "loss": 0.8142,
+      "step": 27873
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4893230650815124e-06,
+      "loss": 0.6635,
+      "step": 27874
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.48867582202985e-06,
+      "loss": 0.7679,
+      "step": 27875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.488028626329262e-06,
+      "loss": 0.7545,
+      "step": 27876
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.487381477984451e-06,
+      "loss": 0.8784,
+      "step": 27877
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4867343770001195e-06,
+      "loss": 0.5793,
+      "step": 27878
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4860873233809842e-06,
+      "loss": 0.8783,
+      "step": 27879
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4854403171317443e-06,
+      "loss": 0.8935,
+      "step": 27880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4847933582571058e-06,
+      "loss": 0.7556,
+      "step": 27881
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.484146446761775e-06,
+      "loss": 0.8493,
+      "step": 27882
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4834995826504524e-06,
+      "loss": 0.8834,
+      "step": 27883
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4828527659278465e-06,
+      "loss": 0.7905,
+      "step": 27884
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.482205996598654e-06,
+      "loss": 0.8492,
+      "step": 27885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.481559274667592e-06,
+      "loss": 0.7287,
+      "step": 27886
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.48091260013935e-06,
+      "loss": 0.953,
+      "step": 27887
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.480265973018636e-06,
+      "loss": 0.5081,
+      "step": 27888
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4796193933101517e-06,
+      "loss": 0.8049,
+      "step": 27889
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478972861018596e-06,
+      "loss": 1.0282,
+      "step": 27890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4783263761486765e-06,
+      "loss": 1.1176,
+      "step": 27891
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.477679938705092e-06,
+      "loss": 0.7019,
+      "step": 27892
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4770335486925444e-06,
+      "loss": 1.0434,
+      "step": 27893
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.476387206115732e-06,
+      "loss": 0.9492,
+      "step": 27894
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.475740910979356e-06,
+      "loss": 0.7834,
+      "step": 27895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4750946632881176e-06,
+      "loss": 0.8909,
+      "step": 27896
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4744484630467146e-06,
+      "loss": 1.0313,
+      "step": 27897
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4738023102598483e-06,
+      "loss": 0.9143,
+      "step": 27898
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4731562049322152e-06,
+      "loss": 0.7953,
+      "step": 27899
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.472510147068515e-06,
+      "loss": 1.0836,
+      "step": 27900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4718641366734463e-06,
+      "loss": 0.9452,
+      "step": 27901
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.471218173751706e-06,
+      "loss": 0.8484,
+      "step": 27902
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4705722583079895e-06,
+      "loss": 0.9919,
+      "step": 27903
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.469926390346999e-06,
+      "loss": 0.9476,
+      "step": 27904
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.46928056987343e-06,
+      "loss": 1.0746,
+      "step": 27905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.468634796891981e-06,
+      "loss": 1.0182,
+      "step": 27906
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4679890714073415e-06,
+      "loss": 0.8115,
+      "step": 27907
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4673433934242072e-06,
+      "loss": 0.9901,
+      "step": 27908
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.466697762947281e-06,
+      "loss": 0.8382,
+      "step": 27909
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4660521799812542e-06,
+      "loss": 0.8121,
+      "step": 27910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.465406644530821e-06,
+      "loss": 0.7822,
+      "step": 27911
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4647611566006768e-06,
+      "loss": 1.0332,
+      "step": 27912
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.464115716195514e-06,
+      "loss": 1.0104,
+      "step": 27913
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4634703233200294e-06,
+      "loss": 0.7793,
+      "step": 27914
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4628249779789105e-06,
+      "loss": 0.8601,
+      "step": 27915
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4621796801768605e-06,
+      "loss": 1.1143,
+      "step": 27916
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.461534429918563e-06,
+      "loss": 0.6621,
+      "step": 27917
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4608892272087125e-06,
+      "loss": 0.9129,
+      "step": 27918
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4602440720520024e-06,
+      "loss": 0.7123,
+      "step": 27919
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4595989644531236e-06,
+      "loss": 0.7466,
+      "step": 27920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.458953904416763e-06,
+      "loss": 0.76,
+      "step": 27921
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4583088919476204e-06,
+      "loss": 0.5979,
+      "step": 27922
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.457663927050382e-06,
+      "loss": 1.105,
+      "step": 27923
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4570190097297374e-06,
+      "loss": 1.065,
+      "step": 27924
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4563741399903784e-06,
+      "loss": 0.927,
+      "step": 27925
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4557293178369924e-06,
+      "loss": 0.7081,
+      "step": 27926
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4550845432742685e-06,
+      "loss": 0.7762,
+      "step": 27927
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.454439816306898e-06,
+      "loss": 0.8094,
+      "step": 27928
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4537951369395683e-06,
+      "loss": 0.8129,
+      "step": 27929
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4531505051769665e-06,
+      "loss": 0.53,
+      "step": 27930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.452505921023781e-06,
+      "loss": 1.5552,
+      "step": 27931
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4518613844847002e-06,
+      "loss": 0.7806,
+      "step": 27932
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4512168955644065e-06,
+      "loss": 0.7687,
+      "step": 27933
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4505724542675958e-06,
+      "loss": 0.79,
+      "step": 27934
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.449928060598947e-06,
+      "loss": 0.6813,
+      "step": 27935
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4492837145631543e-06,
+      "loss": 0.9703,
+      "step": 27936
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.448639416164893e-06,
+      "loss": 0.9352,
+      "step": 27937
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.447995165408854e-06,
+      "loss": 0.9559,
+      "step": 27938
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.447350962299718e-06,
+      "loss": 0.7978,
+      "step": 27939
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.446706806842177e-06,
+      "loss": 1.0452,
+      "step": 27940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4460626990409128e-06,
+      "loss": 0.9835,
+      "step": 27941
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.445418638900607e-06,
+      "loss": 0.5333,
+      "step": 27942
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.444774626425945e-06,
+      "loss": 0.8438,
+      "step": 27943
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4441306616216107e-06,
+      "loss": 0.954,
+      "step": 27944
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4434867444922857e-06,
+      "loss": 0.8663,
+      "step": 27945
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4428428750426533e-06,
+      "loss": 0.7251,
+      "step": 27946
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4421990532773954e-06,
+      "loss": 1.0206,
+      "step": 27947
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.441555279201195e-06,
+      "loss": 1.0787,
+      "step": 27948
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.440911552818732e-06,
+      "loss": 0.8994,
+      "step": 27949
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4402678741346885e-06,
+      "loss": 0.9018,
+      "step": 27950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.439624243153743e-06,
+      "loss": 1.0085,
+      "step": 27951
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.438980659880581e-06,
+      "loss": 0.8867,
+      "step": 27952
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4383371243198805e-06,
+      "loss": 0.8867,
+      "step": 27953
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4376936364763213e-06,
+      "loss": 1.0786,
+      "step": 27954
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4370501963545822e-06,
+      "loss": 0.7334,
+      "step": 27955
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4364068039593466e-06,
+      "loss": 1.1146,
+      "step": 27956
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4357634592952836e-06,
+      "loss": 0.9422,
+      "step": 27957
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4351201623670804e-06,
+      "loss": 0.869,
+      "step": 27958
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.434476913179412e-06,
+      "loss": 0.7493,
+      "step": 27959
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.433833711736957e-06,
+      "loss": 0.8027,
+      "step": 27960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4331905580443924e-06,
+      "loss": 0.8403,
+      "step": 27961
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.432547452106395e-06,
+      "loss": 0.9434,
+      "step": 27962
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.431904393927642e-06,
+      "loss": 0.9752,
+      "step": 27963
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4312613835128063e-06,
+      "loss": 0.9687,
+      "step": 27964
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4306184208665703e-06,
+      "loss": 0.7937,
+      "step": 27965
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.42997550599361e-06,
+      "loss": 0.808,
+      "step": 27966
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4293326388985948e-06,
+      "loss": 0.7694,
+      "step": 27967
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4286898195862018e-06,
+      "loss": 0.8888,
+      "step": 27968
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.428047048061103e-06,
+      "loss": 0.9275,
+      "step": 27969
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.427404324327979e-06,
+      "loss": 0.81,
+      "step": 27970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4267616483915e-06,
+      "loss": 0.7422,
+      "step": 27971
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.426119020256341e-06,
+      "loss": 1.1519,
+      "step": 27972
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4254764399271744e-06,
+      "loss": 0.9423,
+      "step": 27973
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4248339074086736e-06,
+      "loss": 0.9696,
+      "step": 27974
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4241914227055096e-06,
+      "loss": 0.6377,
+      "step": 27975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.423548985822357e-06,
+      "loss": 0.7888,
+      "step": 27976
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.422906596763886e-06,
+      "loss": 0.6965,
+      "step": 27977
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4222642555347686e-06,
+      "loss": 0.952,
+      "step": 27978
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4216219621396764e-06,
+      "loss": 0.9556,
+      "step": 27979
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4209797165832792e-06,
+      "loss": 0.6031,
+      "step": 27980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.420337518870248e-06,
+      "loss": 0.7713,
+      "step": 27981
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4196953690052504e-06,
+      "loss": 0.7254,
+      "step": 27982
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.419053266992961e-06,
+      "loss": 0.7132,
+      "step": 27983
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4184112128380487e-06,
+      "loss": 0.6908,
+      "step": 27984
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4177692065451796e-06,
+      "loss": 0.5185,
+      "step": 27985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4171272481190276e-06,
+      "loss": 0.7492,
+      "step": 27986
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4164853375642505e-06,
+      "loss": 1.086,
+      "step": 27987
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4158434748855263e-06,
+      "loss": 0.917,
+      "step": 27988
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4152016600875193e-06,
+      "loss": 0.9144,
+      "step": 27989
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.414559893174898e-06,
+      "loss": 0.9531,
+      "step": 27990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4139181741523276e-06,
+      "loss": 0.5222,
+      "step": 27991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4132765030244763e-06,
+      "loss": 0.9227,
+      "step": 27992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4126348797960087e-06,
+      "loss": 0.7006,
+      "step": 27993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4119933044715894e-06,
+      "loss": 0.7842,
+      "step": 27994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.411351777055889e-06,
+      "loss": 0.9057,
+      "step": 27995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4107102975535735e-06,
+      "loss": 0.8848,
+      "step": 27996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4100688659693015e-06,
+      "loss": 0.7851,
+      "step": 27997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.40942748230774e-06,
+      "loss": 0.7181,
+      "step": 27998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.408786146573555e-06,
+      "loss": 0.9372,
+      "step": 27999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.408144858771404e-06,
+      "loss": 0.8268,
+      "step": 28000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4075036189059595e-06,
+      "loss": 0.8742,
+      "step": 28001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4068624269818817e-06,
+      "loss": 0.9127,
+      "step": 28002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4062212830038323e-06,
+      "loss": 0.7635,
+      "step": 28003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4055801869764727e-06,
+      "loss": 0.9736,
+      "step": 28004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4049391389044674e-06,
+      "loss": 0.5037,
+      "step": 28005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.404298138792477e-06,
+      "loss": 1.1294,
+      "step": 28006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4036571866451617e-06,
+      "loss": 1.2281,
+      "step": 28007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.403016282467185e-06,
+      "loss": 0.8798,
+      "step": 28008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.402375426263207e-06,
+      "loss": 0.9714,
+      "step": 28009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4017346180378865e-06,
+      "loss": 0.6193,
+      "step": 28010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4010938577958853e-06,
+      "loss": 0.9204,
+      "step": 28011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4004531455418576e-06,
+      "loss": 0.88,
+      "step": 28012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3998124812804724e-06,
+      "loss": 0.6955,
+      "step": 28013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3991718650163828e-06,
+      "loss": 0.751,
+      "step": 28014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3985312967542496e-06,
+      "loss": 0.6923,
+      "step": 28015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3978907764987323e-06,
+      "loss": 0.6569,
+      "step": 28016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3972503042544824e-06,
+      "loss": 0.7118,
+      "step": 28017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3966098800261594e-06,
+      "loss": 0.5876,
+      "step": 28018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.395969503818426e-06,
+      "loss": 0.7538,
+      "step": 28019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3953291756359354e-06,
+      "loss": 0.5757,
+      "step": 28020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3946888954833436e-06,
+      "loss": 0.7584,
+      "step": 28021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.394048663365308e-06,
+      "loss": 0.9068,
+      "step": 28022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.393408479286485e-06,
+      "loss": 0.8143,
+      "step": 28023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.392768343251529e-06,
+      "loss": 0.8654,
+      "step": 28024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3921282552650915e-06,
+      "loss": 0.7972,
+      "step": 28025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.391488215331837e-06,
+      "loss": 0.5786,
+      "step": 28026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.390848223456412e-06,
+      "loss": 0.9058,
+      "step": 28027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3902082796434734e-06,
+      "loss": 0.624,
+      "step": 28028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3895683838976723e-06,
+      "loss": 0.8348,
+      "step": 28029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.388928536223661e-06,
+      "loss": 0.8597,
+      "step": 28030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3882887366260997e-06,
+      "loss": 0.8219,
+      "step": 28031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.387648985109637e-06,
+      "loss": 0.7351,
+      "step": 28032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3870092816789245e-06,
+      "loss": 0.7406,
+      "step": 28033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3863696263386146e-06,
+      "loss": 0.8148,
+      "step": 28034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3857300190933606e-06,
+      "loss": 1.0078,
+      "step": 28035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3850904599478106e-06,
+      "loss": 0.87,
+      "step": 28036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.384450948906619e-06,
+      "loss": 0.624,
+      "step": 28037
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3838114859744332e-06,
+      "loss": 0.8427,
+      "step": 28038
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.383172071155906e-06,
+      "loss": 0.9743,
+      "step": 28039
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3825327044556866e-06,
+      "loss": 0.9259,
+      "step": 28040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.381893385878423e-06,
+      "loss": 1.2508,
+      "step": 28041
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3812541154287658e-06,
+      "loss": 0.9463,
+      "step": 28042
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3806148931113615e-06,
+      "loss": 0.8188,
+      "step": 28043
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3799757189308634e-06,
+      "loss": 0.8648,
+      "step": 28044
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3793365928919165e-06,
+      "loss": 0.9103,
+      "step": 28045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.378697514999173e-06,
+      "loss": 0.6292,
+      "step": 28046
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3780584852572727e-06,
+      "loss": 0.8813,
+      "step": 28047
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3774195036708647e-06,
+      "loss": 0.7078,
+      "step": 28048
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.376780570244599e-06,
+      "loss": 0.5533,
+      "step": 28049
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.376141684983121e-06,
+      "loss": 1.0194,
+      "step": 28050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.375502847891077e-06,
+      "loss": 0.9964,
+      "step": 28051
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3748640589731108e-06,
+      "loss": 1.0037,
+      "step": 28052
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.374225318233869e-06,
+      "loss": 0.8427,
+      "step": 28053
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3735866256779968e-06,
+      "loss": 0.8182,
+      "step": 28054
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.372947981310135e-06,
+      "loss": 0.8311,
+      "step": 28055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3723093851349386e-06,
+      "loss": 0.8504,
+      "step": 28056
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3716708371570406e-06,
+      "loss": 0.9197,
+      "step": 28057
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.371032337381088e-06,
+      "loss": 0.607,
+      "step": 28058
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3703938858117245e-06,
+      "loss": 0.8822,
+      "step": 28059
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3697554824535938e-06,
+      "loss": 0.72,
+      "step": 28060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3691171273113334e-06,
+      "loss": 0.7326,
+      "step": 28061
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3684788203895922e-06,
+      "loss": 0.6885,
+      "step": 28062
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.367840561693011e-06,
+      "loss": 1.0182,
+      "step": 28063
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.367202351226229e-06,
+      "loss": 0.7,
+      "step": 28064
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.366564188993887e-06,
+      "loss": 0.6205,
+      "step": 28065
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3659260750006286e-06,
+      "loss": 0.734,
+      "step": 28066
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.365288009251091e-06,
+      "loss": 0.9204,
+      "step": 28067
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3646499917499165e-06,
+      "loss": 0.8827,
+      "step": 28068
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3640120225017447e-06,
+      "loss": 0.8106,
+      "step": 28069
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.363374101511213e-06,
+      "loss": 0.7756,
+      "step": 28070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3627362287829622e-06,
+      "loss": 0.8081,
+      "step": 28071
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.362098404321632e-06,
+      "loss": 0.4973,
+      "step": 28072
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.361460628131854e-06,
+      "loss": 0.8884,
+      "step": 28073
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.360822900218276e-06,
+      "loss": 0.724,
+      "step": 28074
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.360185220585531e-06,
+      "loss": 0.9663,
+      "step": 28075
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.35954758923826e-06,
+      "loss": 0.8167,
+      "step": 28076
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.358910006181092e-06,
+      "loss": 0.8185,
+      "step": 28077
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3582724714186688e-06,
+      "loss": 0.6267,
+      "step": 28078
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3576349849556223e-06,
+      "loss": 0.7068,
+      "step": 28079
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3569975467965955e-06,
+      "loss": 0.6792,
+      "step": 28080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.356360156946219e-06,
+      "loss": 0.9009,
+      "step": 28081
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3557228154091302e-06,
+      "loss": 0.771,
+      "step": 28082
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3550855221899625e-06,
+      "loss": 1.0465,
+      "step": 28083
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3544482772933505e-06,
+      "loss": 0.8575,
+      "step": 28084
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3538110807239298e-06,
+      "loss": 0.5886,
+      "step": 28085
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3531739324863322e-06,
+      "loss": 0.7926,
+      "step": 28086
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3525368325851916e-06,
+      "loss": 0.6938,
+      "step": 28087
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.351899781025142e-06,
+      "loss": 1.3264,
+      "step": 28088
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.351262777810814e-06,
+      "loss": 0.5464,
+      "step": 28089
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3506258229468427e-06,
+      "loss": 0.874,
+      "step": 28090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.349988916437855e-06,
+      "loss": 0.8742,
+      "step": 28091
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.349352058288491e-06,
+      "loss": 0.7671,
+      "step": 28092
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3487152485033756e-06,
+      "loss": 0.6865,
+      "step": 28093
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.348078487087142e-06,
+      "loss": 0.7407,
+      "step": 28094
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.347441774044421e-06,
+      "loss": 0.7506,
+      "step": 28095
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.346805109379845e-06,
+      "loss": 0.9556,
+      "step": 28096
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.346168493098034e-06,
+      "loss": 0.6,
+      "step": 28097
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.345531925203629e-06,
+      "loss": 0.7177,
+      "step": 28098
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.344895405701254e-06,
+      "loss": 0.8604,
+      "step": 28099
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.344258934595539e-06,
+      "loss": 0.6835,
+      "step": 28100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3436225118911114e-06,
+      "loss": 0.7413,
+      "step": 28101
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3429861375926008e-06,
+      "loss": 0.8611,
+      "step": 28102
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3423498117046337e-06,
+      "loss": 0.6116,
+      "step": 28103
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3417135342318353e-06,
+      "loss": 0.695,
+      "step": 28104
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3410773051788394e-06,
+      "loss": 0.5236,
+      "step": 28105
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.340441124550271e-06,
+      "loss": 0.9426,
+      "step": 28106
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3398049923507514e-06,
+      "loss": 0.7316,
+      "step": 28107
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.339168908584909e-06,
+      "loss": 0.8049,
+      "step": 28108
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3385328732573676e-06,
+      "loss": 1.1127,
+      "step": 28109
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.337896886372757e-06,
+      "loss": 0.8534,
+      "step": 28110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.337260947935701e-06,
+      "loss": 1.0836,
+      "step": 28111
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.336625057950823e-06,
+      "loss": 0.7847,
+      "step": 28112
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.335989216422747e-06,
+      "loss": 0.7267,
+      "step": 28113
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.335353423356098e-06,
+      "loss": 0.9106,
+      "step": 28114
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.334717678755499e-06,
+      "loss": 0.733,
+      "step": 28115
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3340819826255733e-06,
+      "loss": 0.8526,
+      "step": 28116
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3334463349709436e-06,
+      "loss": 0.6363,
+      "step": 28117
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3328107357962315e-06,
+      "loss": 0.5875,
+      "step": 28118
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3321751851060614e-06,
+      "loss": 0.8686,
+      "step": 28119
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.331539682905053e-06,
+      "loss": 0.964,
+      "step": 28120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.330904229197829e-06,
+      "loss": 0.8503,
+      "step": 28121
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.330268823989007e-06,
+      "loss": 0.7178,
+      "step": 28122
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3296334672832133e-06,
+      "loss": 0.7245,
+      "step": 28123
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3289981590850663e-06,
+      "loss": 0.8496,
+      "step": 28124
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3283628993991846e-06,
+      "loss": 0.8001,
+      "step": 28125
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3277276882301935e-06,
+      "loss": 1.1,
+      "step": 28126
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3270925255827004e-06,
+      "loss": 0.7503,
+      "step": 28127
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3264574114613344e-06,
+      "loss": 0.4416,
+      "step": 28128
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3258223458707105e-06,
+      "loss": 1.0755,
+      "step": 28129
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.325187328815449e-06,
+      "loss": 0.5477,
+      "step": 28130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3245523603001663e-06,
+      "loss": 0.5076,
+      "step": 28131
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3239174403294795e-06,
+      "loss": 0.9453,
+      "step": 28132
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3232825689080073e-06,
+      "loss": 0.8794,
+      "step": 28133
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3226477460403615e-06,
+      "loss": 1.0887,
+      "step": 28134
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3220129717311667e-06,
+      "loss": 0.6297,
+      "step": 28135
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3213782459850373e-06,
+      "loss": 0.8463,
+      "step": 28136
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.320743568806585e-06,
+      "loss": 1.1003,
+      "step": 28137
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.320108940200427e-06,
+      "loss": 1.1286,
+      "step": 28138
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.319474360171179e-06,
+      "loss": 0.7786,
+      "step": 28139
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3188398287234504e-06,
+      "loss": 0.822,
+      "step": 28140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3182053458618646e-06,
+      "loss": 0.7938,
+      "step": 28141
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3175709115910315e-06,
+      "loss": 0.84,
+      "step": 28142
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316936525915565e-06,
+      "loss": 0.9369,
+      "step": 28143
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316302188840078e-06,
+      "loss": 0.8882,
+      "step": 28144
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.315667900369184e-06,
+      "loss": 0.957,
+      "step": 28145
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3150336605074953e-06,
+      "loss": 0.907,
+      "step": 28146
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3143994692596237e-06,
+      "loss": 0.8437,
+      "step": 28147
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3137653266301817e-06,
+      "loss": 0.8299,
+      "step": 28148
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3131312326237808e-06,
+      "loss": 0.8874,
+      "step": 28149
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3124971872450328e-06,
+      "loss": 0.9442,
+      "step": 28150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3118631904985476e-06,
+      "loss": 0.8137,
+      "step": 28151
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3112292423889327e-06,
+      "loss": 0.7949,
+      "step": 28152
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.310595342920804e-06,
+      "loss": 1.0772,
+      "step": 28153
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.309961492098769e-06,
+      "loss": 0.8399,
+      "step": 28154
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3093276899274373e-06,
+      "loss": 0.4626,
+      "step": 28155
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.308693936411421e-06,
+      "loss": 0.825,
+      "step": 28156
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3080602315553224e-06,
+      "loss": 0.971,
+      "step": 28157
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3074265753637492e-06,
+      "loss": 0.6512,
+      "step": 28158
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3067929678413157e-06,
+      "loss": 0.8813,
+      "step": 28159
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3061594089926285e-06,
+      "loss": 0.9681,
+      "step": 28160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.305525898822293e-06,
+      "loss": 0.7257,
+      "step": 28161
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3048924373349154e-06,
+      "loss": 0.8934,
+      "step": 28162
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.304259024535105e-06,
+      "loss": 0.8901,
+      "step": 28163
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3036256604274653e-06,
+      "loss": 0.7115,
+      "step": 28164
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3029923450166015e-06,
+      "loss": 0.8696,
+      "step": 28165
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3023590783071256e-06,
+      "loss": 0.7931,
+      "step": 28166
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.301725860303636e-06,
+      "loss": 0.9051,
+      "step": 28167
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.301092691010741e-06,
+      "loss": 0.7725,
+      "step": 28168
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3004595704330423e-06,
+      "loss": 0.6311,
+      "step": 28169
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2998264985751425e-06,
+      "loss": 0.913,
+      "step": 28170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.299193475441652e-06,
+      "loss": 0.8876,
+      "step": 28171
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2985605010371714e-06,
+      "loss": 0.7021,
+      "step": 28172
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.297927575366302e-06,
+      "loss": 0.8807,
+      "step": 28173
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.297294698433647e-06,
+      "loss": 0.789,
+      "step": 28174
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.29666187024381e-06,
+      "loss": 0.8375,
+      "step": 28175
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2960290908013924e-06,
+      "loss": 0.8324,
+      "step": 28176
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2953963601109962e-06,
+      "loss": 0.8446,
+      "step": 28177
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2947636781772217e-06,
+      "loss": 0.9845,
+      "step": 28178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2941310450046703e-06,
+      "loss": 0.6788,
+      "step": 28179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2934984605979424e-06,
+      "loss": 0.5994,
+      "step": 28180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.292865924961639e-06,
+      "loss": 0.849,
+      "step": 28181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2922334381003585e-06,
+      "loss": 0.695,
+      "step": 28182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.291601000018698e-06,
+      "loss": 0.6784,
+      "step": 28183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2909686107212644e-06,
+      "loss": 0.9264,
+      "step": 28184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2903362702126516e-06,
+      "loss": 0.703,
+      "step": 28185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2897039784974615e-06,
+      "loss": 0.8367,
+      "step": 28186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.289071735580286e-06,
+      "loss": 0.8206,
+      "step": 28187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.288439541465723e-06,
+      "loss": 0.7841,
+      "step": 28188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2878073961583768e-06,
+      "loss": 0.6973,
+      "step": 28189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.287175299662839e-06,
+      "loss": 0.8022,
+      "step": 28190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2865432519837094e-06,
+      "loss": 0.8782,
+      "step": 28191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.285911253125582e-06,
+      "loss": 1.0432,
+      "step": 28192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2852793030930543e-06,
+      "loss": 0.8161,
+      "step": 28193
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2846474018907205e-06,
+      "loss": 0.711,
+      "step": 28194
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.284015549523173e-06,
+      "loss": 0.5426,
+      "step": 28195
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.283383745995017e-06,
+      "loss": 0.7125,
+      "step": 28196
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2827519913108375e-06,
+      "loss": 0.9632,
+      "step": 28197
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.282120285475231e-06,
+      "loss": 0.8449,
+      "step": 28198
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2814886284927916e-06,
+      "loss": 0.8801,
+      "step": 28199
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2808570203681135e-06,
+      "loss": 0.6778,
+      "step": 28200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.280225461105785e-06,
+      "loss": 0.645,
+      "step": 28201
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2795939507104067e-06,
+      "loss": 0.9937,
+      "step": 28202
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.278962489186569e-06,
+      "loss": 0.8619,
+      "step": 28203
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.278331076538861e-06,
+      "loss": 1.0622,
+      "step": 28204
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.277699712771876e-06,
+      "loss": 0.5435,
+      "step": 28205
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2770683978902053e-06,
+      "loss": 0.7549,
+      "step": 28206
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.27643713189844e-06,
+      "loss": 0.8716,
+      "step": 28207
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.27580591480117e-06,
+      "loss": 0.8311,
+      "step": 28208
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.275174746602987e-06,
+      "loss": 0.9736,
+      "step": 28209
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2745436273084798e-06,
+      "loss": 0.9414,
+      "step": 28210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2739125569222375e-06,
+      "loss": 0.8075,
+      "step": 28211
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.27328153544885e-06,
+      "loss": 0.7923,
+      "step": 28212
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.272650562892904e-06,
+      "loss": 0.7862,
+      "step": 28213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.272019639258992e-06,
+      "loss": 0.9771,
+      "step": 28214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.271388764551702e-06,
+      "loss": 0.8545,
+      "step": 28215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2707579387756215e-06,
+      "loss": 0.7421,
+      "step": 28216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.270127161935335e-06,
+      "loss": 0.651,
+      "step": 28217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2694964340354295e-06,
+      "loss": 0.9205,
+      "step": 28218
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2688657550804903e-06,
+      "loss": 0.6336,
+      "step": 28219
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.268235125075111e-06,
+      "loss": 0.78,
+      "step": 28220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2676045440238734e-06,
+      "loss": 0.9201,
+      "step": 28221
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.266974011931362e-06,
+      "loss": 0.6327,
+      "step": 28222
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.266343528802164e-06,
+      "loss": 0.8066,
+      "step": 28223
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2657130946408645e-06,
+      "loss": 0.9268,
+      "step": 28224
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2650827094520455e-06,
+      "loss": 0.8146,
+      "step": 28225
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2644523732402945e-06,
+      "loss": 0.7186,
+      "step": 28226
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.263822086010193e-06,
+      "loss": 1.0046,
+      "step": 28227
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.263191847766325e-06,
+      "loss": 0.8148,
+      "step": 28228
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.262561658513275e-06,
+      "loss": 0.8162,
+      "step": 28229
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2619315182556234e-06,
+      "loss": 0.9556,
+      "step": 28230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2613014269979517e-06,
+      "loss": 0.8763,
+      "step": 28231
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.260671384744848e-06,
+      "loss": 1.145,
+      "step": 28232
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.26004139150089e-06,
+      "loss": 0.6535,
+      "step": 28233
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2594114472706586e-06,
+      "loss": 0.9694,
+      "step": 28234
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2587815520587363e-06,
+      "loss": 1.0023,
+      "step": 28235
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2581517058697064e-06,
+      "loss": 0.896,
+      "step": 28236
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.25752190870814e-06,
+      "loss": 0.9522,
+      "step": 28237
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.256892160578625e-06,
+      "loss": 0.96,
+      "step": 28238
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.256262461485741e-06,
+      "loss": 0.9247,
+      "step": 28239
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.255632811434064e-06,
+      "loss": 0.8133,
+      "step": 28240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2550032104281736e-06,
+      "loss": 1.189,
+      "step": 28241
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.254373658472649e-06,
+      "loss": 0.8652,
+      "step": 28242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2537441555720696e-06,
+      "loss": 0.8097,
+      "step": 28243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2531147017310073e-06,
+      "loss": 0.7863,
+      "step": 28244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2524852969540477e-06,
+      "loss": 0.796,
+      "step": 28245
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2518559412457686e-06,
+      "loss": 0.6955,
+      "step": 28246
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.251226634610738e-06,
+      "loss": 1.0974,
+      "step": 28247
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.250597377053537e-06,
+      "loss": 0.8455,
+      "step": 28248
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2499681685787378e-06,
+      "loss": 0.5577,
+      "step": 28249
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.249339009190923e-06,
+      "loss": 0.7313,
+      "step": 28250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2487098988946655e-06,
+      "loss": 0.9118,
+      "step": 28251
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2480808376945383e-06,
+      "loss": 0.9693,
+      "step": 28252
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.247451825595117e-06,
+      "loss": 1.0502,
+      "step": 28253
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2468228626009768e-06,
+      "loss": 0.7829,
+      "step": 28254
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2461939487166894e-06,
+      "loss": 1.0061,
+      "step": 28255
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.24556508394683e-06,
+      "loss": 0.6159,
+      "step": 28256
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.244936268295972e-06,
+      "loss": 1.0201,
+      "step": 28257
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2443075017686875e-06,
+      "loss": 0.9971,
+      "step": 28258
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2436787843695483e-06,
+      "loss": 0.7456,
+      "step": 28259
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.243050116103128e-06,
+      "loss": 0.5338,
+      "step": 28260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2424214969739974e-06,
+      "loss": 0.6786,
+      "step": 28261
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2417929269867256e-06,
+      "loss": 0.9891,
+      "step": 28262
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.241164406145888e-06,
+      "loss": 0.8388,
+      "step": 28263
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.240535934456054e-06,
+      "loss": 0.6107,
+      "step": 28264
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.239907511921793e-06,
+      "loss": 0.7286,
+      "step": 28265
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2392791385476784e-06,
+      "loss": 0.814,
+      "step": 28266
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2386508143382746e-06,
+      "loss": 1.0918,
+      "step": 28267
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.238022539298149e-06,
+      "loss": 0.9255,
+      "step": 28268
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2373943134318777e-06,
+      "loss": 1.0302,
+      "step": 28269
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.236766136744025e-06,
+      "loss": 0.9732,
+      "step": 28270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.23613800923916e-06,
+      "loss": 0.7765,
+      "step": 28271
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.235509930921852e-06,
+      "loss": 0.8585,
+      "step": 28272
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.234881901796666e-06,
+      "loss": 0.8338,
+      "step": 28273
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.23425392186817e-06,
+      "loss": 0.4807,
+      "step": 28274
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2336259911409283e-06,
+      "loss": 0.9206,
+      "step": 28275
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2329981096195153e-06,
+      "loss": 0.741,
+      "step": 28276
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.232370277308489e-06,
+      "loss": 0.7242,
+      "step": 28277
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2317424942124178e-06,
+      "loss": 0.6781,
+      "step": 28278
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.231114760335866e-06,
+      "loss": 0.7605,
+      "step": 28279
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.230487075683396e-06,
+      "loss": 1.0024,
+      "step": 28280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2298594402595796e-06,
+      "loss": 1.0176,
+      "step": 28281
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.229231854068977e-06,
+      "loss": 0.8889,
+      "step": 28282
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2286043171161517e-06,
+      "loss": 0.8201,
+      "step": 28283
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2279768294056678e-06,
+      "loss": 0.7331,
+      "step": 28284
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2273493909420893e-06,
+      "loss": 0.8741,
+      "step": 28285
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2267220017299773e-06,
+      "loss": 0.707,
+      "step": 28286
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2260946617738954e-06,
+      "loss": 0.8382,
+      "step": 28287
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.225467371078406e-06,
+      "loss": 0.8999,
+      "step": 28288
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.224840129648069e-06,
+      "loss": 0.5994,
+      "step": 28289
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2242129374874478e-06,
+      "loss": 0.975,
+      "step": 28290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.223585794601102e-06,
+      "loss": 0.8398,
+      "step": 28291
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2229587009935925e-06,
+      "loss": 0.926,
+      "step": 28292
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.222331656669477e-06,
+      "loss": 0.4105,
+      "step": 28293
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2217046616333213e-06,
+      "loss": 0.7663,
+      "step": 28294
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2210777158896833e-06,
+      "loss": 0.9402,
+      "step": 28295
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2204508194431227e-06,
+      "loss": 0.8998,
+      "step": 28296
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.219823972298194e-06,
+      "loss": 0.6307,
+      "step": 28297
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2191971744594554e-06,
+      "loss": 0.8473,
+      "step": 28298
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.218570425931471e-06,
+      "loss": 1.0794,
+      "step": 28299
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2179437267187953e-06,
+      "loss": 0.8324,
+      "step": 28300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2173170768259852e-06,
+      "loss": 0.7376,
+      "step": 28301
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.216690476257599e-06,
+      "loss": 0.5229,
+      "step": 28302
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2160639250181933e-06,
+      "loss": 0.875,
+      "step": 28303
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2154374231123243e-06,
+      "loss": 0.8571,
+      "step": 28304
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2148109705445442e-06,
+      "loss": 0.5854,
+      "step": 28305
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2141845673194183e-06,
+      "loss": 0.7977,
+      "step": 28306
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2135582134414934e-06,
+      "loss": 0.7407,
+      "step": 28307
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2129319089153266e-06,
+      "loss": 0.9283,
+      "step": 28308
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2123056537454722e-06,
+      "loss": 0.8863,
+      "step": 28309
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2116794479364845e-06,
+      "loss": 1.1625,
+      "step": 28310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.211053291492915e-06,
+      "loss": 0.6436,
+      "step": 28311
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2104271844193225e-06,
+      "loss": 0.7388,
+      "step": 28312
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.209801126720259e-06,
+      "loss": 0.7252,
+      "step": 28313
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2091751184002738e-06,
+      "loss": 0.8354,
+      "step": 28314
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.208549159463923e-06,
+      "loss": 0.7144,
+      "step": 28315
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2079232499157555e-06,
+      "loss": 1.029,
+      "step": 28316
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2072973897603255e-06,
+      "loss": 0.7879,
+      "step": 28317
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.206671579002182e-06,
+      "loss": 0.7642,
+      "step": 28318
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2060458176458774e-06,
+      "loss": 0.972,
+      "step": 28319
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.205420105695963e-06,
+      "loss": 0.7442,
+      "step": 28320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.204794443156988e-06,
+      "loss": 0.5629,
+      "step": 28321
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.204168830033502e-06,
+      "loss": 0.9129,
+      "step": 28322
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2035432663300514e-06,
+      "loss": 0.8635,
+      "step": 28323
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2029177520511924e-06,
+      "loss": 0.9725,
+      "step": 28324
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2022922872014696e-06,
+      "loss": 0.951,
+      "step": 28325
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.201666871785436e-06,
+      "loss": 1.0824,
+      "step": 28326
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.201041505807633e-06,
+      "loss": 0.4944,
+      "step": 28327
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2004161892726104e-06,
+      "loss": 0.4813,
+      "step": 28328
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1997909221849142e-06,
+      "loss": 0.7742,
+      "step": 28329
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1991657045490954e-06,
+      "loss": 0.792,
+      "step": 28330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1985405363697e-06,
+      "loss": 0.564,
+      "step": 28331
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.197915417651273e-06,
+      "loss": 0.8085,
+      "step": 28332
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1972903483983596e-06,
+      "loss": 0.7838,
+      "step": 28333
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.196665328615507e-06,
+      "loss": 0.4779,
+      "step": 28334
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1960403583072596e-06,
+      "loss": 1.001,
+      "step": 28335
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.195415437478162e-06,
+      "loss": 0.959,
+      "step": 28336
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.194790566132758e-06,
+      "loss": 0.6853,
+      "step": 28337
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1941657442755945e-06,
+      "loss": 0.9282,
+      "step": 28338
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.193540971911212e-06,
+      "loss": 0.9437,
+      "step": 28339
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1929162490441566e-06,
+      "loss": 0.9297,
+      "step": 28340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1922915756789663e-06,
+      "loss": 0.9537,
+      "step": 28341
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1916669518201917e-06,
+      "loss": 0.9381,
+      "step": 28342
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.19104237747237e-06,
+      "loss": 0.8913,
+      "step": 28343
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1904178526400443e-06,
+      "loss": 0.8519,
+      "step": 28344
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.189793377327757e-06,
+      "loss": 0.6651,
+      "step": 28345
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1891689515400504e-06,
+      "loss": 0.7211,
+      "step": 28346
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.188544575281458e-06,
+      "loss": 0.7335,
+      "step": 28347
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1879202485565288e-06,
+      "loss": 0.8027,
+      "step": 28348
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1872959713698005e-06,
+      "loss": 0.6637,
+      "step": 28349
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.186671743725812e-06,
+      "loss": 0.8509,
+      "step": 28350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1860475656291034e-06,
+      "loss": 0.7522,
+      "step": 28351
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.185423437084213e-06,
+      "loss": 0.9378,
+      "step": 28352
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1847993580956804e-06,
+      "loss": 0.9124,
+      "step": 28353
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.184175328668041e-06,
+      "loss": 0.8742,
+      "step": 28354
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.183551348805838e-06,
+      "loss": 1.018,
+      "step": 28355
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1829274185136092e-06,
+      "loss": 0.9358,
+      "step": 28356
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.182303537795888e-06,
+      "loss": 0.6529,
+      "step": 28357
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1816797066572114e-06,
+      "loss": 0.825,
+      "step": 28358
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1810559251021132e-06,
+      "loss": 0.6931,
+      "step": 28359
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.180432193135138e-06,
+      "loss": 0.6757,
+      "step": 28360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1798085107608166e-06,
+      "loss": 1.1495,
+      "step": 28361
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1791848779836853e-06,
+      "loss": 0.5057,
+      "step": 28362
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.178561294808279e-06,
+      "loss": 0.8328,
+      "step": 28363
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1779377612391316e-06,
+      "loss": 0.9336,
+      "step": 28364
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1773142772807796e-06,
+      "loss": 0.607,
+      "step": 28365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1766908429377553e-06,
+      "loss": 1.0549,
+      "step": 28366
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1760674582145923e-06,
+      "loss": 0.5038,
+      "step": 28367
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.175444123115825e-06,
+      "loss": 0.6954,
+      "step": 28368
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1748208376459865e-06,
+      "loss": 0.6823,
+      "step": 28369
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1741976018096078e-06,
+      "loss": 1.0913,
+      "step": 28370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1735744156112224e-06,
+      "loss": 0.9146,
+      "step": 28371
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1729512790553595e-06,
+      "loss": 0.7279,
+      "step": 28372
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1723281921465552e-06,
+      "loss": 0.8466,
+      "step": 28373
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1717051548893396e-06,
+      "loss": 0.6555,
+      "step": 28374
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.171082167288241e-06,
+      "loss": 0.9638,
+      "step": 28375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1704592293477953e-06,
+      "loss": 0.8234,
+      "step": 28376
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.169836341072522e-06,
+      "loss": 0.6353,
+      "step": 28377
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.16921350246696e-06,
+      "loss": 0.7713,
+      "step": 28378
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1685907135356362e-06,
+      "loss": 0.6201,
+      "step": 28379
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1679679742830806e-06,
+      "loss": 0.5541,
+      "step": 28380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1673452847138185e-06,
+      "loss": 0.738,
+      "step": 28381
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.166722644832382e-06,
+      "loss": 0.7004,
+      "step": 28382
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.166100054643296e-06,
+      "loss": 0.5519,
+      "step": 28383
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.165477514151086e-06,
+      "loss": 0.7551,
+      "step": 28384
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.164855023360287e-06,
+      "loss": 0.7208,
+      "step": 28385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1642325822754237e-06,
+      "loss": 1.0251,
+      "step": 28386
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1636101909010165e-06,
+      "loss": 0.9419,
+      "step": 28387
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1629878492415957e-06,
+      "loss": 0.8715,
+      "step": 28388
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1623655573016863e-06,
+      "loss": 0.9584,
+      "step": 28389
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.161743315085811e-06,
+      "loss": 0.864,
+      "step": 28390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1611211225985006e-06,
+      "loss": 0.9846,
+      "step": 28391
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1604989798442774e-06,
+      "loss": 0.8145,
+      "step": 28392
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.159876886827664e-06,
+      "loss": 1.0168,
+      "step": 28393
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.159254843553187e-06,
+      "loss": 1.1172,
+      "step": 28394
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.158632850025367e-06,
+      "loss": 0.7018,
+      "step": 28395
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1580109062487287e-06,
+      "loss": 1.089,
+      "step": 28396
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.157389012227796e-06,
+      "loss": 0.8169,
+      "step": 28397
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1567671679670897e-06,
+      "loss": 0.7862,
+      "step": 28398
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1561453734711323e-06,
+      "loss": 0.9138,
+      "step": 28399
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1555236287444457e-06,
+      "loss": 0.749,
+      "step": 28400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.154901933791551e-06,
+      "loss": 0.9091,
+      "step": 28401
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1542802886169666e-06,
+      "loss": 1.0284,
+      "step": 28402
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.15365869322522e-06,
+      "loss": 0.902,
+      "step": 28403
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1530371476208268e-06,
+      "loss": 1.1541,
+      "step": 28404
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.152415651808308e-06,
+      "loss": 0.6528,
+      "step": 28405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.151794205792187e-06,
+      "loss": 0.9512,
+      "step": 28406
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1511728095769746e-06,
+      "loss": 0.9204,
+      "step": 28407
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1505514631671918e-06,
+      "loss": 0.7678,
+      "step": 28408
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1499301665673622e-06,
+      "loss": 0.9654,
+      "step": 28409
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1493089197820015e-06,
+      "loss": 1.1058,
+      "step": 28410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1486877228156264e-06,
+      "loss": 0.9923,
+      "step": 28411
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1480665756727546e-06,
+      "loss": 0.907,
+      "step": 28412
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1474454783579046e-06,
+      "loss": 0.8022,
+      "step": 28413
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1468244308755915e-06,
+      "loss": 0.68,
+      "step": 28414
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.146203433230329e-06,
+      "loss": 0.9599,
+      "step": 28415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1455824854266436e-06,
+      "loss": 0.6609,
+      "step": 28416
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.14496158746904e-06,
+      "loss": 0.704,
+      "step": 28417
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.144340739362036e-06,
+      "loss": 0.8981,
+      "step": 28418
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.143719941110148e-06,
+      "loss": 0.8694,
+      "step": 28419
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1430991927178867e-06,
+      "loss": 0.7128,
+      "step": 28420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1424784941897723e-06,
+      "loss": 0.9741,
+      "step": 28421
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.141857845530316e-06,
+      "loss": 0.8018,
+      "step": 28422
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.141237246744031e-06,
+      "loss": 0.9759,
+      "step": 28423
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1406166978354315e-06,
+      "loss": 0.7641,
+      "step": 28424
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.139996198809028e-06,
+      "loss": 1.0206,
+      "step": 28425
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1393757496693343e-06,
+      "loss": 0.7334,
+      "step": 28426
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1387553504208614e-06,
+      "loss": 0.9872,
+      "step": 28427
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.138135001068122e-06,
+      "loss": 0.8861,
+      "step": 28428
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1375147016156272e-06,
+      "loss": 1.0148,
+      "step": 28429
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.136894452067887e-06,
+      "loss": 0.9226,
+      "step": 28430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1362742524294132e-06,
+      "loss": 1.168,
+      "step": 28431
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1356541027047137e-06,
+      "loss": 0.8987,
+      "step": 28432
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1350340028982983e-06,
+      "loss": 0.5612,
+      "step": 28433
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.13441395301468e-06,
+      "loss": 0.944,
+      "step": 28434
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.133793953058366e-06,
+      "loss": 0.9723,
+      "step": 28435
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1331740030338688e-06,
+      "loss": 0.6967,
+      "step": 28436
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1325541029456894e-06,
+      "loss": 0.7147,
+      "step": 28437
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1319342527983355e-06,
+      "loss": 0.8201,
+      "step": 28438
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1313144525963225e-06,
+      "loss": 0.8496,
+      "step": 28439
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1306947023441524e-06,
+      "loss": 0.7476,
+      "step": 28440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.130075002046333e-06,
+      "loss": 0.8282,
+      "step": 28441
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1294553517073722e-06,
+      "loss": 0.9212,
+      "step": 28442
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1288357513317745e-06,
+      "loss": 0.5221,
+      "step": 28443
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1282162009240467e-06,
+      "loss": 0.7832,
+      "step": 28444
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.127596700488691e-06,
+      "loss": 0.744,
+      "step": 28445
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.126977250030221e-06,
+      "loss": 1.0038,
+      "step": 28446
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.126357849553132e-06,
+      "loss": 0.8453,
+      "step": 28447
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.125738499061932e-06,
+      "loss": 0.781,
+      "step": 28448
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.125119198561125e-06,
+      "loss": 0.7354,
+      "step": 28449
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1244999480552152e-06,
+      "loss": 0.986,
+      "step": 28450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.123880747548702e-06,
+      "loss": 0.9073,
+      "step": 28451
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1232615970460957e-06,
+      "loss": 0.8249,
+      "step": 28452
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.122642496551893e-06,
+      "loss": 0.9645,
+      "step": 28453
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.122023446070599e-06,
+      "loss": 1.1139,
+      "step": 28454
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.121404445606714e-06,
+      "loss": 0.9408,
+      "step": 28455
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.120785495164739e-06,
+      "loss": 1.0468,
+      "step": 28456
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1201665947491764e-06,
+      "loss": 0.8377,
+      "step": 28457
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.119547744364526e-06,
+      "loss": 0.8464,
+      "step": 28458
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.118928944015288e-06,
+      "loss": 0.728,
+      "step": 28459
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1183101937059646e-06,
+      "loss": 0.9538,
+      "step": 28460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.117691493441052e-06,
+      "loss": 0.5714,
+      "step": 28461
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.117072843225052e-06,
+      "loss": 0.8414,
+      "step": 28462
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.116454243062459e-06,
+      "loss": 0.8779,
+      "step": 28463
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.115835692957778e-06,
+      "loss": 0.7165,
+      "step": 28464
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.115217192915505e-06,
+      "loss": 0.792,
+      "step": 28465
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.11459874294014e-06,
+      "loss": 0.8862,
+      "step": 28466
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1139803430361748e-06,
+      "loss": 0.5535,
+      "step": 28467
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.113361993208108e-06,
+      "loss": 0.4276,
+      "step": 28468
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.112743693460436e-06,
+      "loss": 0.7981,
+      "step": 28469
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.112125443797659e-06,
+      "loss": 1.1745,
+      "step": 28470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1115072442242712e-06,
+      "loss": 1.1331,
+      "step": 28471
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1108890947447666e-06,
+      "loss": 1.2922,
+      "step": 28472
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1102709953636423e-06,
+      "loss": 0.7034,
+      "step": 28473
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1096529460853918e-06,
+      "loss": 0.7624,
+      "step": 28474
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.109034946914511e-06,
+      "loss": 0.6689,
+      "step": 28475
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.108416997855492e-06,
+      "loss": 1.0774,
+      "step": 28476
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1077990989128313e-06,
+      "loss": 0.8726,
+      "step": 28477
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1071812500910205e-06,
+      "loss": 0.6934,
+      "step": 28478
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.106563451394553e-06,
+      "loss": 0.664,
+      "step": 28479
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1059457028279215e-06,
+      "loss": 0.8832,
+      "step": 28480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1053280043956157e-06,
+      "loss": 0.903,
+      "step": 28481
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104710356102133e-06,
+      "loss": 0.8812,
+      "step": 28482
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104092757951962e-06,
+      "loss": 0.939,
+      "step": 28483
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1034752099495944e-06,
+      "loss": 0.7004,
+      "step": 28484
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1028577120995216e-06,
+      "loss": 0.6117,
+      "step": 28485
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1022402644062354e-06,
+      "loss": 0.7558,
+      "step": 28486
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1016228668742187e-06,
+      "loss": 0.7931,
+      "step": 28487
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1010055195079693e-06,
+      "loss": 0.7128,
+      "step": 28488
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.100388222311974e-06,
+      "loss": 0.8247,
+      "step": 28489
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0997709752907223e-06,
+      "loss": 1.0022,
+      "step": 28490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0991537784487015e-06,
+      "loss": 0.7946,
+      "step": 28491
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.098536631790401e-06,
+      "loss": 0.8262,
+      "step": 28492
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0979195353203083e-06,
+      "loss": 0.899,
+      "step": 28493
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.097302489042908e-06,
+      "loss": 1.0746,
+      "step": 28494
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0966854929626943e-06,
+      "loss": 1.0071,
+      "step": 28495
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.096068547084152e-06,
+      "loss": 0.5082,
+      "step": 28496
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.095451651411765e-06,
+      "loss": 0.8556,
+      "step": 28497
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.094834805950019e-06,
+      "loss": 0.9856,
+      "step": 28498
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.094218010703397e-06,
+      "loss": 0.8029,
+      "step": 28499
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0936012656763937e-06,
+      "loss": 0.8626,
+      "step": 28500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0929845708734875e-06,
+      "loss": 0.7418,
+      "step": 28501
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0923679262991646e-06,
+      "loss": 0.8621,
+      "step": 28502
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0917513319579095e-06,
+      "loss": 1.0981,
+      "step": 28503
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0911347878542065e-06,
+      "loss": 0.711,
+      "step": 28504
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.090518293992537e-06,
+      "loss": 0.6872,
+      "step": 28505
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.089901850377387e-06,
+      "loss": 0.9327,
+      "step": 28506
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.089285457013238e-06,
+      "loss": 0.7011,
+      "step": 28507
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.088669113904572e-06,
+      "loss": 0.8822,
+      "step": 28508
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0880528210558724e-06,
+      "loss": 0.7771,
+      "step": 28509
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0874365784716197e-06,
+      "loss": 0.8448,
+      "step": 28510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.086820386156296e-06,
+      "loss": 0.6794,
+      "step": 28511
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.086204244114379e-06,
+      "loss": 0.9322,
+      "step": 28512
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0855881523503552e-06,
+      "loss": 0.8768,
+      "step": 28513
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0849721108687025e-06,
+      "loss": 0.597,
+      "step": 28514
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0843561196739013e-06,
+      "loss": 0.663,
+      "step": 28515
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0837401787704324e-06,
+      "loss": 0.9228,
+      "step": 28516
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0831242881627677e-06,
+      "loss": 0.6861,
+      "step": 28517
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0825084478553935e-06,
+      "loss": 0.6982,
+      "step": 28518
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0818926578527864e-06,
+      "loss": 0.835,
+      "step": 28519
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.081276918159424e-06,
+      "loss": 0.8108,
+      "step": 28520
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0806612287797844e-06,
+      "loss": 0.6874,
+      "step": 28521
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0800455897183448e-06,
+      "loss": 0.7995,
+      "step": 28522
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.079430000979581e-06,
+      "loss": 0.9776,
+      "step": 28523
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0788144625679684e-06,
+      "loss": 0.7448,
+      "step": 28524
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.078198974487989e-06,
+      "loss": 1.025,
+      "step": 28525
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.077583536744119e-06,
+      "loss": 0.7496,
+      "step": 28526
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0769681493408256e-06,
+      "loss": 0.6493,
+      "step": 28527
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.076352812282589e-06,
+      "loss": 0.5698,
+      "step": 28528
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.075737525573884e-06,
+      "loss": 1.0241,
+      "step": 28529
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.075122289219181e-06,
+      "loss": 0.8427,
+      "step": 28530
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0745071032229624e-06,
+      "loss": 1.0121,
+      "step": 28531
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.073891967589696e-06,
+      "loss": 0.8387,
+      "step": 28532
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0732768823238555e-06,
+      "loss": 0.777,
+      "step": 28533
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0726618474299162e-06,
+      "loss": 0.787,
+      "step": 28534
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.072046862912349e-06,
+      "loss": 0.7633,
+      "step": 28535
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0714319287756257e-06,
+      "loss": 0.911,
+      "step": 28536
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.070817045024219e-06,
+      "loss": 0.5968,
+      "step": 28537
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0702022116626008e-06,
+      "loss": 0.7472,
+      "step": 28538
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0695874286952408e-06,
+      "loss": 0.841,
+      "step": 28539
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.068972696126611e-06,
+      "loss": 0.8298,
+      "step": 28540
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0683580139611814e-06,
+      "loss": 0.4552,
+      "step": 28541
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0677433822034187e-06,
+      "loss": 1.128,
+      "step": 28542
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0671288008577993e-06,
+      "loss": 0.8998,
+      "step": 28543
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0665142699287896e-06,
+      "loss": 0.7105,
+      "step": 28544
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0658997894208573e-06,
+      "loss": 0.7396,
+      "step": 28545
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0652853593384757e-06,
+      "loss": 0.6761,
+      "step": 28546
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.064670979686105e-06,
+      "loss": 0.6414,
+      "step": 28547
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0640566504682145e-06,
+      "loss": 0.7739,
+      "step": 28548
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.063442371689277e-06,
+      "loss": 0.9623,
+      "step": 28549
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0628281433537576e-06,
+      "loss": 0.8915,
+      "step": 28550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0622139654661208e-06,
+      "loss": 0.8416,
+      "step": 28551
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0615998380308363e-06,
+      "loss": 0.9296,
+      "step": 28552
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0609857610523673e-06,
+      "loss": 0.8838,
+      "step": 28553
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.06037173453518e-06,
+      "loss": 1.0409,
+      "step": 28554
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0597577584837378e-06,
+      "loss": 1.0176,
+      "step": 28555
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0591438329025134e-06,
+      "loss": 0.6145,
+      "step": 28556
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.058529957795964e-06,
+      "loss": 0.7315,
+      "step": 28557
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.057916133168555e-06,
+      "loss": 0.8492,
+      "step": 28558
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0573023590247496e-06,
+      "loss": 0.8171,
+      "step": 28559
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0566886353690106e-06,
+      "loss": 0.9174,
+      "step": 28560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.056074962205805e-06,
+      "loss": 0.8268,
+      "step": 28561
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.055461339539595e-06,
+      "loss": 0.9806,
+      "step": 28562
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.05484776737484e-06,
+      "loss": 0.6638,
+      "step": 28563
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0542342457160037e-06,
+      "loss": 1.0497,
+      "step": 28564
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.053620774567546e-06,
+      "loss": 0.7442,
+      "step": 28565
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.053007353933931e-06,
+      "loss": 0.9423,
+      "step": 28566
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.052393983819617e-06,
+      "loss": 0.629,
+      "step": 28567
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0517806642290673e-06,
+      "loss": 0.7378,
+      "step": 28568
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0511673951667377e-06,
+      "loss": 0.7177,
+      "step": 28569
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0505541766370915e-06,
+      "loss": 1.013,
+      "step": 28570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.049941008644588e-06,
+      "loss": 0.8742,
+      "step": 28571
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.049327891193684e-06,
+      "loss": 0.787,
+      "step": 28572
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0487148242888367e-06,
+      "loss": 0.6257,
+      "step": 28573
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0481018079345102e-06,
+      "loss": 0.9108,
+      "step": 28574
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.047488842135159e-06,
+      "loss": 0.6061,
+      "step": 28575
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0468759268952453e-06,
+      "loss": 1.0118,
+      "step": 28576
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.046263062219218e-06,
+      "loss": 0.8358,
+      "step": 28577
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0456502481115346e-06,
+      "loss": 0.6268,
+      "step": 28578
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0450374845766584e-06,
+      "loss": 1.1081,
+      "step": 28579
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.044424771619041e-06,
+      "loss": 0.7996,
+      "step": 28580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0438121092431406e-06,
+      "loss": 0.6193,
+      "step": 28581
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0431994974534097e-06,
+      "loss": 0.5497,
+      "step": 28582
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.042586936254306e-06,
+      "loss": 0.7492,
+      "step": 28583
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0419744256502813e-06,
+      "loss": 0.9717,
+      "step": 28584
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0413619656457883e-06,
+      "loss": 1.0012,
+      "step": 28585
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0407495562452903e-06,
+      "loss": 0.9516,
+      "step": 28586
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.040137197453232e-06,
+      "loss": 1.1548,
+      "step": 28587
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.039524889274068e-06,
+      "loss": 0.9683,
+      "step": 28588
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.038912631712252e-06,
+      "loss": 1.1002,
+      "step": 28589
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.038300424772237e-06,
+      "loss": 0.9782,
+      "step": 28590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.037688268458471e-06,
+      "loss": 1.018,
+      "step": 28591
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.037076162775412e-06,
+      "loss": 0.8201,
+      "step": 28592
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.036464107727508e-06,
+      "loss": 0.8882,
+      "step": 28593
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.03585210331921e-06,
+      "loss": 0.7715,
+      "step": 28594
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.03524014955497e-06,
+      "loss": 0.5187,
+      "step": 28595
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0346282464392365e-06,
+      "loss": 1.0099,
+      "step": 28596
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.03401639397646e-06,
+      "loss": 0.5857,
+      "step": 28597
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.03340459217109e-06,
+      "loss": 1.006,
+      "step": 28598
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0327928410275753e-06,
+      "loss": 1.2766,
+      "step": 28599
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0321811405503645e-06,
+      "loss": 1.1386,
+      "step": 28600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0315694907439062e-06,
+      "loss": 0.7133,
+      "step": 28601
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.030957891612648e-06,
+      "loss": 0.9234,
+      "step": 28602
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.030346343161036e-06,
+      "loss": 0.5888,
+      "step": 28603
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0297348453935214e-06,
+      "loss": 0.5765,
+      "step": 28604
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0291233983145494e-06,
+      "loss": 0.9123,
+      "step": 28605
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0285120019285697e-06,
+      "loss": 0.9891,
+      "step": 28606
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0279006562400213e-06,
+      "loss": 0.6779,
+      "step": 28607
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.027289361253354e-06,
+      "loss": 0.7645,
+      "step": 28608
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0266781169730088e-06,
+      "loss": 0.9024,
+      "step": 28609
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.026066923403438e-06,
+      "loss": 0.7968,
+      "step": 28610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.025455780549084e-06,
+      "loss": 0.894,
+      "step": 28611
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.024844688414389e-06,
+      "loss": 0.8253,
+      "step": 28612
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.024233647003798e-06,
+      "loss": 0.5953,
+      "step": 28613
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0236226563217554e-06,
+      "loss": 0.9296,
+      "step": 28614
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0230117163727024e-06,
+      "loss": 0.9073,
+      "step": 28615
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.022400827161084e-06,
+      "loss": 0.7489,
+      "step": 28616
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0217899886913404e-06,
+      "loss": 0.6428,
+      "step": 28617
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0211792009679163e-06,
+      "loss": 0.6626,
+      "step": 28618
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0205684639952514e-06,
+      "loss": 1.1409,
+      "step": 28619
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.019957777777788e-06,
+      "loss": 0.9142,
+      "step": 28620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0193471423199626e-06,
+      "loss": 0.8276,
+      "step": 28621
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.018736557626224e-06,
+      "loss": 0.8039,
+      "step": 28622
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.018126023701009e-06,
+      "loss": 0.5071,
+      "step": 28623
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0175155405487553e-06,
+      "loss": 0.7497,
+      "step": 28624
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0169051081739054e-06,
+      "loss": 0.8125,
+      "step": 28625
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0162947265808994e-06,
+      "loss": 0.805,
+      "step": 28626
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.015684395774168e-06,
+      "loss": 0.8365,
+      "step": 28627
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0150741157581575e-06,
+      "loss": 0.9983,
+      "step": 28628
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0144638865373045e-06,
+      "loss": 0.7737,
+      "step": 28629
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0138537081160446e-06,
+      "loss": 0.755,
+      "step": 28630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.013243580498818e-06,
+      "loss": 0.7768,
+      "step": 28631
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.012633503690059e-06,
+      "loss": 0.892,
+      "step": 28632
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.012023477694205e-06,
+      "loss": 0.9721,
+      "step": 28633
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.01141350251569e-06,
+      "loss": 0.8691,
+      "step": 28634
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.010803578158954e-06,
+      "loss": 0.7207,
+      "step": 28635
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0101937046284345e-06,
+      "loss": 1.0859,
+      "step": 28636
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0095838819285593e-06,
+      "loss": 0.772,
+      "step": 28637
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.008974110063766e-06,
+      "loss": 0.8487,
+      "step": 28638
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.008364389038486e-06,
+      "loss": 0.9872,
+      "step": 28639
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.007754718857161e-06,
+      "loss": 0.5208,
+      "step": 28640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.007145099524219e-06,
+      "loss": 0.7276,
+      "step": 28641
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0065355310440946e-06,
+      "loss": 1.1019,
+      "step": 28642
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.005926013421221e-06,
+      "loss": 0.8261,
+      "step": 28643
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0053165466600288e-06,
+      "loss": 0.9985,
+      "step": 28644
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.004707130764952e-06,
+      "loss": 0.6209,
+      "step": 28645
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0040977657404224e-06,
+      "loss": 0.9611,
+      "step": 28646
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.00348845159087e-06,
+      "loss": 0.7369,
+      "step": 28647
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0028791883207277e-06,
+      "loss": 0.493,
+      "step": 28648
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.002269975934423e-06,
+      "loss": 0.7837,
+      "step": 28649
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.00166081443639e-06,
+      "loss": 0.7778,
+      "step": 28650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.001051703831056e-06,
+      "loss": 0.8452,
+      "step": 28651
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0004426441228475e-06,
+      "loss": 0.84,
+      "step": 28652
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9998336353162006e-06,
+      "loss": 0.9877,
+      "step": 28653
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.999224677415542e-06,
+      "loss": 0.8591,
+      "step": 28654
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9986157704252984e-06,
+      "loss": 0.9503,
+      "step": 28655
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9980069143499015e-06,
+      "loss": 0.9128,
+      "step": 28656
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9973981091937686e-06,
+      "loss": 0.8913,
+      "step": 28657
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.996789354961338e-06,
+      "loss": 0.8572,
+      "step": 28658
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9961806516570337e-06,
+      "loss": 0.802,
+      "step": 28659
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9955719992852806e-06,
+      "loss": 0.9417,
+      "step": 28660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9949633978505064e-06,
+      "loss": 0.6047,
+      "step": 28661
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.994354847357136e-06,
+      "loss": 0.9577,
+      "step": 28662
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9937463478095953e-06,
+      "loss": 0.6269,
+      "step": 28663
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.993137899212305e-06,
+      "loss": 0.8736,
+      "step": 28664
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9925295015696978e-06,
+      "loss": 1.0639,
+      "step": 28665
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.991921154886198e-06,
+      "loss": 0.6566,
+      "step": 28666
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.991312859166222e-06,
+      "loss": 0.6789,
+      "step": 28667
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.990704614414197e-06,
+      "loss": 0.5512,
+      "step": 28668
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9900964206345472e-06,
+      "loss": 0.822,
+      "step": 28669
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9894882778316903e-06,
+      "loss": 0.9561,
+      "step": 28670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.988880186010057e-06,
+      "loss": 0.8051,
+      "step": 28671
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.988272145174066e-06,
+      "loss": 1.0199,
+      "step": 28672
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9876641553281383e-06,
+      "loss": 0.7155,
+      "step": 28673
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9870562164766946e-06,
+      "loss": 0.6554,
+      "step": 28674
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9864483286241585e-06,
+      "loss": 0.8611,
+      "step": 28675
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.985840491774948e-06,
+      "loss": 0.8748,
+      "step": 28676
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.985232705933484e-06,
+      "loss": 0.7097,
+      "step": 28677
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9846249711041874e-06,
+      "loss": 0.9235,
+      "step": 28678
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9840172872914774e-06,
+      "loss": 1.006,
+      "step": 28679
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9834096544997725e-06,
+      "loss": 0.9439,
+      "step": 28680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9828020727334915e-06,
+      "loss": 0.582,
+      "step": 28681
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.98219454199705e-06,
+      "loss": 0.7502,
+      "step": 28682
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.981587062294872e-06,
+      "loss": 0.9172,
+      "step": 28683
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9809796336313725e-06,
+      "loss": 0.8997,
+      "step": 28684
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9803722560109683e-06,
+      "loss": 0.6573,
+      "step": 28685
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9797649294380803e-06,
+      "loss": 0.8221,
+      "step": 28686
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9791576539171175e-06,
+      "loss": 0.807,
+      "step": 28687
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.978550429452497e-06,
+      "loss": 0.7318,
+      "step": 28688
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.977943256048641e-06,
+      "loss": 0.9473,
+      "step": 28689
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9773361337099605e-06,
+      "loss": 0.8315,
+      "step": 28690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.976729062440873e-06,
+      "loss": 0.812,
+      "step": 28691
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9761220422457902e-06,
+      "loss": 0.8682,
+      "step": 28692
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.975515073129128e-06,
+      "loss": 0.7201,
+      "step": 28693
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9749081550953007e-06,
+      "loss": 0.7875,
+      "step": 28694
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9743012881487187e-06,
+      "loss": 0.896,
+      "step": 28695
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9736944722938042e-06,
+      "loss": 0.9144,
+      "step": 28696
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9730877075349597e-06,
+      "loss": 0.7731,
+      "step": 28697
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.972480993876602e-06,
+      "loss": 0.9126,
+      "step": 28698
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9718743313231435e-06,
+      "loss": 1.0017,
+      "step": 28699
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.971267719878992e-06,
+      "loss": 0.9447,
+      "step": 28700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9706611595485644e-06,
+      "loss": 0.8992,
+      "step": 28701
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9700546503362706e-06,
+      "loss": 0.9167,
+      "step": 28702
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.969448192246519e-06,
+      "loss": 1.0116,
+      "step": 28703
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9688417852837214e-06,
+      "loss": 0.8962,
+      "step": 28704
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9682354294522875e-06,
+      "loss": 0.8403,
+      "step": 28705
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9676291247566257e-06,
+      "loss": 0.8895,
+      "step": 28706
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9670228712011472e-06,
+      "loss": 0.8571,
+      "step": 28707
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9664166687902584e-06,
+      "loss": 0.9613,
+      "step": 28708
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.965810517528368e-06,
+      "loss": 0.9049,
+      "step": 28709
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.965204417419886e-06,
+      "loss": 0.8682,
+      "step": 28710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.964598368469217e-06,
+      "loss": 1.1133,
+      "step": 28711
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.963992370680772e-06,
+      "loss": 0.8497,
+      "step": 28712
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9633864240589507e-06,
+      "loss": 0.6945,
+      "step": 28713
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.962780528608169e-06,
+      "loss": 0.7878,
+      "step": 28714
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.962174684332828e-06,
+      "loss": 0.6197,
+      "step": 28715
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9615688912373387e-06,
+      "loss": 1.1894,
+      "step": 28716
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9609631493260973e-06,
+      "loss": 1.152,
+      "step": 28717
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.960357458603511e-06,
+      "loss": 1.0298,
+      "step": 28718
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9597518190739903e-06,
+      "loss": 0.5278,
+      "step": 28719
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.959146230741936e-06,
+      "loss": 1.0562,
+      "step": 28720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.958540693611752e-06,
+      "loss": 0.5934,
+      "step": 28721
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9579352076878433e-06,
+      "loss": 1.148,
+      "step": 28722
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.957329772974611e-06,
+      "loss": 0.9437,
+      "step": 28723
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9567243894764576e-06,
+      "loss": 1.1323,
+      "step": 28724
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.956119057197785e-06,
+      "loss": 0.6029,
+      "step": 28725
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.955513776143002e-06,
+      "loss": 0.7772,
+      "step": 28726
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9549085463165016e-06,
+      "loss": 0.8815,
+      "step": 28727
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9543033677226896e-06,
+      "loss": 0.7274,
+      "step": 28728
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9536982403659663e-06,
+      "loss": 0.8934,
+      "step": 28729
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.953093164250731e-06,
+      "loss": 0.777,
+      "step": 28730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.952488139381382e-06,
+      "loss": 0.8285,
+      "step": 28731
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.951883165762325e-06,
+      "loss": 0.8245,
+      "step": 28732
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.951278243397956e-06,
+      "loss": 0.5477,
+      "step": 28733
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.950673372292674e-06,
+      "loss": 0.6354,
+      "step": 28734
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.950068552450879e-06,
+      "loss": 0.9294,
+      "step": 28735
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.949463783876968e-06,
+      "loss": 0.5672,
+      "step": 28736
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9488590665753393e-06,
+      "loss": 0.9298,
+      "step": 28737
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.94825440055039e-06,
+      "loss": 0.9828,
+      "step": 28738
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9476497858065177e-06,
+      "loss": 0.8276,
+      "step": 28739
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9470452223481206e-06,
+      "loss": 0.8986,
+      "step": 28740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.946440710179592e-06,
+      "loss": 0.688,
+      "step": 28741
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9458362493053316e-06,
+      "loss": 0.7897,
+      "step": 28742
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.945231839729729e-06,
+      "loss": 0.5558,
+      "step": 28743
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9446274814571875e-06,
+      "loss": 1.0078,
+      "step": 28744
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9440231744920987e-06,
+      "loss": 0.9406,
+      "step": 28745
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9434189188388594e-06,
+      "loss": 0.8274,
+      "step": 28746
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9428147145018584e-06,
+      "loss": 0.7849,
+      "step": 28747
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9422105614854922e-06,
+      "loss": 0.9954,
+      "step": 28748
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.941606459794153e-06,
+      "loss": 0.7407,
+      "step": 28749
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9410024094322363e-06,
+      "loss": 0.9322,
+      "step": 28750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.940398410404136e-06,
+      "loss": 0.738,
+      "step": 28751
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9397944627142417e-06,
+      "loss": 0.7593,
+      "step": 28752
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9391905663669453e-06,
+      "loss": 1.0383,
+      "step": 28753
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9385867213666396e-06,
+      "loss": 0.8055,
+      "step": 28754
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9379829277177152e-06,
+      "loss": 0.8366,
+      "step": 28755
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9373791854245636e-06,
+      "loss": 0.9906,
+      "step": 28756
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.936775494491575e-06,
+      "loss": 0.8405,
+      "step": 28757
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.936171854923139e-06,
+      "loss": 0.7507,
+      "step": 28758
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9355682667236463e-06,
+      "loss": 0.8129,
+      "step": 28759
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.934964729897485e-06,
+      "loss": 0.6722,
+      "step": 28760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.934361244449042e-06,
+      "loss": 0.6761,
+      "step": 28761
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9337578103827113e-06,
+      "loss": 0.9878,
+      "step": 28762
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.933154427702879e-06,
+      "loss": 1.1317,
+      "step": 28763
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9325510964139324e-06,
+      "loss": 0.7615,
+      "step": 28764
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9319478165202587e-06,
+      "loss": 0.8077,
+      "step": 28765
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.931344588026248e-06,
+      "loss": 0.6106,
+      "step": 28766
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9307414109362795e-06,
+      "loss": 0.9436,
+      "step": 28767
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.930138285254747e-06,
+      "loss": 0.8792,
+      "step": 28768
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9295352109860333e-06,
+      "loss": 0.8946,
+      "step": 28769
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9289321881345257e-06,
+      "loss": 0.8101,
+      "step": 28770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9283292167046083e-06,
+      "loss": 0.8289,
+      "step": 28771
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9277262967006657e-06,
+      "loss": 0.6308,
+      "step": 28772
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.927123428127083e-06,
+      "loss": 0.7934,
+      "step": 28773
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9265206109882404e-06,
+      "loss": 0.4606,
+      "step": 28774
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.925917845288533e-06,
+      "loss": 0.9176,
+      "step": 28775
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9253151310323326e-06,
+      "loss": 0.8016,
+      "step": 28776
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9247124682240267e-06,
+      "loss": 0.5386,
+      "step": 28777
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9241098568679973e-06,
+      "loss": 0.7139,
+      "step": 28778
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.923507296968626e-06,
+      "loss": 0.8056,
+      "step": 28779
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.922904788530293e-06,
+      "loss": 0.8505,
+      "step": 28780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.922302331557385e-06,
+      "loss": 0.7623,
+      "step": 28781
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.921699926054281e-06,
+      "loss": 0.7087,
+      "step": 28782
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.921097572025361e-06,
+      "loss": 0.6392,
+      "step": 28783
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9204952694750054e-06,
+      "loss": 0.6701,
+      "step": 28784
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9198930184075944e-06,
+      "loss": 0.9463,
+      "step": 28785
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9192908188275072e-06,
+      "loss": 0.3994,
+      "step": 28786
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.918688670739124e-06,
+      "loss": 0.8769,
+      "step": 28787
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9180865741468233e-06,
+      "loss": 0.8013,
+      "step": 28788
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9174845290549823e-06,
+      "loss": 0.8145,
+      "step": 28789
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.916882535467982e-06,
+      "loss": 0.7952,
+      "step": 28790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9162805933901973e-06,
+      "loss": 0.7285,
+      "step": 28791
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9156787028260036e-06,
+      "loss": 0.8966,
+      "step": 28792
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9150768637797843e-06,
+      "loss": 1.0411,
+      "step": 28793
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9144750762559137e-06,
+      "loss": 1.0928,
+      "step": 28794
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9138733402587694e-06,
+      "loss": 0.6465,
+      "step": 28795
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.913271655792722e-06,
+      "loss": 0.7056,
+      "step": 28796
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9126700228621508e-06,
+      "loss": 0.6927,
+      "step": 28797
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.912068441471426e-06,
+      "loss": 0.8534,
+      "step": 28798
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9114669116249307e-06,
+      "loss": 0.7976,
+      "step": 28799
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9108654333270346e-06,
+      "loss": 0.6872,
+      "step": 28800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.910264006582112e-06,
+      "loss": 0.7867,
+      "step": 28801
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9096626313945377e-06,
+      "loss": 0.7669,
+      "step": 28802
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9090613077686834e-06,
+      "loss": 0.7815,
+      "step": 28803
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.908460035708923e-06,
+      "loss": 0.831,
+      "step": 28804
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9078588152196284e-06,
+      "loss": 0.8963,
+      "step": 28805
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.907257646305173e-06,
+      "loss": 1.0126,
+      "step": 28806
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9066565289699267e-06,
+      "loss": 1.0687,
+      "step": 28807
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9060554632182625e-06,
+      "loss": 0.5229,
+      "step": 28808
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.90545444905455e-06,
+      "loss": 0.7975,
+      "step": 28809
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9048534864831577e-06,
+      "loss": 0.9668,
+      "step": 28810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9042525755084628e-06,
+      "loss": 0.7245,
+      "step": 28811
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.90365171613483e-06,
+      "loss": 0.8471,
+      "step": 28812
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.90305090836663e-06,
+      "loss": 0.9778,
+      "step": 28813
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9024501522082326e-06,
+      "loss": 0.9853,
+      "step": 28814
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.901849447664008e-06,
+      "loss": 0.961,
+      "step": 28815
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.901248794738317e-06,
+      "loss": 0.9203,
+      "step": 28816
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.900648193435536e-06,
+      "loss": 0.822,
+      "step": 28817
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.900047643760029e-06,
+      "loss": 0.7719,
+      "step": 28818
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8994471457161655e-06,
+      "loss": 0.8333,
+      "step": 28819
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8988466993083096e-06,
+      "loss": 0.8067,
+      "step": 28820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.898246304540829e-06,
+      "loss": 0.9944,
+      "step": 28821
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8976459614180906e-06,
+      "loss": 1.0399,
+      "step": 28822
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.897045669944456e-06,
+      "loss": 0.8045,
+      "step": 28823
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.896445430124297e-06,
+      "loss": 0.7388,
+      "step": 28824
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8958452419619787e-06,
+      "loss": 0.8949,
+      "step": 28825
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8952451054618602e-06,
+      "loss": 0.8174,
+      "step": 28826
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8946450206283083e-06,
+      "loss": 0.8193,
+      "step": 28827
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.894044987465683e-06,
+      "loss": 0.8883,
+      "step": 28828
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.893445005978356e-06,
+      "loss": 0.7342,
+      "step": 28829
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.892845076170685e-06,
+      "loss": 0.6893,
+      "step": 28830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8922451980470335e-06,
+      "loss": 0.7179,
+      "step": 28831
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8916453716117652e-06,
+      "loss": 0.9209,
+      "step": 28832
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8910455968692406e-06,
+      "loss": 0.5977,
+      "step": 28833
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.890445873823822e-06,
+      "loss": 0.8998,
+      "step": 28834
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.88984620247987e-06,
+      "loss": 0.6442,
+      "step": 28835
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.889246582841747e-06,
+      "loss": 0.6249,
+      "step": 28836
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.888647014913811e-06,
+      "loss": 1.0456,
+      "step": 28837
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.888047498700425e-06,
+      "loss": 0.78,
+      "step": 28838
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8874480342059476e-06,
+      "loss": 0.9396,
+      "step": 28839
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.886848621434737e-06,
+      "loss": 0.8659,
+      "step": 28840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8862492603911496e-06,
+      "loss": 0.5895,
+      "step": 28841
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.885649951079551e-06,
+      "loss": 1.0288,
+      "step": 28842
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8850506935042966e-06,
+      "loss": 0.5031,
+      "step": 28843
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.884451487669744e-06,
+      "loss": 0.6883,
+      "step": 28844
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8838523335802525e-06,
+      "loss": 0.7338,
+      "step": 28845
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8832532312401705e-06,
+      "loss": 1.0377,
+      "step": 28846
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8826541806538656e-06,
+      "loss": 1.0068,
+      "step": 28847
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8820551818256893e-06,
+      "loss": 0.883,
+      "step": 28848
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.881456234759997e-06,
+      "loss": 0.9466,
+      "step": 28849
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8808573394611474e-06,
+      "loss": 0.9573,
+      "step": 28850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8802584959334922e-06,
+      "loss": 0.5562,
+      "step": 28851
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.879659704181389e-06,
+      "loss": 0.7496,
+      "step": 28852
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8790609642091883e-06,
+      "loss": 0.8729,
+      "step": 28853
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8784622760212487e-06,
+      "loss": 0.9789,
+      "step": 28854
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8778636396219262e-06,
+      "loss": 0.7538,
+      "step": 28855
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8772650550155667e-06,
+      "loss": 0.9893,
+      "step": 28856
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.876666522206527e-06,
+      "loss": 0.7331,
+      "step": 28857
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.87606804119916e-06,
+      "loss": 0.6665,
+      "step": 28858
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.875469611997813e-06,
+      "loss": 1.2311,
+      "step": 28859
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8748712346068464e-06,
+      "loss": 0.9668,
+      "step": 28860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8742729090306067e-06,
+      "loss": 0.8233,
+      "step": 28861
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8736746352734456e-06,
+      "loss": 0.7757,
+      "step": 28862
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.873076413339715e-06,
+      "loss": 1.0223,
+      "step": 28863
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.872478243233763e-06,
+      "loss": 0.8999,
+      "step": 28864
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.87188012495994e-06,
+      "loss": 0.7302,
+      "step": 28865
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8712820585225976e-06,
+      "loss": 1.0191,
+      "step": 28866
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.870684043926084e-06,
+      "loss": 0.6353,
+      "step": 28867
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8700860811747468e-06,
+      "loss": 0.9616,
+      "step": 28868
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.869488170272936e-06,
+      "loss": 0.9484,
+      "step": 28869
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.868890311224999e-06,
+      "loss": 0.9792,
+      "step": 28870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8682925040352794e-06,
+      "loss": 0.9799,
+      "step": 28871
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.867694748708133e-06,
+      "loss": 0.8333,
+      "step": 28872
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8670970452479018e-06,
+      "loss": 1.1382,
+      "step": 28873
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8664993936589326e-06,
+      "loss": 1.0342,
+      "step": 28874
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.865901793945576e-06,
+      "loss": 0.8315,
+      "step": 28875
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8653042461121703e-06,
+      "loss": 0.9868,
+      "step": 28876
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.864706750163061e-06,
+      "loss": 0.7852,
+      "step": 28877
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8641093061026006e-06,
+      "loss": 0.8254,
+      "step": 28878
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8635119139351287e-06,
+      "loss": 0.5014,
+      "step": 28879
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.862914573664991e-06,
+      "loss": 0.6955,
+      "step": 28880
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.862317285296531e-06,
+      "loss": 1.0764,
+      "step": 28881
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8617200488340923e-06,
+      "loss": 0.8061,
+      "step": 28882
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8611228642820187e-06,
+      "loss": 0.8191,
+      "step": 28883
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.860525731644649e-06,
+      "loss": 0.8982,
+      "step": 28884
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8599286509263346e-06,
+      "loss": 1.0506,
+      "step": 28885
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8593316221314096e-06,
+      "loss": 0.8074,
+      "step": 28886
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.858734645264217e-06,
+      "loss": 0.8033,
+      "step": 28887
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8581377203291005e-06,
+      "loss": 1.1538,
+      "step": 28888
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8575408473303956e-06,
+      "loss": 0.5322,
+      "step": 28889
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8569440262724502e-06,
+      "loss": 0.7014,
+      "step": 28890
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8563472571596008e-06,
+      "loss": 0.617,
+      "step": 28891
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.855750539996187e-06,
+      "loss": 0.8422,
+      "step": 28892
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8551538747865493e-06,
+      "loss": 0.7822,
+      "step": 28893
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.854557261535026e-06,
+      "loss": 0.6907,
+      "step": 28894
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8539607002459558e-06,
+      "loss": 0.8927,
+      "step": 28895
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8533641909236763e-06,
+      "loss": 0.8295,
+      "step": 28896
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8527677335725267e-06,
+      "loss": 0.8826,
+      "step": 28897
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8521713281968443e-06,
+      "loss": 0.861,
+      "step": 28898
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8515749748009645e-06,
+      "loss": 0.8414,
+      "step": 28899
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8509786733892264e-06,
+      "loss": 0.9013,
+      "step": 28900
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.850382423965965e-06,
+      "loss": 0.8534,
+      "step": 28901
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8497862265355125e-06,
+      "loss": 0.809,
+      "step": 28902
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.849190081102212e-06,
+      "loss": 0.649,
+      "step": 28903
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.848593987670396e-06,
+      "loss": 0.6723,
+      "step": 28904
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8479979462444017e-06,
+      "loss": 0.9051,
+      "step": 28905
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.847401956828557e-06,
+      "loss": 0.9679,
+      "step": 28906
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.846806019427197e-06,
+      "loss": 0.9607,
+      "step": 28907
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8462101340446603e-06,
+      "loss": 0.6407,
+      "step": 28908
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8456143006852786e-06,
+      "loss": 1.004,
+      "step": 28909
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.845018519353384e-06,
+      "loss": 0.9372,
+      "step": 28910
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8444227900533094e-06,
+      "loss": 0.7025,
+      "step": 28911
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8438271127893867e-06,
+      "loss": 0.9195,
+      "step": 28912
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8432314875659473e-06,
+      "loss": 0.9157,
+      "step": 28913
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8426359143873204e-06,
+      "loss": 0.8511,
+      "step": 28914
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8420403932578465e-06,
+      "loss": 0.8852,
+      "step": 28915
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8414449241818477e-06,
+      "loss": 0.8266,
+      "step": 28916
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8408495071636545e-06,
+      "loss": 0.7478,
+      "step": 28917
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8402541422076e-06,
+      "loss": 0.7307,
+      "step": 28918
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8396588293180115e-06,
+      "loss": 1.0163,
+      "step": 28919
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8390635684992163e-06,
+      "loss": 0.9817,
+      "step": 28920
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8384683597555497e-06,
+      "loss": 0.8877,
+      "step": 28921
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.837873203091336e-06,
+      "loss": 0.8116,
+      "step": 28922
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.837278098510904e-06,
+      "loss": 0.7988,
+      "step": 28923
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.83668304601858e-06,
+      "loss": 0.7907,
+      "step": 28924
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8360880456186936e-06,
+      "loss": 0.8323,
+      "step": 28925
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8354930973155702e-06,
+      "loss": 1.0976,
+      "step": 28926
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.834898201113536e-06,
+      "loss": 0.7692,
+      "step": 28927
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8343033570169185e-06,
+      "loss": 0.8861,
+      "step": 28928
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8337085650300423e-06,
+      "loss": 0.8926,
+      "step": 28929
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.833113825157233e-06,
+      "loss": 0.9323,
+      "step": 28930
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.832519137402816e-06,
+      "loss": 0.5574,
+      "step": 28931
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8319245017711127e-06,
+      "loss": 0.9043,
+      "step": 28932
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8313299182664543e-06,
+      "loss": 0.8623,
+      "step": 28933
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.83073538689316e-06,
+      "loss": 0.7864,
+      "step": 28934
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8301409076555574e-06,
+      "loss": 0.7181,
+      "step": 28935
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8295464805579633e-06,
+      "loss": 0.8576,
+      "step": 28936
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8289521056047033e-06,
+      "loss": 0.9667,
+      "step": 28937
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.828357782800096e-06,
+      "loss": 0.9519,
+      "step": 28938
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8277635121484714e-06,
+      "loss": 0.8035,
+      "step": 28939
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.827169293654147e-06,
+      "loss": 0.7005,
+      "step": 28940
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8265751273214438e-06,
+      "loss": 0.9037,
+      "step": 28941
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.825981013154683e-06,
+      "loss": 0.7027,
+      "step": 28942
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8253869511581845e-06,
+      "loss": 0.6898,
+      "step": 28943
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8247929413362684e-06,
+      "loss": 0.924,
+      "step": 28944
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.824198983693256e-06,
+      "loss": 0.5594,
+      "step": 28945
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8236050782334647e-06,
+      "loss": 0.7684,
+      "step": 28946
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.823011224961215e-06,
+      "loss": 0.4795,
+      "step": 28947
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8224174238808245e-06,
+      "loss": 0.9187,
+      "step": 28948
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.821823674996611e-06,
+      "loss": 0.7486,
+      "step": 28949
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.821229978312889e-06,
+      "loss": 1.0717,
+      "step": 28950
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8206363338339837e-06,
+      "loss": 0.878,
+      "step": 28951
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8200427415642083e-06,
+      "loss": 0.8882,
+      "step": 28952
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8194492015078802e-06,
+      "loss": 0.6237,
+      "step": 28953
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.818855713669314e-06,
+      "loss": 0.6856,
+      "step": 28954
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.818262278052829e-06,
+      "loss": 0.8756,
+      "step": 28955
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8176688946627327e-06,
+      "loss": 0.9466,
+      "step": 28956
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.81707556350335e-06,
+      "loss": 0.5683,
+      "step": 28957
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8164822845789895e-06,
+      "loss": 0.9172,
+      "step": 28958
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.815889057893969e-06,
+      "loss": 0.9698,
+      "step": 28959
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8152958834526e-06,
+      "loss": 0.9025,
+      "step": 28960
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.814702761259198e-06,
+      "loss": 0.8688,
+      "step": 28961
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8141096913180744e-06,
+      "loss": 0.6783,
+      "step": 28962
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8135166736335405e-06,
+      "loss": 1.0084,
+      "step": 28963
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.812923708209915e-06,
+      "loss": 0.7295,
+      "step": 28964
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8123307950515087e-06,
+      "loss": 0.9435,
+      "step": 28965
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8117379341626293e-06,
+      "loss": 0.5747,
+      "step": 28966
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8111451255475888e-06,
+      "loss": 0.7425,
+      "step": 28967
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.810552369210696e-06,
+      "loss": 0.9214,
+      "step": 28968
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8099596651562677e-06,
+      "loss": 0.8199,
+      "step": 28969
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8093670133886118e-06,
+      "loss": 0.8651,
+      "step": 28970
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8087744139120376e-06,
+      "loss": 0.823,
+      "step": 28971
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8081818667308536e-06,
+      "loss": 0.8096,
+      "step": 28972
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.807589371849371e-06,
+      "loss": 0.8038,
+      "step": 28973
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8069969292718956e-06,
+      "loss": 0.927,
+      "step": 28974
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8064045390027384e-06,
+      "loss": 0.7427,
+      "step": 28975
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8058122010462065e-06,
+      "loss": 0.8993,
+      "step": 28976
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.805219915406606e-06,
+      "loss": 0.7024,
+      "step": 28977
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8046276820882456e-06,
+      "loss": 0.8992,
+      "step": 28978
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8040355010954323e-06,
+      "loss": 0.6454,
+      "step": 28979
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8034433724324716e-06,
+      "loss": 0.7076,
+      "step": 28980
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.802851296103667e-06,
+      "loss": 0.8359,
+      "step": 28981
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.802259272113329e-06,
+      "loss": 0.8608,
+      "step": 28982
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8016673004657624e-06,
+      "loss": 1.0253,
+      "step": 28983
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.801075381165269e-06,
+      "loss": 0.7579,
+      "step": 28984
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8004835142161592e-06,
+      "loss": 0.8034,
+      "step": 28985
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7998916996227265e-06,
+      "loss": 0.7524,
+      "step": 28986
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7992999373892838e-06,
+      "loss": 0.974,
+      "step": 28987
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7987082275201306e-06,
+      "loss": 0.5564,
+      "step": 28988
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.798116570019572e-06,
+      "loss": 0.5716,
+      "step": 28989
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.797524964891909e-06,
+      "loss": 0.6288,
+      "step": 28990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7969334121414447e-06,
+      "loss": 0.7978,
+      "step": 28991
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.79634191177248e-06,
+      "loss": 0.969,
+      "step": 28992
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.795750463789314e-06,
+      "loss": 0.4776,
+      "step": 28993
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.795159068196254e-06,
+      "loss": 0.9915,
+      "step": 28994
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7945677249976e-06,
+      "loss": 0.5972,
+      "step": 28995
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.793976434197646e-06,
+      "loss": 1.1526,
+      "step": 28996
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7933851958006964e-06,
+      "loss": 0.6348,
+      "step": 28997
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7927940098110505e-06,
+      "loss": 0.9297,
+      "step": 28998
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7922028762330035e-06,
+      "loss": 0.8771,
+      "step": 28999
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.79161179507086e-06,
+      "loss": 0.8187,
+      "step": 29000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.791020766328916e-06,
+      "loss": 0.8476,
+      "step": 29001
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7904297900114695e-06,
+      "loss": 0.889,
+      "step": 29002
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7898388661228195e-06,
+      "loss": 0.9764,
+      "step": 29003
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.78924799466726e-06,
+      "loss": 0.8228,
+      "step": 29004
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.788657175649091e-06,
+      "loss": 0.6251,
+      "step": 29005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7880664090726073e-06,
+      "loss": 0.7733,
+      "step": 29006
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7874756949421057e-06,
+      "loss": 0.8633,
+      "step": 29007
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7868850332618824e-06,
+      "loss": 0.6859,
+      "step": 29008
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.786294424036231e-06,
+      "loss": 1.0137,
+      "step": 29009
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7857038672694492e-06,
+      "loss": 0.741,
+      "step": 29010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7851133629658257e-06,
+      "loss": 1.082,
+      "step": 29011
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7845229111296625e-06,
+      "loss": 0.9343,
+      "step": 29012
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7839325117652494e-06,
+      "loss": 0.9825,
+      "step": 29013
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.783342164876881e-06,
+      "loss": 0.8211,
+      "step": 29014
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7827518704688526e-06,
+      "loss": 0.7461,
+      "step": 29015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.782161628545451e-06,
+      "loss": 1.0767,
+      "step": 29016
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.781571439110968e-06,
+      "loss": 0.8947,
+      "step": 29017
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.780981302169703e-06,
+      "loss": 0.9005,
+      "step": 29018
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7803912177259427e-06,
+      "loss": 0.6996,
+      "step": 29019
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7798011857839802e-06,
+      "loss": 0.9655,
+      "step": 29020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.779211206348105e-06,
+      "loss": 0.9643,
+      "step": 29021
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7786212794226074e-06,
+      "loss": 0.6233,
+      "step": 29022
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7780314050117786e-06,
+      "loss": 0.7571,
+      "step": 29023
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.777441583119903e-06,
+      "loss": 0.6803,
+      "step": 29024
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.776851813751281e-06,
+      "loss": 0.823,
+      "step": 29025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.776262096910193e-06,
+      "loss": 0.4486,
+      "step": 29026
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.775672432600928e-06,
+      "loss": 0.9117,
+      "step": 29027
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7750828208277747e-06,
+      "loss": 0.8387,
+      "step": 29028
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.774493261595018e-06,
+      "loss": 0.8774,
+      "step": 29029
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7739037549069534e-06,
+      "loss": 0.7794,
+      "step": 29030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.773314300767862e-06,
+      "loss": 0.8598,
+      "step": 29031
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.772724899182032e-06,
+      "loss": 0.8705,
+      "step": 29032
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7721355501537493e-06,
+      "loss": 0.6644,
+      "step": 29033
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7715462536872995e-06,
+      "loss": 0.8568,
+      "step": 29034
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7709570097869678e-06,
+      "loss": 1.0953,
+      "step": 29035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.770367818457039e-06,
+      "loss": 0.9687,
+      "step": 29036
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.769778679701799e-06,
+      "loss": 0.8211,
+      "step": 29037
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.769189593525531e-06,
+      "loss": 1.0303,
+      "step": 29038
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7686005599325194e-06,
+      "loss": 0.7991,
+      "step": 29039
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7680115789270478e-06,
+      "loss": 0.9354,
+      "step": 29040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7674226505133983e-06,
+      "loss": 0.9583,
+      "step": 29041
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.766833774695852e-06,
+      "loss": 1.0383,
+      "step": 29042
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7662449514786973e-06,
+      "loss": 0.8123,
+      "step": 29043
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.765656180866211e-06,
+      "loss": 0.7018,
+      "step": 29044
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7650674628626805e-06,
+      "loss": 0.8432,
+      "step": 29045
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7644787974723797e-06,
+      "loss": 0.7056,
+      "step": 29046
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.763890184699589e-06,
+      "loss": 0.5455,
+      "step": 29047
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7633016245485955e-06,
+      "loss": 0.8614,
+      "step": 29048
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.762713117023678e-06,
+      "loss": 0.8911,
+      "step": 29049
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.762124662129112e-06,
+      "loss": 0.7974,
+      "step": 29050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7615362598691795e-06,
+      "loss": 0.7634,
+      "step": 29051
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.760947910248161e-06,
+      "loss": 0.6965,
+      "step": 29052
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7603596132703316e-06,
+      "loss": 0.8854,
+      "step": 29053
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7597713689399687e-06,
+      "loss": 0.8027,
+      "step": 29054
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7591831772613576e-06,
+      "loss": 1.0552,
+      "step": 29055
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.758595038238768e-06,
+      "loss": 0.8743,
+      "step": 29056
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7580069518764796e-06,
+      "loss": 0.8109,
+      "step": 29057
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.757418918178769e-06,
+      "loss": 0.9653,
+      "step": 29058
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.756830937149911e-06,
+      "loss": 0.8444,
+      "step": 29059
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7562430087941816e-06,
+      "loss": 0.7514,
+      "step": 29060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7556551331158588e-06,
+      "loss": 0.881,
+      "step": 29061
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7550673101192172e-06,
+      "loss": 0.762,
+      "step": 29062
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.754479539808531e-06,
+      "loss": 0.7585,
+      "step": 29063
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.753891822188074e-06,
+      "loss": 0.812,
+      "step": 29064
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.75330415726212e-06,
+      "loss": 0.9084,
+      "step": 29065
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7527165450349423e-06,
+      "loss": 1.0613,
+      "step": 29066
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.752128985510816e-06,
+      "loss": 0.805,
+      "step": 29067
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.751541478694012e-06,
+      "loss": 0.5511,
+      "step": 29068
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.750954024588803e-06,
+      "loss": 0.6791,
+      "step": 29069
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.750366623199462e-06,
+      "loss": 0.9938,
+      "step": 29070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7497792745302586e-06,
+      "loss": 1.2429,
+      "step": 29071
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.749191978585464e-06,
+      "loss": 1.0622,
+      "step": 29072
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7486047353693523e-06,
+      "loss": 0.8495,
+      "step": 29073
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7480175448861933e-06,
+      "loss": 1.2775,
+      "step": 29074
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.747430407140258e-06,
+      "loss": 0.6325,
+      "step": 29075
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7468433221358115e-06,
+      "loss": 0.7057,
+      "step": 29076
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.746256289877126e-06,
+      "loss": 0.839,
+      "step": 29077
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7456693103684675e-06,
+      "loss": 0.714,
+      "step": 29078
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7450823836141107e-06,
+      "loss": 0.8482,
+      "step": 29079
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.744495509618319e-06,
+      "loss": 0.8968,
+      "step": 29080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7439086883853626e-06,
+      "loss": 0.6934,
+      "step": 29081
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.743321919919508e-06,
+      "loss": 0.6827,
+      "step": 29082
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.742735204225022e-06,
+      "loss": 0.8144,
+      "step": 29083
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.742148541306172e-06,
+      "loss": 0.9347,
+      "step": 29084
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7415619311672236e-06,
+      "loss": 0.5908,
+      "step": 29085
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7409753738124423e-06,
+      "loss": 0.5958,
+      "step": 29086
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7403888692460946e-06,
+      "loss": 0.9193,
+      "step": 29087
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7398024174724447e-06,
+      "loss": 0.9338,
+      "step": 29088
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7392160184957585e-06,
+      "loss": 0.7587,
+      "step": 29089
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7386296723202966e-06,
+      "loss": 0.6207,
+      "step": 29090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.738043378950329e-06,
+      "loss": 0.8181,
+      "step": 29091
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.737457138390116e-06,
+      "loss": 0.9279,
+      "step": 29092
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7368709506439215e-06,
+      "loss": 0.7344,
+      "step": 29093
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7362848157160084e-06,
+      "loss": 0.9646,
+      "step": 29094
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.735698733610641e-06,
+      "loss": 1.0281,
+      "step": 29095
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7351127043320735e-06,
+      "loss": 0.716,
+      "step": 29096
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7345267278845767e-06,
+      "loss": 0.6419,
+      "step": 29097
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.733940804272407e-06,
+      "loss": 1.0949,
+      "step": 29098
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7333549334998276e-06,
+      "loss": 0.8974,
+      "step": 29099
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7327691155710978e-06,
+      "loss": 1.028,
+      "step": 29100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.732183350490478e-06,
+      "loss": 0.6597,
+      "step": 29101
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7315976382622278e-06,
+      "loss": 0.7068,
+      "step": 29102
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.731011978890604e-06,
+      "loss": 0.765,
+      "step": 29103
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.730426372379872e-06,
+      "loss": 1.0063,
+      "step": 29104
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7298408187342884e-06,
+      "loss": 0.9191,
+      "step": 29105
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.729255317958107e-06,
+      "loss": 0.6211,
+      "step": 29106
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.728669870055589e-06,
+      "loss": 0.7035,
+      "step": 29107
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7280844750309877e-06,
+      "loss": 1.0963,
+      "step": 29108
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.727499132888566e-06,
+      "loss": 0.8149,
+      "step": 29109
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7269138436325795e-06,
+      "loss": 0.5623,
+      "step": 29110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7263286072672823e-06,
+      "loss": 0.8293,
+      "step": 29111
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7257434237969306e-06,
+      "loss": 0.7415,
+      "step": 29112
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7251582932257813e-06,
+      "loss": 0.9573,
+      "step": 29113
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7245732155580893e-06,
+      "loss": 0.964,
+      "step": 29114
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.723988190798108e-06,
+      "loss": 0.8114,
+      "step": 29115
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7234032189500924e-06,
+      "loss": 0.9114,
+      "step": 29116
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.722818300018297e-06,
+      "loss": 0.8101,
+      "step": 29117
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7222334340069756e-06,
+      "loss": 0.9249,
+      "step": 29118
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7216486209203806e-06,
+      "loss": 0.776,
+      "step": 29119
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.721063860762765e-06,
+      "loss": 0.6688,
+      "step": 29120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.720479153538378e-06,
+      "loss": 1.0013,
+      "step": 29121
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7198944992514774e-06,
+      "loss": 1.0163,
+      "step": 29122
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7193098979063127e-06,
+      "loss": 0.9794,
+      "step": 29123
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7187253495071354e-06,
+      "loss": 0.8132,
+      "step": 29124
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.718140854058199e-06,
+      "loss": 0.7828,
+      "step": 29125
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7175564115637453e-06,
+      "loss": 0.7724,
+      "step": 29126
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7169720220280325e-06,
+      "loss": 0.7352,
+      "step": 29127
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7163876854553074e-06,
+      "loss": 1.0189,
+      "step": 29128
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7158034018498204e-06,
+      "loss": 0.7219,
+      "step": 29129
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7152191712158207e-06,
+      "loss": 0.8531,
+      "step": 29130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7146349935575556e-06,
+      "loss": 0.7663,
+      "step": 29131
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.714050868879273e-06,
+      "loss": 0.8899,
+      "step": 29132
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.713466797185219e-06,
+      "loss": 0.725,
+      "step": 29133
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7128827784796475e-06,
+      "loss": 0.8618,
+      "step": 29134
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.712298812766804e-06,
+      "loss": 0.7848,
+      "step": 29135
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.711714900050929e-06,
+      "loss": 0.8549,
+      "step": 29136
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7111310403362735e-06,
+      "loss": 0.6232,
+      "step": 29137
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7105472336270833e-06,
+      "loss": 0.692,
+      "step": 29138
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7099634799275986e-06,
+      "loss": 0.6618,
+      "step": 29139
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.709379779242073e-06,
+      "loss": 0.791,
+      "step": 29140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.708796131574747e-06,
+      "loss": 0.8102,
+      "step": 29141
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.708212536929865e-06,
+      "loss": 0.8409,
+      "step": 29142
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.707628995311671e-06,
+      "loss": 0.906,
+      "step": 29143
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.707045506724408e-06,
+      "loss": 0.9216,
+      "step": 29144
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.706462071172322e-06,
+      "loss": 0.6481,
+      "step": 29145
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7058786886596523e-06,
+      "loss": 0.9541,
+      "step": 29146
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.705295359190643e-06,
+      "loss": 0.5818,
+      "step": 29147
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7047120827695363e-06,
+      "loss": 0.7688,
+      "step": 29148
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.704128859400573e-06,
+      "loss": 0.7877,
+      "step": 29149
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7035456890879954e-06,
+      "loss": 1.0624,
+      "step": 29150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.70296257183604e-06,
+      "loss": 0.6471,
+      "step": 29151
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7023795076489543e-06,
+      "loss": 0.8567,
+      "step": 29152
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.701796496530975e-06,
+      "loss": 0.5491,
+      "step": 29153
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7012135384863424e-06,
+      "loss": 0.8903,
+      "step": 29154
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.700630633519299e-06,
+      "loss": 0.8977,
+      "step": 29155
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7000477816340766e-06,
+      "loss": 0.7852,
+      "step": 29156
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6994649828349153e-06,
+      "loss": 0.7161,
+      "step": 29157
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.698882237126057e-06,
+      "loss": 1.0631,
+      "step": 29158
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6982995445117376e-06,
+      "loss": 1.2542,
+      "step": 29159
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.697716904996196e-06,
+      "loss": 0.7902,
+      "step": 29160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6971343185836673e-06,
+      "loss": 1.0165,
+      "step": 29161
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6965517852783897e-06,
+      "loss": 0.6223,
+      "step": 29162
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.695969305084597e-06,
+      "loss": 0.7433,
+      "step": 29163
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6953868780065252e-06,
+      "loss": 0.8227,
+      "step": 29164
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6948045040484162e-06,
+      "loss": 0.9305,
+      "step": 29165
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6942221832144968e-06,
+      "loss": 1.0673,
+      "step": 29166
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6936399155090054e-06,
+      "loss": 0.9203,
+      "step": 29167
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.693057700936176e-06,
+      "loss": 0.5526,
+      "step": 29168
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.692475539500239e-06,
+      "loss": 0.8101,
+      "step": 29169
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6918934312054347e-06,
+      "loss": 0.8543,
+      "step": 29170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.691311376055993e-06,
+      "loss": 0.7505,
+      "step": 29171
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.690729374056147e-06,
+      "loss": 0.6793,
+      "step": 29172
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.690147425210128e-06,
+      "loss": 1.0832,
+      "step": 29173
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.689565529522169e-06,
+      "loss": 1.0189,
+      "step": 29174
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6889836869965016e-06,
+      "loss": 0.8647,
+      "step": 29175
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.688401897637355e-06,
+      "loss": 1.0282,
+      "step": 29176
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.687820161448963e-06,
+      "loss": 0.7359,
+      "step": 29177
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6872384784355553e-06,
+      "loss": 0.7828,
+      "step": 29178
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.686656848601361e-06,
+      "loss": 0.957,
+      "step": 29179
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68607527195061e-06,
+      "loss": 0.5861,
+      "step": 29180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.685493748487532e-06,
+      "loss": 0.6163,
+      "step": 29181
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6849122782163527e-06,
+      "loss": 0.9785,
+      "step": 29182
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6843308611413066e-06,
+      "loss": 0.9851,
+      "step": 29183
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6837494972666177e-06,
+      "loss": 0.8073,
+      "step": 29184
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68316818659652e-06,
+      "loss": 0.8674,
+      "step": 29185
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.682586929135231e-06,
+      "loss": 0.7481,
+      "step": 29186
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68200572488698e-06,
+      "loss": 0.8405,
+      "step": 29187
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6814245738559996e-06,
+      "loss": 0.9268,
+      "step": 29188
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.680843476046512e-06,
+      "loss": 0.6694,
+      "step": 29189
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6802624314627436e-06,
+      "loss": 0.7425,
+      "step": 29190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.67968144010892e-06,
+      "loss": 0.9691,
+      "step": 29191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6791005019892645e-06,
+      "loss": 1.1233,
+      "step": 29192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.678519617108004e-06,
+      "loss": 0.639,
+      "step": 29193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6779387854693594e-06,
+      "loss": 0.7117,
+      "step": 29194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.677358007077563e-06,
+      "loss": 0.8972,
+      "step": 29195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6767772819368286e-06,
+      "loss": 0.8619,
+      "step": 29196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6761966100513824e-06,
+      "loss": 0.6841,
+      "step": 29197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6756159914254485e-06,
+      "loss": 0.6962,
+      "step": 29198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6750354260632494e-06,
+      "loss": 0.7657,
+      "step": 29199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6744549139690025e-06,
+      "loss": 0.8316,
+      "step": 29200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6738744551469353e-06,
+      "loss": 0.933,
+      "step": 29201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.673294049601267e-06,
+      "loss": 0.8212,
+      "step": 29202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6727136973362177e-06,
+      "loss": 0.8104,
+      "step": 29203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6721333983560084e-06,
+      "loss": 0.8144,
+      "step": 29204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6715531526648585e-06,
+      "loss": 0.7049,
+      "step": 29205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6709729602669887e-06,
+      "loss": 1.0441,
+      "step": 29206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6703928211666165e-06,
+      "loss": 0.9624,
+      "step": 29207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6698127353679625e-06,
+      "loss": 0.8795,
+      "step": 29208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.669232702875244e-06,
+      "loss": 0.7977,
+      "step": 29209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6686527236926787e-06,
+      "loss": 0.5249,
+      "step": 29210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6680727978244847e-06,
+      "loss": 0.7124,
+      "step": 29211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.667492925274877e-06,
+      "loss": 0.9204,
+      "step": 29212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.666913106048079e-06,
+      "loss": 0.7647,
+      "step": 29213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6663333401483015e-06,
+      "loss": 0.6709,
+      "step": 29214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.665753627579767e-06,
+      "loss": 0.8103,
+      "step": 29215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6651739683466825e-06,
+      "loss": 0.7545,
+      "step": 29216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6645943624532676e-06,
+      "loss": 0.6634,
+      "step": 29217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6640148099037343e-06,
+      "loss": 0.9466,
+      "step": 29218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6634353107023027e-06,
+      "loss": 0.977,
+      "step": 29219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6628558648531845e-06,
+      "loss": 0.9521,
+      "step": 29220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.662276472360593e-06,
+      "loss": 0.6811,
+      "step": 29221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.661697133228741e-06,
+      "loss": 0.708,
+      "step": 29222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6611178474618437e-06,
+      "loss": 0.822,
+      "step": 29223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.660538615064112e-06,
+      "loss": 0.7707,
+      "step": 29224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6599594360397586e-06,
+      "loss": 0.6494,
+      "step": 29225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6593803103929947e-06,
+      "loss": 0.9306,
+      "step": 29226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6588012381280327e-06,
+      "loss": 1.0206,
+      "step": 29227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6582222192490836e-06,
+      "loss": 0.8901,
+      "step": 29228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.657643253760358e-06,
+      "loss": 0.8012,
+      "step": 29229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6570643416660623e-06,
+      "loss": 0.9333,
+      "step": 29230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6564854829704134e-06,
+      "loss": 0.6605,
+      "step": 29231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6559066776776177e-06,
+      "loss": 0.9247,
+      "step": 29232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.655327925791883e-06,
+      "loss": 0.8113,
+      "step": 29233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6547492273174202e-06,
+      "loss": 0.6889,
+      "step": 29234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.654170582258441e-06,
+      "loss": 0.9936,
+      "step": 29235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.653591990619141e-06,
+      "loss": 0.7798,
+      "step": 29236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.65301345240374e-06,
+      "loss": 0.9076,
+      "step": 29237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6524349676164406e-06,
+      "loss": 1.0475,
+      "step": 29238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.65185653626145e-06,
+      "loss": 0.8057,
+      "step": 29239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6512781583429748e-06,
+      "loss": 1.1404,
+      "step": 29240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.65069983386522e-06,
+      "loss": 0.9375,
+      "step": 29241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.650121562832394e-06,
+      "loss": 0.8436,
+      "step": 29242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.649543345248695e-06,
+      "loss": 0.7463,
+      "step": 29243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6489651811183383e-06,
+      "loss": 0.8673,
+      "step": 29244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.648387070445525e-06,
+      "loss": 0.84,
+      "step": 29245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6478090132344535e-06,
+      "loss": 0.8801,
+      "step": 29246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6472310094893326e-06,
+      "loss": 0.7084,
+      "step": 29247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6466530592143604e-06,
+      "loss": 0.8639,
+      "step": 29248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.646075162413747e-06,
+      "loss": 1.1037,
+      "step": 29249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.645497319091692e-06,
+      "loss": 0.6637,
+      "step": 29250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.644919529252398e-06,
+      "loss": 0.5763,
+      "step": 29251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.644341792900064e-06,
+      "loss": 0.7172,
+      "step": 29252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6437641100388945e-06,
+      "loss": 0.9198,
+      "step": 29253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.643186480673089e-06,
+      "loss": 0.5897,
+      "step": 29254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.642608904806848e-06,
+      "loss": 0.9777,
+      "step": 29255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.642031382444372e-06,
+      "loss": 1.1729,
+      "step": 29256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.641453913589862e-06,
+      "loss": 0.7062,
+      "step": 29257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6408764982475154e-06,
+      "loss": 0.9457,
+      "step": 29258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6402991364215325e-06,
+      "loss": 0.7185,
+      "step": 29259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.639721828116112e-06,
+      "loss": 1.15,
+      "step": 29260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.639144573335447e-06,
+      "loss": 0.5489,
+      "step": 29261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6385673720837447e-06,
+      "loss": 0.5819,
+      "step": 29262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.637990224365198e-06,
+      "loss": 0.8485,
+      "step": 29263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.637413130184002e-06,
+      "loss": 0.5543,
+      "step": 29264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.63683608954436e-06,
+      "loss": 0.6563,
+      "step": 29265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.636259102450458e-06,
+      "loss": 0.9735,
+      "step": 29266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6356821689064995e-06,
+      "loss": 0.8713,
+      "step": 29267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6351052889166782e-06,
+      "loss": 0.8435,
+      "step": 29268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.63452846248519e-06,
+      "loss": 0.7168,
+      "step": 29269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6339516896162276e-06,
+      "loss": 0.875,
+      "step": 29270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6333749703139868e-06,
+      "loss": 0.7722,
+      "step": 29271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.632798304582661e-06,
+      "loss": 0.9235,
+      "step": 29272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.632221692426441e-06,
+      "loss": 0.9321,
+      "step": 29273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6316451338495263e-06,
+      "loss": 0.9438,
+      "step": 29274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6310686288561096e-06,
+      "loss": 0.7606,
+      "step": 29275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.630492177450378e-06,
+      "loss": 0.5933,
+      "step": 29276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.629915779636524e-06,
+      "loss": 0.6216,
+      "step": 29277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6293394354187417e-06,
+      "loss": 0.7317,
+      "step": 29278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6287631448012185e-06,
+      "loss": 0.8517,
+      "step": 29279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6281869077881507e-06,
+      "loss": 0.8725,
+      "step": 29280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6276107243837267e-06,
+      "loss": 1.069,
+      "step": 29281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.627034594592136e-06,
+      "loss": 0.9566,
+      "step": 29282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6264585184175693e-06,
+      "loss": 0.7265,
+      "step": 29283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.625882495864215e-06,
+      "loss": 0.5377,
+      "step": 29284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.625306526936261e-06,
+      "loss": 0.7813,
+      "step": 29285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6247306116378978e-06,
+      "loss": 0.6978,
+      "step": 29286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6241547499733123e-06,
+      "loss": 0.8797,
+      "step": 29287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6235789419466918e-06,
+      "loss": 0.9733,
+      "step": 29288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6230031875622253e-06,
+      "loss": 0.8456,
+      "step": 29289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6224274868240983e-06,
+      "loss": 0.9591,
+      "step": 29290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.621851839736498e-06,
+      "loss": 0.8606,
+      "step": 29291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6212762463036078e-06,
+      "loss": 0.9633,
+      "step": 29292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6207007065296188e-06,
+      "loss": 0.7111,
+      "step": 29293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6201252204187143e-06,
+      "loss": 1.0197,
+      "step": 29294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.619549787975081e-06,
+      "loss": 0.7249,
+      "step": 29295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.618974409202898e-06,
+      "loss": 0.8707,
+      "step": 29296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.618399084106351e-06,
+      "loss": 0.6701,
+      "step": 29297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6178238126896273e-06,
+      "loss": 0.9295,
+      "step": 29298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.617248594956909e-06,
+      "loss": 0.7484,
+      "step": 29299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.616673430912379e-06,
+      "loss": 0.8611,
+      "step": 29300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6160983205602196e-06,
+      "loss": 0.99,
+      "step": 29301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6155232639046134e-06,
+      "loss": 0.8898,
+      "step": 29302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.614948260949741e-06,
+      "loss": 1.0686,
+      "step": 29303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.614373311699783e-06,
+      "loss": 0.7369,
+      "step": 29304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.613798416158928e-06,
+      "loss": 0.7016,
+      "step": 29305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6132235743313483e-06,
+      "loss": 0.4684,
+      "step": 29306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6126487862212267e-06,
+      "loss": 0.6289,
+      "step": 29307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6120740518327447e-06,
+      "loss": 0.9241,
+      "step": 29308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6114993711700786e-06,
+      "loss": 0.7073,
+      "step": 29309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6109247442374088e-06,
+      "loss": 0.564,
+      "step": 29310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6103501710389155e-06,
+      "loss": 0.9685,
+      "step": 29311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6097756515787775e-06,
+      "loss": 0.7415,
+      "step": 29312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.609201185861172e-06,
+      "loss": 0.9629,
+      "step": 29313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.608626773890276e-06,
+      "loss": 1.1075,
+      "step": 29314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.608052415670267e-06,
+      "loss": 0.6345,
+      "step": 29315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6074781112053216e-06,
+      "loss": 0.6703,
+      "step": 29316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6069038604996156e-06,
+      "loss": 0.6416,
+      "step": 29317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6063296635573266e-06,
+      "loss": 0.7687,
+      "step": 29318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.605755520382629e-06,
+      "loss": 0.7774,
+      "step": 29319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6051814309796976e-06,
+      "loss": 0.7111,
+      "step": 29320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6046073953527094e-06,
+      "loss": 0.5571,
+      "step": 29321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.604033413505833e-06,
+      "loss": 0.7057,
+      "step": 29322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6034594854432515e-06,
+      "loss": 0.6607,
+      "step": 29323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6028856111691327e-06,
+      "loss": 0.9111,
+      "step": 29324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.602311790687655e-06,
+      "loss": 1.0481,
+      "step": 29325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6017380240029844e-06,
+      "loss": 0.9734,
+      "step": 29326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6011643111192953e-06,
+      "loss": 0.8279,
+      "step": 29327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6005906520407585e-06,
+      "loss": 0.8177,
+      "step": 29328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6000170467715525e-06,
+      "loss": 0.5554,
+      "step": 29329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5994434953158433e-06,
+      "loss": 0.736,
+      "step": 29330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.598869997677802e-06,
+      "loss": 0.7138,
+      "step": 29331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5982965538616013e-06,
+      "loss": 0.6979,
+      "step": 29332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5977231638714097e-06,
+      "loss": 0.9849,
+      "step": 29333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.597149827711396e-06,
+      "loss": 0.7577,
+      "step": 29334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.596576545385732e-06,
+      "loss": 0.983,
+      "step": 29335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5960033168985855e-06,
+      "loss": 0.8716,
+      "step": 29336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.595430142254125e-06,
+      "loss": 0.8954,
+      "step": 29337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5948570214565183e-06,
+      "loss": 0.9045,
+      "step": 29338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.594283954509934e-06,
+      "loss": 0.8151,
+      "step": 29339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.593710941418537e-06,
+      "loss": 0.6669,
+      "step": 29340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5931379821864977e-06,
+      "loss": 0.8052,
+      "step": 29341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.592565076817982e-06,
+      "loss": 0.4684,
+      "step": 29342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.591992225317157e-06,
+      "loss": 0.991,
+      "step": 29343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5914194276881865e-06,
+      "loss": 0.6218,
+      "step": 29344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5908466839352386e-06,
+      "loss": 1.2345,
+      "step": 29345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.590273994062471e-06,
+      "loss": 0.8063,
+      "step": 29346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.589701358074057e-06,
+      "loss": 0.6247,
+      "step": 29347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.589128775974158e-06,
+      "loss": 0.7176,
+      "step": 29348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.588556247766937e-06,
+      "loss": 1.0318,
+      "step": 29349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5879837734565583e-06,
+      "loss": 0.6164,
+      "step": 29350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5874113530471844e-06,
+      "loss": 0.9076,
+      "step": 29351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.586838986542978e-06,
+      "loss": 0.7447,
+      "step": 29352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.586266673948099e-06,
+      "loss": 0.7295,
+      "step": 29353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5856944152667153e-06,
+      "loss": 0.6743,
+      "step": 29354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.585122210502987e-06,
+      "loss": 0.7652,
+      "step": 29355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.584550059661072e-06,
+      "loss": 0.8799,
+      "step": 29356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.583977962745131e-06,
+      "loss": 0.9592,
+      "step": 29357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.583405919759324e-06,
+      "loss": 0.7992,
+      "step": 29358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5828339307078156e-06,
+      "loss": 0.8563,
+      "step": 29359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5822619955947613e-06,
+      "loss": 0.6601,
+      "step": 29360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.581690114424322e-06,
+      "loss": 0.6663,
+      "step": 29361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5811182872006567e-06,
+      "loss": 0.8736,
+      "step": 29362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.580546513927923e-06,
+      "loss": 0.837,
+      "step": 29363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.579974794610278e-06,
+      "loss": 0.9525,
+      "step": 29364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.579403129251882e-06,
+      "loss": 0.7881,
+      "step": 29365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.578831517856889e-06,
+      "loss": 0.9424,
+      "step": 29366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.578259960429458e-06,
+      "loss": 0.9075,
+      "step": 29367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5776884569737435e-06,
+      "loss": 0.8811,
+      "step": 29368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5771170074939043e-06,
+      "loss": 0.8574,
+      "step": 29369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5765456119940933e-06,
+      "loss": 1.085,
+      "step": 29370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5759742704784652e-06,
+      "loss": 0.9093,
+      "step": 29371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5754029829511786e-06,
+      "loss": 0.7817,
+      "step": 29372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5748317494163868e-06,
+      "loss": 0.7831,
+      "step": 29373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.574260569878242e-06,
+      "loss": 0.9304,
+      "step": 29374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.573689444340903e-06,
+      "loss": 0.6898,
+      "step": 29375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.573118372808513e-06,
+      "loss": 0.7619,
+      "step": 29376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.572547355285233e-06,
+      "loss": 0.7172,
+      "step": 29377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.571976391775214e-06,
+      "loss": 0.7917,
+      "step": 29378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5714054822826073e-06,
+      "loss": 0.9809,
+      "step": 29379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5708346268115648e-06,
+      "loss": 0.6291,
+      "step": 29380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.570263825366238e-06,
+      "loss": 0.6305,
+      "step": 29381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5696930779507787e-06,
+      "loss": 1.018,
+      "step": 29382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5691223845693323e-06,
+      "loss": 1.1359,
+      "step": 29383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5685517452260566e-06,
+      "loss": 0.7583,
+      "step": 29384
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5679811599251003e-06,
+      "loss": 0.7428,
+      "step": 29385
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.567410628670608e-06,
+      "loss": 0.7339,
+      "step": 29386
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5668401514667306e-06,
+      "loss": 0.7658,
+      "step": 29387
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5662697283176176e-06,
+      "loss": 0.9022,
+      "step": 29388
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.565699359227413e-06,
+      "loss": 1.083,
+      "step": 29389
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.565129044200271e-06,
+      "loss": 0.9459,
+      "step": 29390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5645587832403363e-06,
+      "loss": 0.9492,
+      "step": 29391
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.563988576351756e-06,
+      "loss": 1.0494,
+      "step": 29392
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5634184235386773e-06,
+      "loss": 0.7403,
+      "step": 29393
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.562848324805245e-06,
+      "loss": 0.7943,
+      "step": 29394
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5622782801556055e-06,
+      "loss": 0.9026,
+      "step": 29395
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.561708289593904e-06,
+      "loss": 0.8839,
+      "step": 29396
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.561138353124286e-06,
+      "loss": 0.9065,
+      "step": 29397
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.560568470750896e-06,
+      "loss": 0.7477,
+      "step": 29398
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.559998642477879e-06,
+      "loss": 1.0829,
+      "step": 29399
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.559428868309377e-06,
+      "loss": 0.7226,
+      "step": 29400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.558859148249532e-06,
+      "loss": 0.693,
+      "step": 29401
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.558289482302492e-06,
+      "loss": 0.7497,
+      "step": 29402
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5577198704723983e-06,
+      "loss": 0.9254,
+      "step": 29403
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.557150312763391e-06,
+      "loss": 1.041,
+      "step": 29404
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5565808091796163e-06,
+      "loss": 0.7819,
+      "step": 29405
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.55601135972521e-06,
+      "loss": 1.2251,
+      "step": 29406
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.555441964404313e-06,
+      "loss": 0.7033,
+      "step": 29407
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.554872623221072e-06,
+      "loss": 0.7451,
+      "step": 29408
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.554303336179623e-06,
+      "loss": 0.786,
+      "step": 29409
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.553734103284108e-06,
+      "loss": 0.6144,
+      "step": 29410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.553164924538665e-06,
+      "loss": 0.6647,
+      "step": 29411
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5525957999474326e-06,
+      "loss": 0.8737,
+      "step": 29412
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.552026729514552e-06,
+      "loss": 0.4973,
+      "step": 29413
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5514577132441567e-06,
+      "loss": 0.9554,
+      "step": 29414
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5508887511403936e-06,
+      "loss": 0.7466,
+      "step": 29415
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.550319843207392e-06,
+      "loss": 0.6189,
+      "step": 29416
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5497509894492922e-06,
+      "loss": 0.9769,
+      "step": 29417
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5491821898702295e-06,
+      "loss": 0.3638,
+      "step": 29418
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5486134444743394e-06,
+      "loss": 0.8632,
+      "step": 29419
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5480447532657625e-06,
+      "loss": 0.9359,
+      "step": 29420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5474761162486304e-06,
+      "loss": 0.7715,
+      "step": 29421
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.546907533427081e-06,
+      "loss": 0.9689,
+      "step": 29422
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5463390048052472e-06,
+      "loss": 0.9284,
+      "step": 29423
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.545770530387265e-06,
+      "loss": 0.7527,
+      "step": 29424
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5452021101772662e-06,
+      "loss": 0.784,
+      "step": 29425
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5446337441793857e-06,
+      "loss": 0.9443,
+      "step": 29426
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5440654323977555e-06,
+      "loss": 0.6585,
+      "step": 29427
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.543497174836511e-06,
+      "loss": 1.0662,
+      "step": 29428
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5429289714997816e-06,
+      "loss": 0.5887,
+      "step": 29429
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.542360822391702e-06,
+      "loss": 0.8411,
+      "step": 29430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5417927275164025e-06,
+      "loss": 0.8649,
+      "step": 29431
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.541224686878011e-06,
+      "loss": 0.7693,
+      "step": 29432
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5406567004806647e-06,
+      "loss": 0.7771,
+      "step": 29433
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5400887683284925e-06,
+      "loss": 0.7079,
+      "step": 29434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5395208904256253e-06,
+      "loss": 0.8569,
+      "step": 29435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5389530667761875e-06,
+      "loss": 0.9304,
+      "step": 29436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5383852973843082e-06,
+      "loss": 0.7821,
+      "step": 29437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5378175822541227e-06,
+      "loss": 0.919,
+      "step": 29438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5372499213897573e-06,
+      "loss": 0.8949,
+      "step": 29439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.536682314795339e-06,
+      "loss": 0.966,
+      "step": 29440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.536114762474995e-06,
+      "loss": 0.5889,
+      "step": 29441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5355472644328538e-06,
+      "loss": 0.9348,
+      "step": 29442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5349798206730413e-06,
+      "loss": 0.8752,
+      "step": 29443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5344124311996822e-06,
+      "loss": 0.4217,
+      "step": 29444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5338450960169105e-06,
+      "loss": 0.4757,
+      "step": 29445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.533277815128844e-06,
+      "loss": 0.7956,
+      "step": 29446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5327105885396097e-06,
+      "loss": 0.8221,
+      "step": 29447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.532143416253333e-06,
+      "loss": 0.7766,
+      "step": 29448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5315762982741386e-06,
+      "loss": 0.4315,
+      "step": 29449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.531009234606149e-06,
+      "loss": 0.722,
+      "step": 29450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.530442225253492e-06,
+      "loss": 0.8095,
+      "step": 29451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.529875270220288e-06,
+      "loss": 0.7823,
+      "step": 29452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.529308369510661e-06,
+      "loss": 0.7296,
+      "step": 29453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5287415231287325e-06,
+      "loss": 0.96,
+      "step": 29454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5281747310786264e-06,
+      "loss": 0.8221,
+      "step": 29455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5276079933644637e-06,
+      "loss": 0.7514,
+      "step": 29456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5270413099903644e-06,
+      "loss": 0.7454,
+      "step": 29457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.52647468096045e-06,
+      "loss": 0.842,
+      "step": 29458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5259081062788426e-06,
+      "loss": 0.9628,
+      "step": 29459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.525341585949662e-06,
+      "loss": 0.8573,
+      "step": 29460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5247751199770267e-06,
+      "loss": 0.721,
+      "step": 29461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5242087083650536e-06,
+      "loss": 0.9599,
+      "step": 29462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5236423511178687e-06,
+      "loss": 0.6464,
+      "step": 29463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5230760482395854e-06,
+      "loss": 0.93,
+      "step": 29464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5225097997343272e-06,
+      "loss": 0.7185,
+      "step": 29465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5219436056062063e-06,
+      "loss": 0.7991,
+      "step": 29466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5213774658593403e-06,
+      "loss": 0.6891,
+      "step": 29467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.520811380497845e-06,
+      "loss": 0.9974,
+      "step": 29468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.520245349525844e-06,
+      "loss": 0.9923,
+      "step": 29469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.519679372947449e-06,
+      "loss": 0.6502,
+      "step": 29470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5191134507667758e-06,
+      "loss": 0.9786,
+      "step": 29471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5185475829879404e-06,
+      "loss": 0.8514,
+      "step": 29472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.517981769615058e-06,
+      "loss": 0.8222,
+      "step": 29473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5174160106522426e-06,
+      "loss": 0.9247,
+      "step": 29474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5168503061036086e-06,
+      "loss": 0.8757,
+      "step": 29475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5162846559732712e-06,
+      "loss": 0.7428,
+      "step": 29476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5157190602653424e-06,
+      "loss": 0.798,
+      "step": 29477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.515153518983936e-06,
+      "loss": 0.9379,
+      "step": 29478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.514588032133164e-06,
+      "loss": 0.8533,
+      "step": 29479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5140225997171353e-06,
+      "loss": 0.7592,
+      "step": 29480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.51345722173997e-06,
+      "loss": 1.1414,
+      "step": 29481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5128918982057737e-06,
+      "loss": 0.9107,
+      "step": 29482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.51232662911866e-06,
+      "loss": 0.9929,
+      "step": 29483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5117614144827383e-06,
+      "loss": 0.7238,
+      "step": 29484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.511196254302123e-06,
+      "loss": 0.9055,
+      "step": 29485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.510631148580913e-06,
+      "loss": 0.7691,
+      "step": 29486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5100660973232283e-06,
+      "loss": 0.8772,
+      "step": 29487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.509501100533174e-06,
+      "loss": 0.9263,
+      "step": 29488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.508936158214861e-06,
+      "loss": 0.7656,
+      "step": 29489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5083712703723952e-06,
+      "loss": 0.8745,
+      "step": 29490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.507806437009885e-06,
+      "loss": 0.554,
+      "step": 29491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5072416581314384e-06,
+      "loss": 0.6008,
+      "step": 29492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.50667693374116e-06,
+      "loss": 0.7512,
+      "step": 29493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.506112263843162e-06,
+      "loss": 0.8568,
+      "step": 29494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5055476484415507e-06,
+      "loss": 0.6983,
+      "step": 29495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.504983087540427e-06,
+      "loss": 0.9258,
+      "step": 29496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.504418581143897e-06,
+      "loss": 0.5362,
+      "step": 29497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.503854129256066e-06,
+      "loss": 0.719,
+      "step": 29498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.503289731881042e-06,
+      "loss": 0.8197,
+      "step": 29499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5027253890229286e-06,
+      "loss": 0.6878,
+      "step": 29500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.502161100685828e-06,
+      "loss": 0.648,
+      "step": 29501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5015968668738443e-06,
+      "loss": 0.9379,
+      "step": 29502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5010326875910816e-06,
+      "loss": 1.1285,
+      "step": 29503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5004685628416413e-06,
+      "loss": 0.7959,
+      "step": 29504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.499904492629627e-06,
+      "loss": 0.8889,
+      "step": 29505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.49934047695914e-06,
+      "loss": 0.8782,
+      "step": 29506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4987765158342814e-06,
+      "loss": 0.8784,
+      "step": 29507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.498212609259153e-06,
+      "loss": 0.9874,
+      "step": 29508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.497648757237856e-06,
+      "loss": 0.7582,
+      "step": 29509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4970849597744905e-06,
+      "loss": 1.1335,
+      "step": 29510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4965212168731534e-06,
+      "loss": 0.7581,
+      "step": 29511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.49595752853795e-06,
+      "loss": 0.8813,
+      "step": 29512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.495393894772976e-06,
+      "loss": 0.9428,
+      "step": 29513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4948303155823315e-06,
+      "loss": 0.6506,
+      "step": 29514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.494266790970118e-06,
+      "loss": 0.8669,
+      "step": 29515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.493703320940424e-06,
+      "loss": 0.8115,
+      "step": 29516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.493139905497355e-06,
+      "loss": 0.6422,
+      "step": 29517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.492576544645007e-06,
+      "loss": 0.6818,
+      "step": 29518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.492013238387475e-06,
+      "loss": 0.6128,
+      "step": 29519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4914499867288577e-06,
+      "loss": 0.7387,
+      "step": 29520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4908867896732493e-06,
+      "loss": 0.8116,
+      "step": 29521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4903236472247462e-06,
+      "loss": 1.0124,
+      "step": 29522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.489760559387441e-06,
+      "loss": 0.8107,
+      "step": 29523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4891975261654323e-06,
+      "loss": 1.2087,
+      "step": 29524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4886345475628172e-06,
+      "loss": 0.7736,
+      "step": 29525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4880716235836832e-06,
+      "loss": 1.109,
+      "step": 29526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4875087542321265e-06,
+      "loss": 0.7508,
+      "step": 29527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4869459395122395e-06,
+      "loss": 0.6038,
+      "step": 29528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4863831794281147e-06,
+      "loss": 0.714,
+      "step": 29529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4858204739838474e-06,
+      "loss": 0.6892,
+      "step": 29530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4852578231835276e-06,
+      "loss": 0.7612,
+      "step": 29531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4846952270312486e-06,
+      "loss": 0.6445,
+      "step": 29532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4841326855310998e-06,
+      "loss": 0.5986,
+      "step": 29533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4835701986871728e-06,
+      "loss": 0.894,
+      "step": 29534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.483007766503558e-06,
+      "loss": 0.6169,
+      "step": 29535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.482445388984346e-06,
+      "loss": 0.7626,
+      "step": 29536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.481883066133626e-06,
+      "loss": 0.6133,
+      "step": 29537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4813207979554856e-06,
+      "loss": 0.6361,
+      "step": 29538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.480758584454017e-06,
+      "loss": 0.7511,
+      "step": 29539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4801964256333054e-06,
+      "loss": 0.809,
+      "step": 29540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4796343214974394e-06,
+      "loss": 0.9026,
+      "step": 29541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.479072272050509e-06,
+      "loss": 0.9365,
+      "step": 29542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4785102772966017e-06,
+      "loss": 0.7998,
+      "step": 29543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4779483372398017e-06,
+      "loss": 0.6148,
+      "step": 29544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4773864518842006e-06,
+      "loss": 0.8095,
+      "step": 29545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4768246212338764e-06,
+      "loss": 0.7826,
+      "step": 29546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4762628452929172e-06,
+      "loss": 0.5788,
+      "step": 29547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4757011240654127e-06,
+      "loss": 0.8959,
+      "step": 29548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4751394575554445e-06,
+      "loss": 1.1304,
+      "step": 29549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.474577845767099e-06,
+      "loss": 0.8621,
+      "step": 29550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4740162887044584e-06,
+      "loss": 0.9143,
+      "step": 29551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4734547863716063e-06,
+      "loss": 0.6792,
+      "step": 29552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4728933387726286e-06,
+      "loss": 0.717,
+      "step": 29553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4723319459116025e-06,
+      "loss": 0.5391,
+      "step": 29554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4717706077926195e-06,
+      "loss": 1.0963,
+      "step": 29555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4712093244197545e-06,
+      "loss": 0.9461,
+      "step": 29556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4706480957970912e-06,
+      "loss": 1.1072,
+      "step": 29557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4700869219287117e-06,
+      "loss": 0.597,
+      "step": 29558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.469525802818693e-06,
+      "loss": 0.7393,
+      "step": 29559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.468964738471121e-06,
+      "loss": 0.8219,
+      "step": 29560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4684037288900753e-06,
+      "loss": 0.7615,
+      "step": 29561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.467842774079633e-06,
+      "loss": 0.99,
+      "step": 29562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4672818740438744e-06,
+      "loss": 0.984,
+      "step": 29563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.466721028786878e-06,
+      "loss": 0.9291,
+      "step": 29564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4661602383127235e-06,
+      "loss": 0.7754,
+      "step": 29565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4655995026254888e-06,
+      "loss": 0.6803,
+      "step": 29566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4650388217292507e-06,
+      "loss": 0.9371,
+      "step": 29567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4644781956280873e-06,
+      "loss": 0.9188,
+      "step": 29568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4639176243260753e-06,
+      "loss": 1.0055,
+      "step": 29569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.463357107827291e-06,
+      "loss": 0.5761,
+      "step": 29570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.462796646135811e-06,
+      "loss": 0.7632,
+      "step": 29571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4622362392557076e-06,
+      "loss": 1.0073,
+      "step": 29572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.461675887191062e-06,
+      "loss": 0.8141,
+      "step": 29573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4611155899459462e-06,
+      "loss": 0.9203,
+      "step": 29574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.460555347524438e-06,
+      "loss": 0.9339,
+      "step": 29575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4599951599306072e-06,
+      "loss": 0.7122,
+      "step": 29576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.459435027168524e-06,
+      "loss": 0.9347,
+      "step": 29577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4588749492422704e-06,
+      "loss": 0.8173,
+      "step": 29578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.458314926155916e-06,
+      "loss": 0.7169,
+      "step": 29579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4577549579135318e-06,
+      "loss": 0.8131,
+      "step": 29580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4571950445191916e-06,
+      "loss": 0.5524,
+      "step": 29581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4566351859769678e-06,
+      "loss": 0.8994,
+      "step": 29582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.456075382290929e-06,
+      "loss": 0.8118,
+      "step": 29583
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4555156334651455e-06,
+      "loss": 0.9591,
+      "step": 29584
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.454955939503696e-06,
+      "loss": 0.7014,
+      "step": 29585
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.454396300410642e-06,
+      "loss": 0.9636,
+      "step": 29586
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4538367161900555e-06,
+      "loss": 0.9026,
+      "step": 29587
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.453277186846007e-06,
+      "loss": 0.7128,
+      "step": 29588
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4527177123825643e-06,
+      "loss": 0.8714,
+      "step": 29589
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4521582928037935e-06,
+      "loss": 0.9603,
+      "step": 29590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.45159892811377e-06,
+      "loss": 0.9898,
+      "step": 29591
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.451039618316555e-06,
+      "loss": 0.9925,
+      "step": 29592
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.45048036341622e-06,
+      "loss": 0.6306,
+      "step": 29593
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4499211634168294e-06,
+      "loss": 1.0475,
+      "step": 29594
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.449362018322451e-06,
+      "loss": 1.05,
+      "step": 29595
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4488029281371505e-06,
+      "loss": 0.8451,
+      "step": 29596
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.448243892864993e-06,
+      "loss": 0.9874,
+      "step": 29597
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4476849125100445e-06,
+      "loss": 0.4297,
+      "step": 29598
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.447125987076371e-06,
+      "loss": 0.9268,
+      "step": 29599
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.446567116568035e-06,
+      "loss": 0.9428,
+      "step": 29600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4460083009891024e-06,
+      "loss": 0.7194,
+      "step": 29601
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.445449540343633e-06,
+      "loss": 0.806,
+      "step": 29602
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.444890834635697e-06,
+      "loss": 0.8604,
+      "step": 29603
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4443321838693523e-06,
+      "loss": 0.698,
+      "step": 29604
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.443773588048668e-06,
+      "loss": 0.7403,
+      "step": 29605
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.443215047177697e-06,
+      "loss": 0.8905,
+      "step": 29606
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4426565612605056e-06,
+      "loss": 0.8695,
+      "step": 29607
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4420981303011516e-06,
+      "loss": 0.8587,
+      "step": 29608
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4415397543037022e-06,
+      "loss": 0.7186,
+      "step": 29609
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.440981433272216e-06,
+      "loss": 0.8457,
+      "step": 29610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4404231672107514e-06,
+      "loss": 0.7552,
+      "step": 29611
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4398649561233687e-06,
+      "loss": 0.7483,
+      "step": 29612
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4393068000141283e-06,
+      "loss": 0.6057,
+      "step": 29613
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.438748698887089e-06,
+      "loss": 0.5513,
+      "step": 29614
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4381906527463082e-06,
+      "loss": 0.6314,
+      "step": 29615
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4376326615958444e-06,
+      "loss": 1.0389,
+      "step": 29616
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4370747254397554e-06,
+      "loss": 0.9663,
+      "step": 29617
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4365168442820997e-06,
+      "loss": 0.9537,
+      "step": 29618
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4359590181269332e-06,
+      "loss": 0.7812,
+      "step": 29619
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4354012469783096e-06,
+      "loss": 0.6577,
+      "step": 29620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4348435308402907e-06,
+      "loss": 0.8192,
+      "step": 29621
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4342858697169304e-06,
+      "loss": 0.6977,
+      "step": 29622
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4337282636122827e-06,
+      "loss": 0.9788,
+      "step": 29623
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.433170712530404e-06,
+      "loss": 1.1305,
+      "step": 29624
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.43261321647535e-06,
+      "loss": 0.7395,
+      "step": 29625
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4320557754511685e-06,
+      "loss": 0.9806,
+      "step": 29626
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4314983894619204e-06,
+      "loss": 0.5969,
+      "step": 29627
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4309410585116566e-06,
+      "loss": 0.681,
+      "step": 29628
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4303837826044295e-06,
+      "loss": 0.7395,
+      "step": 29629
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4298265617442918e-06,
+      "loss": 0.82,
+      "step": 29630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4292693959352966e-06,
+      "loss": 1.0157,
+      "step": 29631
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4287122851814947e-06,
+      "loss": 0.9683,
+      "step": 29632
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4281552294869347e-06,
+      "loss": 0.9245,
+      "step": 29633
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.427598228855674e-06,
+      "loss": 0.8597,
+      "step": 29634
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.427041283291763e-06,
+      "loss": 0.7893,
+      "step": 29635
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4264843927992456e-06,
+      "loss": 0.5682,
+      "step": 29636
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.425927557382175e-06,
+      "loss": 1.0292,
+      "step": 29637
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4253707770445966e-06,
+      "loss": 0.6878,
+      "step": 29638
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.424814051790567e-06,
+      "loss": 0.8203,
+      "step": 29639
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.42425738162413e-06,
+      "loss": 0.7043,
+      "step": 29640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.423700766549335e-06,
+      "loss": 0.8748,
+      "step": 29641
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.423144206570229e-06,
+      "loss": 1.0493,
+      "step": 29642
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4225877016908605e-06,
+      "loss": 0.805,
+      "step": 29643
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4220312519152757e-06,
+      "loss": 0.5443,
+      "step": 29644
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.42147485724752e-06,
+      "loss": 0.834,
+      "step": 29645
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4209185176916415e-06,
+      "loss": 0.934,
+      "step": 29646
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.420362233251685e-06,
+      "loss": 0.8705,
+      "step": 29647
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.419806003931696e-06,
+      "loss": 0.9816,
+      "step": 29648
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.419249829735719e-06,
+      "loss": 0.612,
+      "step": 29649
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4186937106678e-06,
+      "loss": 0.894,
+      "step": 29650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4181376467319785e-06,
+      "loss": 0.5393,
+      "step": 29651
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.417581637932306e-06,
+      "loss": 0.8996,
+      "step": 29652
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.417025684272821e-06,
+      "loss": 0.9023,
+      "step": 29653
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.416469785757567e-06,
+      "loss": 0.9897,
+      "step": 29654
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4159139423905898e-06,
+      "loss": 0.8862,
+      "step": 29655
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4153581541759243e-06,
+      "loss": 0.6454,
+      "step": 29656
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.414802421117618e-06,
+      "loss": 0.8038,
+      "step": 29657
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.41424674321971e-06,
+      "loss": 0.7033,
+      "step": 29658
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4136911204862446e-06,
+      "loss": 0.8083,
+      "step": 29659
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4131355529212576e-06,
+      "loss": 0.7082,
+      "step": 29660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4125800405287925e-06,
+      "loss": 1.0098,
+      "step": 29661
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4120245833128875e-06,
+      "loss": 0.4526,
+      "step": 29662
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.411469181277579e-06,
+      "loss": 0.8766,
+      "step": 29663
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4109138344269136e-06,
+      "loss": 0.8509,
+      "step": 29664
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4103585427649267e-06,
+      "loss": 0.8395,
+      "step": 29665
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.409803306295654e-06,
+      "loss": 0.9088,
+      "step": 29666
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4092481250231335e-06,
+      "loss": 0.77,
+      "step": 29667
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.408692998951404e-06,
+      "loss": 0.7447,
+      "step": 29668
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4081379280844985e-06,
+      "loss": 0.6945,
+      "step": 29669
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4075829124264606e-06,
+      "loss": 0.5507,
+      "step": 29670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4070279519813235e-06,
+      "loss": 0.9837,
+      "step": 29671
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4064730467531207e-06,
+      "loss": 0.7139,
+      "step": 29672
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.40591819674589e-06,
+      "loss": 1.1301,
+      "step": 29673
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.405363401963665e-06,
+      "loss": 0.6984,
+      "step": 29674
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4048086624104816e-06,
+      "loss": 1.0571,
+      "step": 29675
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.404253978090373e-06,
+      "loss": 0.7743,
+      "step": 29676
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.403699349007372e-06,
+      "loss": 0.7177,
+      "step": 29677
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4031447751655134e-06,
+      "loss": 0.7884,
+      "step": 29678
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4025902565688295e-06,
+      "loss": 0.5503,
+      "step": 29679
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.402035793221353e-06,
+      "loss": 0.9716,
+      "step": 29680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4014813851271133e-06,
+      "loss": 0.9046,
+      "step": 29681
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4009270322901477e-06,
+      "loss": 0.5421,
+      "step": 29682
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4003727347144845e-06,
+      "loss": 0.8726,
+      "step": 29683
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3998184924041546e-06,
+      "loss": 0.8271,
+      "step": 29684
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3992643053631904e-06,
+      "loss": 0.8846,
+      "step": 29685
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3987101735956187e-06,
+      "loss": 0.8782,
+      "step": 29686
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3981560971054675e-06,
+      "loss": 0.8487,
+      "step": 29687
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.397602075896772e-06,
+      "loss": 1.044,
+      "step": 29688
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.397048109973559e-06,
+      "loss": 0.7196,
+      "step": 29689
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.396494199339856e-06,
+      "loss": 0.7422,
+      "step": 29690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.395940343999691e-06,
+      "loss": 0.7363,
+      "step": 29691
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3953865439570934e-06,
+      "loss": 0.8905,
+      "step": 29692
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.394832799216088e-06,
+      "loss": 0.798,
+      "step": 29693
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.394279109780704e-06,
+      "loss": 0.9139,
+      "step": 29694
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3937254756549668e-06,
+      "loss": 0.8834,
+      "step": 29695
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.393171896842903e-06,
+      "loss": 1.0401,
+      "step": 29696
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3926183733485365e-06,
+      "loss": 1.0158,
+      "step": 29697
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.392064905175895e-06,
+      "loss": 0.968,
+      "step": 29698
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3915114923289995e-06,
+      "loss": 0.5832,
+      "step": 29699
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3909581348118803e-06,
+      "loss": 0.7338,
+      "step": 29700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.390404832628558e-06,
+      "loss": 0.9213,
+      "step": 29701
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.389851585783057e-06,
+      "loss": 0.9369,
+      "step": 29702
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3892983942794e-06,
+      "loss": 0.9605,
+      "step": 29703
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.388745258121613e-06,
+      "loss": 0.6651,
+      "step": 29704
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3881921773137107e-06,
+      "loss": 0.5722,
+      "step": 29705
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3876391518597232e-06,
+      "loss": 0.9337,
+      "step": 29706
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.387086181763668e-06,
+      "loss": 0.7398,
+      "step": 29707
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3865332670295682e-06,
+      "loss": 0.839,
+      "step": 29708
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3859804076614445e-06,
+      "loss": 1.0577,
+      "step": 29709
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.385427603663316e-06,
+      "loss": 0.6792,
+      "step": 29710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.384874855039203e-06,
+      "loss": 0.686,
+      "step": 29711
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3843221617931244e-06,
+      "loss": 0.6198,
+      "step": 29712
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3837695239291024e-06,
+      "loss": 0.6931,
+      "step": 29713
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3832169414511575e-06,
+      "loss": 0.6468,
+      "step": 29714
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3826644143633017e-06,
+      "loss": 0.784,
+      "step": 29715
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.382111942669556e-06,
+      "loss": 0.8025,
+      "step": 29716
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3815595263739355e-06,
+      "loss": 0.8893,
+      "step": 29717
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3810071654804636e-06,
+      "loss": 1.0648,
+      "step": 29718
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.380454859993153e-06,
+      "loss": 0.6561,
+      "step": 29719
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3799026099160207e-06,
+      "loss": 0.792,
+      "step": 29720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.379350415253083e-06,
+      "loss": 0.7671,
+      "step": 29721
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.378798276008356e-06,
+      "loss": 0.741,
+      "step": 29722
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3782461921858536e-06,
+      "loss": 0.8222,
+      "step": 29723
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3776941637895912e-06,
+      "loss": 1.0659,
+      "step": 29724
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3771421908235837e-06,
+      "loss": 0.8448,
+      "step": 29725
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3765902732918444e-06,
+      "loss": 0.8824,
+      "step": 29726
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.376038411198387e-06,
+      "loss": 1.0528,
+      "step": 29727
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.375486604547225e-06,
+      "loss": 0.9504,
+      "step": 29728
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.374934853342371e-06,
+      "loss": 0.7831,
+      "step": 29729
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3743831575878352e-06,
+      "loss": 0.545,
+      "step": 29730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3738315172876337e-06,
+      "loss": 0.8568,
+      "step": 29731
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3732799324457776e-06,
+      "loss": 0.9589,
+      "step": 29732
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3727284030662756e-06,
+      "loss": 0.9706,
+      "step": 29733
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3721769291531427e-06,
+      "loss": 0.9801,
+      "step": 29734
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3716255107103814e-06,
+      "loss": 1.0702,
+      "step": 29735
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3710741477420085e-06,
+      "loss": 0.7627,
+      "step": 29736
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.370522840252032e-06,
+      "loss": 0.7512,
+      "step": 29737
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3699715882444605e-06,
+      "loss": 0.6625,
+      "step": 29738
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3694203917233027e-06,
+      "loss": 1.0419,
+      "step": 29739
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.368869250692567e-06,
+      "loss": 1.0036,
+      "step": 29740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3683181651562627e-06,
+      "loss": 1.1179,
+      "step": 29741
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.367767135118392e-06,
+      "loss": 0.7114,
+      "step": 29742
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3672161605829693e-06,
+      "loss": 0.9727,
+      "step": 29743
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3666652415540016e-06,
+      "loss": 0.626,
+      "step": 29744
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.366114378035489e-06,
+      "loss": 0.7659,
+      "step": 29745
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3655635700314406e-06,
+      "loss": 1.3394,
+      "step": 29746
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3650128175458607e-06,
+      "loss": 0.6948,
+      "step": 29747
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3644621205827534e-06,
+      "loss": 0.6545,
+      "step": 29748
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.363911479146128e-06,
+      "loss": 0.7469,
+      "step": 29749
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3633608932399865e-06,
+      "loss": 0.8943,
+      "step": 29750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.362810362868333e-06,
+      "loss": 0.7826,
+      "step": 29751
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.36225988803517e-06,
+      "loss": 0.9772,
+      "step": 29752
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3617094687445007e-06,
+      "loss": 0.5312,
+      "step": 29753
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3611591050003292e-06,
+      "loss": 0.8315,
+      "step": 29754
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.360608796806657e-06,
+      "loss": 0.6356,
+      "step": 29755
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.360058544167486e-06,
+      "loss": 0.8055,
+      "step": 29756
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3595083470868174e-06,
+      "loss": 0.9208,
+      "step": 29757
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3589582055686523e-06,
+      "loss": 0.7372,
+      "step": 29758
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.358408119616993e-06,
+      "loss": 1.0428,
+      "step": 29759
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3578580892358337e-06,
+      "loss": 0.783,
+      "step": 29760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.357308114429183e-06,
+      "loss": 0.4912,
+      "step": 29761
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3567581952010353e-06,
+      "loss": 0.567,
+      "step": 29762
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3562083315553917e-06,
+      "loss": 0.8423,
+      "step": 29763
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3556585234962535e-06,
+      "loss": 0.9128,
+      "step": 29764
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.355108771027611e-06,
+      "loss": 0.6213,
+      "step": 29765
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3545590741534653e-06,
+      "loss": 0.8307,
+      "step": 29766
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3540094328778164e-06,
+      "loss": 0.8349,
+      "step": 29767
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.35345984720466e-06,
+      "loss": 0.9404,
+      "step": 29768
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.352910317137994e-06,
+      "loss": 1.1869,
+      "step": 29769
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.352360842681812e-06,
+      "loss": 0.7485,
+      "step": 29770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3518114238401124e-06,
+      "loss": 0.6744,
+      "step": 29771
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.351262060616888e-06,
+      "loss": 0.6589,
+      "step": 29772
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3507127530161332e-06,
+      "loss": 0.9807,
+      "step": 29773
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.350163501041851e-06,
+      "loss": 0.7772,
+      "step": 29774
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3496143046980256e-06,
+      "loss": 0.6314,
+      "step": 29775
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.349065163988655e-06,
+      "loss": 0.8758,
+      "step": 29776
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.348516078917732e-06,
+      "loss": 0.8492,
+      "step": 29777
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.347967049489247e-06,
+      "loss": 0.7244,
+      "step": 29778
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3474180757071973e-06,
+      "loss": 0.4939,
+      "step": 29779
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3468691575755743e-06,
+      "loss": 0.7818,
+      "step": 29780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3463202950983675e-06,
+      "loss": 0.8811,
+      "step": 29781
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3457714882795712e-06,
+      "loss": 1.2327,
+      "step": 29782
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3452227371231728e-06,
+      "loss": 0.9389,
+      "step": 29783
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.344674041633166e-06,
+      "loss": 0.8455,
+      "step": 29784
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3441254018135394e-06,
+      "loss": 0.6677,
+      "step": 29785
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3435768176682827e-06,
+      "loss": 0.9187,
+      "step": 29786
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3430282892013858e-06,
+      "loss": 0.6248,
+      "step": 29787
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3424798164168373e-06,
+      "loss": 0.5632,
+      "step": 29788
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3419313993186255e-06,
+      "loss": 0.7435,
+      "step": 29789
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3413830379107395e-06,
+      "loss": 0.9988,
+      "step": 29790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3408347321971637e-06,
+      "loss": 0.6309,
+      "step": 29791
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3402864821818916e-06,
+      "loss": 0.6071,
+      "step": 29792
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3397382878689058e-06,
+      "loss": 0.7606,
+      "step": 29793
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.339190149262197e-06,
+      "loss": 0.7172,
+      "step": 29794
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.338642066365746e-06,
+      "loss": 0.8872,
+      "step": 29795
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3380940391835395e-06,
+      "loss": 0.6381,
+      "step": 29796
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3375460677195617e-06,
+      "loss": 1.1622,
+      "step": 29797
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.336998151977803e-06,
+      "loss": 0.6616,
+      "step": 29798
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3364502919622444e-06,
+      "loss": 1.0356,
+      "step": 29799
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3359024876768697e-06,
+      "loss": 0.8564,
+      "step": 29800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3353547391256636e-06,
+      "loss": 0.946,
+      "step": 29801
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3348070463126094e-06,
+      "loss": 0.7991,
+      "step": 29802
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3342594092416893e-06,
+      "loss": 0.8282,
+      "step": 29803
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3337118279168857e-06,
+      "loss": 0.6055,
+      "step": 29804
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3331643023421813e-06,
+      "loss": 0.6139,
+      "step": 29805
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.332616832521558e-06,
+      "loss": 0.7303,
+      "step": 29806
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3320694184589965e-06,
+      "loss": 0.7276,
+      "step": 29807
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3315220601584765e-06,
+      "loss": 0.8594,
+      "step": 29808
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3309747576239783e-06,
+      "loss": 0.8568,
+      "step": 29809
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.330427510859484e-06,
+      "loss": 0.7403,
+      "step": 29810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3298803198689747e-06,
+      "loss": 0.7064,
+      "step": 29811
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.329333184656427e-06,
+      "loss": 0.9101,
+      "step": 29812
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3287861052258198e-06,
+      "loss": 0.7047,
+      "step": 29813
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.328239081581134e-06,
+      "loss": 0.9165,
+      "step": 29814
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3276921137263408e-06,
+      "loss": 1.0351,
+      "step": 29815
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.327145201665425e-06,
+      "loss": 0.8418,
+      "step": 29816
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3265983454023624e-06,
+      "loss": 0.7731,
+      "step": 29817
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.326051544941127e-06,
+      "loss": 0.8484,
+      "step": 29818
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3255048002856983e-06,
+      "loss": 0.8945,
+      "step": 29819
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.324958111440051e-06,
+      "loss": 0.7427,
+      "step": 29820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3244114784081594e-06,
+      "loss": 0.6525,
+      "step": 29821
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.323864901193997e-06,
+      "loss": 0.8471,
+      "step": 29822
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3233183798015457e-06,
+      "loss": 0.6541,
+      "step": 29823
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.322771914234778e-06,
+      "loss": 0.881,
+      "step": 29824
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3222255044976625e-06,
+      "loss": 0.7018,
+      "step": 29825
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.321679150594175e-06,
+      "loss": 0.7577,
+      "step": 29826
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.321132852528287e-06,
+      "loss": 0.7377,
+      "step": 29827
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3205866103039775e-06,
+      "loss": 0.9075,
+      "step": 29828
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.320040423925214e-06,
+      "loss": 0.9822,
+      "step": 29829
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3194942933959696e-06,
+      "loss": 0.9354,
+      "step": 29830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.318948218720215e-06,
+      "loss": 0.8391,
+      "step": 29831
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3184021999019225e-06,
+      "loss": 0.7185,
+      "step": 29832
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3178562369450617e-06,
+      "loss": 0.8537,
+      "step": 29833
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3173103298536037e-06,
+      "loss": 0.6705,
+      "step": 29834
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.316764478631518e-06,
+      "loss": 0.8488,
+      "step": 29835
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.316218683282775e-06,
+      "loss": 0.8792,
+      "step": 29836
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3156729438113423e-06,
+      "loss": 0.7782,
+      "step": 29837
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3151272602211894e-06,
+      "loss": 0.642,
+      "step": 29838
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.314581632516284e-06,
+      "loss": 0.721,
+      "step": 29839
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.314036060700593e-06,
+      "loss": 0.6002,
+      "step": 29840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3134905447780865e-06,
+      "loss": 0.9227,
+      "step": 29841
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3129450847527315e-06,
+      "loss": 0.7447,
+      "step": 29842
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.312399680628492e-06,
+      "loss": 0.7578,
+      "step": 29843
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3118543324093403e-06,
+      "loss": 0.8037,
+      "step": 29844
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.311309040099231e-06,
+      "loss": 0.784,
+      "step": 29845
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3107638037021395e-06,
+      "loss": 0.8425,
+      "step": 29846
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.310218623222027e-06,
+      "loss": 0.8392,
+      "step": 29847
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3096734986628586e-06,
+      "loss": 1.0901,
+      "step": 29848
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.309128430028599e-06,
+      "loss": 0.8435,
+      "step": 29849
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.30858341732321e-06,
+      "loss": 0.6109,
+      "step": 29850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.308038460550659e-06,
+      "loss": 0.8899,
+      "step": 29851
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3074935597149017e-06,
+      "loss": 0.9405,
+      "step": 29852
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3069487148199086e-06,
+      "loss": 0.7794,
+      "step": 29853
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3064039258696426e-06,
+      "loss": 0.7838,
+      "step": 29854
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.305859192868057e-06,
+      "loss": 1.0302,
+      "step": 29855
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3053145158191194e-06,
+      "loss": 0.493,
+      "step": 29856
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3047698947267884e-06,
+      "loss": 0.9251,
+      "step": 29857
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3042253295950235e-06,
+      "loss": 0.7982,
+      "step": 29858
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3036808204277884e-06,
+      "loss": 0.8499,
+      "step": 29859
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3031363672290406e-06,
+      "loss": 0.8948,
+      "step": 29860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.30259197000274e-06,
+      "loss": 0.8808,
+      "step": 29861
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.302047628752846e-06,
+      "loss": 0.9048,
+      "step": 29862
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.301503343483317e-06,
+      "loss": 0.6107,
+      "step": 29863
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.300959114198109e-06,
+      "loss": 0.7581,
+      "step": 29864
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.300414940901182e-06,
+      "loss": 1.1673,
+      "step": 29865
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.299870823596493e-06,
+      "loss": 0.8632,
+      "step": 29866
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2993267622879976e-06,
+      "loss": 0.8696,
+      "step": 29867
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2987827569796538e-06,
+      "loss": 0.8863,
+      "step": 29868
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.298238807675417e-06,
+      "loss": 0.8505,
+      "step": 29869
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2976949143792404e-06,
+      "loss": 0.6947,
+      "step": 29870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2971510770950834e-06,
+      "loss": 1.0362,
+      "step": 29871
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2966072958269003e-06,
+      "loss": 0.8271,
+      "step": 29872
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.296063570578644e-06,
+      "loss": 0.8873,
+      "step": 29873
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.295519901354272e-06,
+      "loss": 0.9808,
+      "step": 29874
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2949762881577307e-06,
+      "loss": 0.8995,
+      "step": 29875
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.294432730992976e-06,
+      "loss": 0.5727,
+      "step": 29876
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.293889229863964e-06,
+      "loss": 0.8556,
+      "step": 29877
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2933457847746453e-06,
+      "loss": 0.8489,
+      "step": 29878
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2928023957289724e-06,
+      "loss": 0.8771,
+      "step": 29879
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.292259062730897e-06,
+      "loss": 0.6863,
+      "step": 29880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2917157857843686e-06,
+      "loss": 0.6472,
+      "step": 29881
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.291172564893339e-06,
+      "loss": 0.9568,
+      "step": 29882
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.290629400061757e-06,
+      "loss": 0.8026,
+      "step": 29883
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2900862912935785e-06,
+      "loss": 0.5812,
+      "step": 29884
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2895432385927454e-06,
+      "loss": 0.6881,
+      "step": 29885
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2890002419632107e-06,
+      "loss": 0.5642,
+      "step": 29886
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2884573014089227e-06,
+      "loss": 0.6691,
+      "step": 29887
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2879144169338275e-06,
+      "loss": 1.1052,
+      "step": 29888
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.287371588541877e-06,
+      "loss": 0.8584,
+      "step": 29889
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2868288162370165e-06,
+      "loss": 1.0658,
+      "step": 29890
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.286286100023194e-06,
+      "loss": 1.0286,
+      "step": 29891
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2857434399043543e-06,
+      "loss": 0.8136,
+      "step": 29892
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2852008358844467e-06,
+      "loss": 0.8513,
+      "step": 29893
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.284658287967414e-06,
+      "loss": 0.7729,
+      "step": 29894
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2841157961572034e-06,
+      "loss": 1.0271,
+      "step": 29895
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2835733604577603e-06,
+      "loss": 0.712,
+      "step": 29896
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2830309808730277e-06,
+      "loss": 1.0138,
+      "step": 29897
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2824886574069515e-06,
+      "loss": 1.0129,
+      "step": 29898
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.281946390063474e-06,
+      "loss": 0.6999,
+      "step": 29899
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2814041788465404e-06,
+      "loss": 0.9612,
+      "step": 29900
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.280862023760089e-06,
+      "loss": 1.2179,
+      "step": 29901
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.280319924808069e-06,
+      "loss": 0.8028,
+      "step": 29902
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2797778819944195e-06,
+      "loss": 0.6249,
+      "step": 29903
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2792358953230863e-06,
+      "loss": 0.8034,
+      "step": 29904
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2786939647980033e-06,
+      "loss": 0.753,
+      "step": 29905
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2781520904231134e-06,
+      "loss": 0.584,
+      "step": 29906
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2776102722023617e-06,
+      "loss": 0.9779,
+      "step": 29907
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.277068510139685e-06,
+      "loss": 0.7896,
+      "step": 29908
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.276526804239024e-06,
+      "loss": 0.755,
+      "step": 29909
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2759851545043175e-06,
+      "loss": 0.9125,
+      "step": 29910
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2754435609395055e-06,
+      "loss": 0.7121,
+      "step": 29911
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2749020235485253e-06,
+      "loss": 0.8952,
+      "step": 29912
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.274360542335312e-06,
+      "loss": 0.9968,
+      "step": 29913
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.273819117303813e-06,
+      "loss": 0.9931,
+      "step": 29914
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.273277748457956e-06,
+      "loss": 0.9098,
+      "step": 29915
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.272736435801681e-06,
+      "loss": 0.9022,
+      "step": 29916
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2721951793389263e-06,
+      "loss": 0.7094,
+      "step": 29917
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.271653979073626e-06,
+      "loss": 1.0799,
+      "step": 29918
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2711128350097123e-06,
+      "loss": 0.6289,
+      "step": 29919
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.270571747151128e-06,
+      "loss": 0.7997,
+      "step": 29920
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.270030715501803e-06,
+      "loss": 0.5717,
+      "step": 29921
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2694897400656744e-06,
+      "loss": 0.7022,
+      "step": 29922
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2689488208466747e-06,
+      "loss": 0.9217,
+      "step": 29923
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2684079578487373e-06,
+      "loss": 0.6283,
+      "step": 29924
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2678671510757953e-06,
+      "loss": 0.649,
+      "step": 29925
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.267326400531783e-06,
+      "loss": 0.461,
+      "step": 29926
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2667857062206323e-06,
+      "loss": 1.0368,
+      "step": 29927
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2662450681462733e-06,
+      "loss": 0.868,
+      "step": 29928
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.26570448631264e-06,
+      "loss": 0.7899,
+      "step": 29929
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.265163960723662e-06,
+      "loss": 0.9857,
+      "step": 29930
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2646234913832677e-06,
+      "loss": 0.7679,
+      "step": 29931
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2640830782953924e-06,
+      "loss": 0.6395,
+      "step": 29932
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.263542721463966e-06,
+      "loss": 1.0431,
+      "step": 29933
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.263002420892918e-06,
+      "loss": 0.8596,
+      "step": 29934
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.262462176586172e-06,
+      "loss": 0.693,
+      "step": 29935
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2619219885476606e-06,
+      "loss": 0.7278,
+      "step": 29936
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2613818567813094e-06,
+      "loss": 0.7578,
+      "step": 29937
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2608417812910522e-06,
+      "loss": 0.7495,
+      "step": 29938
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.260301762080812e-06,
+      "loss": 0.8328,
+      "step": 29939
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.259761799154516e-06,
+      "loss": 0.6262,
+      "step": 29940
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2592218925160935e-06,
+      "loss": 0.553,
+      "step": 29941
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.258682042169468e-06,
+      "loss": 0.9304,
+      "step": 29942
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.258142248118567e-06,
+      "loss": 0.9178,
+      "step": 29943
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2576025103673148e-06,
+      "loss": 0.7089,
+      "step": 29944
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.257062828919636e-06,
+      "loss": 0.8307,
+      "step": 29945
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.256523203779457e-06,
+      "loss": 0.7252,
+      "step": 29946
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2559836349507015e-06,
+      "loss": 0.9196,
+      "step": 29947
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2554441224372937e-06,
+      "loss": 0.9003,
+      "step": 29948
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.254904666243152e-06,
+      "loss": 0.7814,
+      "step": 29949
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.254365266372207e-06,
+      "loss": 0.7377,
+      "step": 29950
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2538259228283776e-06,
+      "loss": 0.6535,
+      "step": 29951
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.253286635615587e-06,
+      "loss": 0.9086,
+      "step": 29952
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.252747404737755e-06,
+      "loss": 0.8826,
+      "step": 29953
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.252208230198808e-06,
+      "loss": 0.6553,
+      "step": 29954
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.251669112002657e-06,
+      "loss": 0.9243,
+      "step": 29955
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.251130050153232e-06,
+      "loss": 0.9188,
+      "step": 29956
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.25059104465445e-06,
+      "loss": 0.9913,
+      "step": 29957
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2500520955102288e-06,
+      "loss": 1.035,
+      "step": 29958
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2495132027244915e-06,
+      "loss": 1.0642,
+      "step": 29959
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.248974366301153e-06,
+      "loss": 0.5195,
+      "step": 29960
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2484355862441353e-06,
+      "loss": 1.0044,
+      "step": 29961
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2478968625573506e-06,
+      "loss": 0.8536,
+      "step": 29962
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2473581952447243e-06,
+      "loss": 0.8293,
+      "step": 29963
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2468195843101726e-06,
+      "loss": 0.7401,
+      "step": 29964
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2462810297576075e-06,
+      "loss": 0.7645,
+      "step": 29965
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2457425315909475e-06,
+      "loss": 0.6908,
+      "step": 29966
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.245204089814106e-06,
+      "loss": 0.7656,
+      "step": 29967
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.244665704431005e-06,
+      "loss": 0.7858,
+      "step": 29968
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.244127375445555e-06,
+      "loss": 1.1406,
+      "step": 29969
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.243589102861673e-06,
+      "loss": 0.9012,
+      "step": 29970
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2430508866832724e-06,
+      "loss": 0.8611,
+      "step": 29971
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2425127269142665e-06,
+      "loss": 1.0655,
+      "step": 29972
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2419746235585694e-06,
+      "loss": 1.0408,
+      "step": 29973
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.241436576620095e-06,
+      "loss": 0.8787,
+      "step": 29974
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2408985861027554e-06,
+      "loss": 1.113,
+      "step": 29975
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2403606520104627e-06,
+      "loss": 1.0541,
+      "step": 29976
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.239822774347129e-06,
+      "loss": 0.5378,
+      "step": 29977
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.239284953116665e-06,
+      "loss": 0.84,
+      "step": 29978
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2387471883229837e-06,
+      "loss": 0.8517,
+      "step": 29979
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2382094799699914e-06,
+      "loss": 0.742,
+      "step": 29980
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.237671828061605e-06,
+      "loss": 0.9615,
+      "step": 29981
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2371342326017307e-06,
+      "loss": 0.7803,
+      "step": 29982
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2365966935942784e-06,
+      "loss": 0.9711,
+      "step": 29983
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.23605921104316e-06,
+      "loss": 0.7525,
+      "step": 29984
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.235521784952275e-06,
+      "loss": 0.7914,
+      "step": 29985
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2349844153255407e-06,
+      "loss": 0.9519,
+      "step": 29986
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2344471021668622e-06,
+      "loss": 0.7561,
+      "step": 29987
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2339098454801465e-06,
+      "loss": 0.7553,
+      "step": 29988
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2333726452693017e-06,
+      "loss": 0.5209,
+      "step": 29989
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.232835501538233e-06,
+      "loss": 0.4476,
+      "step": 29990
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2322984142908467e-06,
+      "loss": 1.0163,
+      "step": 29991
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.231761383531047e-06,
+      "loss": 0.9142,
+      "step": 29992
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.231224409262743e-06,
+      "loss": 0.8781,
+      "step": 29993
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2306874914898414e-06,
+      "loss": 0.9702,
+      "step": 29994
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.230150630216239e-06,
+      "loss": 1.0359,
+      "step": 29995
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2296138254458454e-06,
+      "loss": 0.5944,
+      "step": 29996
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.229077077182562e-06,
+      "loss": 0.8873,
+      "step": 29997
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2285403854302912e-06,
+      "loss": 0.734,
+      "step": 29998
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2280037501929396e-06,
+      "loss": 0.6474,
+      "step": 29999
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.227467171474409e-06,
+      "loss": 0.6786,
+      "step": 30000
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2269306492785992e-06,
+      "loss": 1.0083,
+      "step": 30001
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.226394183609413e-06,
+      "loss": 0.8167,
+      "step": 30002
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.225857774470751e-06,
+      "loss": 0.815,
+      "step": 30003
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2253214218665153e-06,
+      "loss": 0.8741,
+      "step": 30004
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.224785125800606e-06,
+      "loss": 0.8365,
+      "step": 30005
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2242488862769217e-06,
+      "loss": 0.7539,
+      "step": 30006
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2237127032993643e-06,
+      "loss": 0.6582,
+      "step": 30007
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2231765768718306e-06,
+      "loss": 0.9963,
+      "step": 30008
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2226405069982206e-06,
+      "loss": 0.7436,
+      "step": 30009
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.22210449368243e-06,
+      "loss": 1.1021,
+      "step": 30010
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.221568536928361e-06,
+      "loss": 0.7241,
+      "step": 30011
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2210326367399106e-06,
+      "loss": 0.7715,
+      "step": 30012
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2204967931209744e-06,
+      "loss": 0.7116,
+      "step": 30013
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2199610060754517e-06,
+      "loss": 0.8678,
+      "step": 30014
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2194252756072343e-06,
+      "loss": 0.8552,
+      "step": 30015
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2188896017202176e-06,
+      "loss": 0.7374,
+      "step": 30016
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2183539844183023e-06,
+      "loss": 0.6845,
+      "step": 30017
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.21781842370538e-06,
+      "loss": 1.0046,
+      "step": 30018
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2172829195853475e-06,
+      "loss": 0.8057,
+      "step": 30019
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2167474720620974e-06,
+      "loss": 0.7499,
+      "step": 30020
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.216212081139525e-06,
+      "loss": 0.6459,
+      "step": 30021
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.215676746821521e-06,
+      "loss": 0.675,
+      "step": 30022
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2151414691119788e-06,
+      "loss": 0.9404,
+      "step": 30023
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.214606248014798e-06,
+      "loss": 0.685,
+      "step": 30024
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2140710835338607e-06,
+      "loss": 0.9857,
+      "step": 30025
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2135359756730645e-06,
+      "loss": 1.1577,
+      "step": 30026
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2130009244362993e-06,
+      "loss": 0.7972,
+      "step": 30027
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.212465929827452e-06,
+      "loss": 0.5831,
+      "step": 30028
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2119309918504217e-06,
+      "loss": 0.7789,
+      "step": 30029
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2113961105090933e-06,
+      "loss": 1.004,
+      "step": 30030
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2108612858073574e-06,
+      "loss": 0.8125,
+      "step": 30031
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.210326517749103e-06,
+      "loss": 0.8523,
+      "step": 30032
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2097918063382195e-06,
+      "loss": 0.7916,
+      "step": 30033
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2092571515785945e-06,
+      "loss": 0.9552,
+      "step": 30034
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2087225534741175e-06,
+      "loss": 0.6414,
+      "step": 30035
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2081880120286746e-06,
+      "loss": 0.7796,
+      "step": 30036
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2076535272461543e-06,
+      "loss": 0.6469,
+      "step": 30037
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2071190991304413e-06,
+      "loss": 0.8081,
+      "step": 30038
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2065847276854258e-06,
+      "loss": 0.6869,
+      "step": 30039
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.206050412914991e-06,
+      "loss": 0.6961,
+      "step": 30040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2055161548230185e-06,
+      "loss": 0.7011,
+      "step": 30041
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.204981953413403e-06,
+      "loss": 0.7624,
+      "step": 30042
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2044478086900233e-06,
+      "loss": 0.7205,
+      "step": 30043
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2039137206567686e-06,
+      "loss": 0.7909,
+      "step": 30044
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2033796893175152e-06,
+      "loss": 0.5502,
+      "step": 30045
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.202845714676148e-06,
+      "loss": 0.8002,
+      "step": 30046
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.202311796736556e-06,
+      "loss": 0.7182,
+      "step": 30047
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.201777935502617e-06,
+      "loss": 0.4412,
+      "step": 30048
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2012441309782163e-06,
+      "loss": 0.8769,
+      "step": 30049
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200710383167234e-06,
+      "loss": 0.6112,
+      "step": 30050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.200176692073551e-06,
+      "loss": 0.5358,
+      "step": 30051
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.199643057701051e-06,
+      "loss": 0.8748,
+      "step": 30052
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.199109480053608e-06,
+      "loss": 0.7373,
+      "step": 30053
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.198575959135114e-06,
+      "loss": 0.8392,
+      "step": 30054
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1980424949494394e-06,
+      "loss": 0.885,
+      "step": 30055
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1975090875004658e-06,
+      "loss": 0.8038,
+      "step": 30056
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.196975736792073e-06,
+      "loss": 0.7939,
+      "step": 30057
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1964424428281385e-06,
+      "loss": 0.6695,
+      "step": 30058
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1959092056125385e-06,
+      "loss": 0.8081,
+      "step": 30059
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.195376025149156e-06,
+      "loss": 0.7469,
+      "step": 30060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.194842901441866e-06,
+      "loss": 0.8414,
+      "step": 30061
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1943098344945446e-06,
+      "loss": 0.6638,
+      "step": 30062
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.193776824311069e-06,
+      "loss": 0.7353,
+      "step": 30063
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1932438708953165e-06,
+      "loss": 0.5273,
+      "step": 30064
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1927109742511597e-06,
+      "loss": 0.8332,
+      "step": 30065
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1921781343824765e-06,
+      "loss": 0.9594,
+      "step": 30066
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1916453512931414e-06,
+      "loss": 0.6809,
+      "step": 30067
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1911126249870273e-06,
+      "loss": 0.8136,
+      "step": 30068
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1905799554680108e-06,
+      "loss": 0.4054,
+      "step": 30069
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1900473427399627e-06,
+      "loss": 0.6637,
+      "step": 30070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1895147868067557e-06,
+      "loss": 0.9544,
+      "step": 30071
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1889822876722667e-06,
+      "loss": 0.7628,
+      "step": 30072
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1884498453403657e-06,
+      "loss": 0.8077,
+      "step": 30073
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.187917459814929e-06,
+      "loss": 0.7316,
+      "step": 30074
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1873851310998194e-06,
+      "loss": 0.7107,
+      "step": 30075
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1868528591989146e-06,
+      "loss": 0.8299,
+      "step": 30076
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1863206441160787e-06,
+      "loss": 0.8449,
+      "step": 30077
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1857884858551915e-06,
+      "loss": 0.8476,
+      "step": 30078
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1852563844201167e-06,
+      "loss": 0.87,
+      "step": 30079
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1847243398147266e-06,
+      "loss": 0.964,
+      "step": 30080
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.184192352042888e-06,
+      "loss": 0.5946,
+      "step": 30081
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.18366042110847e-06,
+      "loss": 0.8744,
+      "step": 30082
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.183128547015343e-06,
+      "loss": 0.7581,
+      "step": 30083
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1825967297673724e-06,
+      "loss": 1.1209,
+      "step": 30084
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.182064969368426e-06,
+      "loss": 0.7336,
+      "step": 30085
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.181533265822372e-06,
+      "loss": 0.7419,
+      "step": 30086
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.181001619133076e-06,
+      "loss": 0.6243,
+      "step": 30087
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1804700293044047e-06,
+      "loss": 0.8252,
+      "step": 30088
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1799384963402205e-06,
+      "loss": 0.8756,
+      "step": 30089
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.179407020244395e-06,
+      "loss": 0.754,
+      "step": 30090
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.17887560102079e-06,
+      "loss": 0.6979,
+      "step": 30091
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.17834423867327e-06,
+      "loss": 0.9611,
+      "step": 30092
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1778129332056998e-06,
+      "loss": 0.8269,
+      "step": 30093
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1772816846219457e-06,
+      "loss": 0.7002,
+      "step": 30094
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.176750492925861e-06,
+      "loss": 1.0121,
+      "step": 30095
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.17621935812132e-06,
+      "loss": 0.7692,
+      "step": 30096
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1756882802121792e-06,
+      "loss": 0.8489,
+      "step": 30097
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.175157259202302e-06,
+      "loss": 0.8402,
+      "step": 30098
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174626295095551e-06,
+      "loss": 0.6774,
+      "step": 30099
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174095387895786e-06,
+      "loss": 0.5838,
+      "step": 30100
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1735645376068693e-06,
+      "loss": 0.7268,
+      "step": 30101
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1730337442326564e-06,
+      "loss": 0.54,
+      "step": 30102
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.172503007777015e-06,
+      "loss": 0.8827,
+      "step": 30103
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1719723282438033e-06,
+      "loss": 0.7253,
+      "step": 30104
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1714417056368752e-06,
+      "loss": 0.7349,
+      "step": 30105
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1709111399600925e-06,
+      "loss": 0.4983,
+      "step": 30106
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1703806312173103e-06,
+      "loss": 0.7333,
+      "step": 30107
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.169850179412393e-06,
+      "loss": 0.6846,
+      "step": 30108
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.169319784549194e-06,
+      "loss": 0.9673,
+      "step": 30109
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.168789446631572e-06,
+      "loss": 0.9041,
+      "step": 30110
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1682591656633823e-06,
+      "loss": 0.6672,
+      "step": 30111
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1677289416484815e-06,
+      "loss": 0.6283,
+      "step": 30112
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.167198774590725e-06,
+      "loss": 0.7514,
+      "step": 30113
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1666686644939703e-06,
+      "loss": 0.9415,
+      "step": 30114
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1661386113620696e-06,
+      "loss": 0.8128,
+      "step": 30115
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1656086151988796e-06,
+      "loss": 0.8214,
+      "step": 30116
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.165078676008253e-06,
+      "loss": 0.9302,
+      "step": 30117
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1645487937940456e-06,
+      "loss": 0.9042,
+      "step": 30118
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1640189685601087e-06,
+      "loss": 0.7778,
+      "step": 30119
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1634892003102935e-06,
+      "loss": 0.6895,
+      "step": 30120
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1629594890484573e-06,
+      "loss": 0.8702,
+      "step": 30121
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1624298347784502e-06,
+      "loss": 0.8128,
+      "step": 30122
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1619002375041243e-06,
+      "loss": 0.8729,
+      "step": 30123
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1613706972293326e-06,
+      "loss": 0.6913,
+      "step": 30124
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.160841213957918e-06,
+      "loss": 0.9223,
+      "step": 30125
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1603117876937394e-06,
+      "loss": 0.8804,
+      "step": 30126
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1597824184406435e-06,
+      "loss": 0.8574,
+      "step": 30127
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1592531062024813e-06,
+      "loss": 0.8704,
+      "step": 30128
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.158723850983101e-06,
+      "loss": 0.8918,
+      "step": 30129
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.158194652786352e-06,
+      "loss": 0.7586,
+      "step": 30130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1576655116160814e-06,
+      "loss": 0.7987,
+      "step": 30131
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1571364274761365e-06,
+      "loss": 0.7211,
+      "step": 30132
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1566074003703685e-06,
+      "loss": 0.4912,
+      "step": 30133
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.156078430302625e-06,
+      "loss": 1.0706,
+      "step": 30134
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.155549517276747e-06,
+      "loss": 0.8672,
+      "step": 30135
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.155020661296585e-06,
+      "loss": 1.1075,
+      "step": 30136
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1544918623659837e-06,
+      "loss": 0.9324,
+      "step": 30137
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1539631204887855e-06,
+      "loss": 0.8285,
+      "step": 30138
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.153434435668843e-06,
+      "loss": 0.9538,
+      "step": 30139
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.152905807909995e-06,
+      "loss": 1.1419,
+      "step": 30140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1523772372160877e-06,
+      "loss": 0.8464,
+      "step": 30141
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1518487235909657e-06,
+      "loss": 1.0205,
+      "step": 30142
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.15132026703847e-06,
+      "loss": 1.1624,
+      "step": 30143
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.150791867562445e-06,
+      "loss": 0.9648,
+      "step": 30144
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1502635251667335e-06,
+      "loss": 0.8359,
+      "step": 30145
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1497352398551776e-06,
+      "loss": 0.7928,
+      "step": 30146
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1492070116316187e-06,
+      "loss": 1.0343,
+      "step": 30147
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1486788404998983e-06,
+      "loss": 0.6983,
+      "step": 30148
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1481507264638566e-06,
+      "loss": 0.8479,
+      "step": 30149
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1476226695273326e-06,
+      "loss": 0.8276,
+      "step": 30150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.147094669694171e-06,
+      "loss": 1.0261,
+      "step": 30151
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1465667269682088e-06,
+      "loss": 1.0305,
+      "step": 30152
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.146038841353286e-06,
+      "loss": 0.7592,
+      "step": 30153
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.145511012853243e-06,
+      "loss": 0.6799,
+      "step": 30154
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.144983241471914e-06,
+      "loss": 0.7545,
+      "step": 30155
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.144455527213136e-06,
+      "loss": 0.7145,
+      "step": 30156
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.143927870080753e-06,
+      "loss": 0.7779,
+      "step": 30157
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1434002700785983e-06,
+      "loss": 0.6741,
+      "step": 30158
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.14287272721051e-06,
+      "loss": 1.087,
+      "step": 30159
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1423452414803235e-06,
+      "loss": 0.7392,
+      "step": 30160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1418178128918742e-06,
+      "loss": 0.8403,
+      "step": 30161
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.141290441449e-06,
+      "loss": 0.6447,
+      "step": 30162
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1407631271555306e-06,
+      "loss": 0.4491,
+      "step": 30163
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.140235870015309e-06,
+      "loss": 0.8152,
+      "step": 30164
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1397086700321635e-06,
+      "loss": 0.9191,
+      "step": 30165
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1391815272099293e-06,
+      "loss": 0.7705,
+      "step": 30166
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.13865444155244e-06,
+      "loss": 0.6655,
+      "step": 30167
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1381274130635255e-06,
+      "loss": 0.8695,
+      "step": 30168
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.137600441747024e-06,
+      "loss": 0.8455,
+      "step": 30169
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1370735276067666e-06,
+      "loss": 0.9014,
+      "step": 30170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.136546670646582e-06,
+      "loss": 0.7439,
+      "step": 30171
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1360198708703037e-06,
+      "loss": 0.8586,
+      "step": 30172
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1354931282817627e-06,
+      "loss": 0.5942,
+      "step": 30173
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.134966442884788e-06,
+      "loss": 0.8127,
+      "step": 30174
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.134439814683212e-06,
+      "loss": 0.9315,
+      "step": 30175
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1339132436808628e-06,
+      "loss": 0.9402,
+      "step": 30176
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1333867298815692e-06,
+      "loss": 0.8725,
+      "step": 30177
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.132860273289161e-06,
+      "loss": 0.7836,
+      "step": 30178
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1323338739074673e-06,
+      "loss": 0.6692,
+      "step": 30179
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1318075317403152e-06,
+      "loss": 1.0477,
+      "step": 30180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.13128124679153e-06,
+      "loss": 0.6154,
+      "step": 30181
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.130755019064943e-06,
+      "loss": 0.9397,
+      "step": 30182
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1302288485643797e-06,
+      "loss": 0.864,
+      "step": 30183
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1297027352936693e-06,
+      "loss": 0.8652,
+      "step": 30184
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1291766792566316e-06,
+      "loss": 0.7463,
+      "step": 30185
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1286506804570927e-06,
+      "loss": 0.6193,
+      "step": 30186
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1281247388988825e-06,
+      "loss": 0.6935,
+      "step": 30187
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.127598854585824e-06,
+      "loss": 0.8964,
+      "step": 30188
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.12707302752174e-06,
+      "loss": 0.8423,
+      "step": 30189
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.126547257710455e-06,
+      "loss": 0.6584,
+      "step": 30190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1260215451557933e-06,
+      "loss": 1.0293,
+      "step": 30191
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1254958898615774e-06,
+      "loss": 0.827,
+      "step": 30192
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1249702918316274e-06,
+      "loss": 1.083,
+      "step": 30193
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.124444751069773e-06,
+      "loss": 0.9526,
+      "step": 30194
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.123919267579828e-06,
+      "loss": 0.8782,
+      "step": 30195
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.123393841365616e-06,
+      "loss": 1.0323,
+      "step": 30196
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1228684724309602e-06,
+      "loss": 0.7382,
+      "step": 30197
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1223431607796784e-06,
+      "loss": 0.6954,
+      "step": 30198
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1218179064155896e-06,
+      "loss": 0.5651,
+      "step": 30199
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1212927093425186e-06,
+      "loss": 0.6322,
+      "step": 30200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.120767569564283e-06,
+      "loss": 0.3001,
+      "step": 30201
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1202424870846983e-06,
+      "loss": 0.5992,
+      "step": 30202
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.119717461907587e-06,
+      "loss": 0.9483,
+      "step": 30203
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1191924940367647e-06,
+      "loss": 0.6838,
+      "step": 30204
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1186675834760507e-06,
+      "loss": 1.0865,
+      "step": 30205
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.11814273022926e-06,
+      "loss": 0.6799,
+      "step": 30206
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1176179343002112e-06,
+      "loss": 0.9101,
+      "step": 30207
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.11709319569272e-06,
+      "loss": 1.0658,
+      "step": 30208
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.116568514410602e-06,
+      "loss": 0.4434,
+      "step": 30209
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1160438904576743e-06,
+      "loss": 0.5439,
+      "step": 30210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.115519323837747e-06,
+      "loss": 0.7783,
+      "step": 30211
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.114994814554643e-06,
+      "loss": 0.6606,
+      "step": 30212
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.114470362612171e-06,
+      "loss": 0.7162,
+      "step": 30213
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1139459680141494e-06,
+      "loss": 0.6455,
+      "step": 30214
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1134216307643853e-06,
+      "loss": 0.8009,
+      "step": 30215
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.112897350866695e-06,
+      "loss": 0.875,
+      "step": 30216
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.112373128324888e-06,
+      "loss": 0.6078,
+      "step": 30217
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.111848963142783e-06,
+      "loss": 0.9063,
+      "step": 30218
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1113248553241884e-06,
+      "loss": 0.9292,
+      "step": 30219
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1108008048729146e-06,
+      "loss": 0.7764,
+      "step": 30220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.110276811792774e-06,
+      "loss": 0.9156,
+      "step": 30221
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1097528760875763e-06,
+      "loss": 0.7245,
+      "step": 30222
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1092289977611325e-06,
+      "loss": 0.9159,
+      "step": 30223
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.108705176817251e-06,
+      "loss": 0.9453,
+      "step": 30224
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.108181413259741e-06,
+      "loss": 0.47,
+      "step": 30225
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1076577070924144e-06,
+      "loss": 1.0161,
+      "step": 30226
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.107134058319076e-06,
+      "loss": 0.6517,
+      "step": 30227
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.106610466943535e-06,
+      "loss": 0.909,
+      "step": 30228
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1060869329695966e-06,
+      "loss": 0.9671,
+      "step": 30229
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.105563456401073e-06,
+      "loss": 0.9199,
+      "step": 30230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.105040037241769e-06,
+      "loss": 0.8555,
+      "step": 30231
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1045166754954914e-06,
+      "loss": 0.7872,
+      "step": 30232
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.103993371166044e-06,
+      "loss": 0.8812,
+      "step": 30233
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1034701242572374e-06,
+      "loss": 0.6639,
+      "step": 30234
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.102946934772867e-06,
+      "loss": 0.8386,
+      "step": 30235
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1024238027167467e-06,
+      "loss": 0.9569,
+      "step": 30236
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1019007280926762e-06,
+      "loss": 1.0902,
+      "step": 30237
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.101377710904462e-06,
+      "loss": 0.7698,
+      "step": 30238
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1008547511559053e-06,
+      "loss": 0.6312,
+      "step": 30239
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1003318488508107e-06,
+      "loss": 0.9282,
+      "step": 30240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0998090039929796e-06,
+      "loss": 0.6509,
+      "step": 30241
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0992862165862115e-06,
+      "loss": 0.8456,
+      "step": 30242
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0987634866343155e-06,
+      "loss": 0.7169,
+      "step": 30243
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0982408141410904e-06,
+      "loss": 0.7009,
+      "step": 30244
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0977181991103336e-06,
+      "loss": 0.6381,
+      "step": 30245
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0971956415458473e-06,
+      "loss": 0.911,
+      "step": 30246
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0966731414514297e-06,
+      "loss": 0.8498,
+      "step": 30247
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0961506988308843e-06,
+      "loss": 0.6386,
+      "step": 30248
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0956283136880097e-06,
+      "loss": 0.8825,
+      "step": 30249
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.095105986026603e-06,
+      "loss": 0.8365,
+      "step": 30250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0945837158504644e-06,
+      "loss": 0.9048,
+      "step": 30251
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0940615031633903e-06,
+      "loss": 0.8845,
+      "step": 30252
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0935393479691777e-06,
+      "loss": 0.6329,
+      "step": 30253
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.093017250271627e-06,
+      "loss": 0.9199,
+      "step": 30254
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.092495210074532e-06,
+      "loss": 0.676,
+      "step": 30255
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0919732273816896e-06,
+      "loss": 0.6899,
+      "step": 30256
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.091451302196896e-06,
+      "loss": 0.6914,
+      "step": 30257
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.090929434523947e-06,
+      "loss": 0.586,
+      "step": 30258
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0904076243666383e-06,
+      "loss": 0.7996,
+      "step": 30259
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0898858717287594e-06,
+      "loss": 0.9134,
+      "step": 30260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0893641766141128e-06,
+      "loss": 0.8909,
+      "step": 30261
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.088842539026489e-06,
+      "loss": 0.9487,
+      "step": 30262
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0883209589696786e-06,
+      "loss": 0.9472,
+      "step": 30263
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0877994364474805e-06,
+      "loss": 1.0449,
+      "step": 30264
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0872779714636783e-06,
+      "loss": 0.9942,
+      "step": 30265
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.086756564022072e-06,
+      "loss": 0.7643,
+      "step": 30266
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0862352141264507e-06,
+      "loss": 0.8113,
+      "step": 30267
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0857139217806067e-06,
+      "loss": 1.1528,
+      "step": 30268
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.085192686988329e-06,
+      "loss": 0.9103,
+      "step": 30269
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0846715097534087e-06,
+      "loss": 0.8703,
+      "step": 30270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0841503900796356e-06,
+      "loss": 0.7962,
+      "step": 30271
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0836293279707985e-06,
+      "loss": 0.7591,
+      "step": 30272
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0831083234306902e-06,
+      "loss": 0.6421,
+      "step": 30273
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0825873764631e-06,
+      "loss": 0.9613,
+      "step": 30274
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.082066487071811e-06,
+      "loss": 0.8175,
+      "step": 30275
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.081545655260614e-06,
+      "loss": 0.8905,
+      "step": 30276
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0810248810332956e-06,
+      "loss": 1.0051,
+      "step": 30277
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0805041643936417e-06,
+      "loss": 0.8904,
+      "step": 30278
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.079983505345443e-06,
+      "loss": 0.8283,
+      "step": 30279
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0794629038924843e-06,
+      "loss": 0.5314,
+      "step": 30280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0789423600385506e-06,
+      "loss": 0.9738,
+      "step": 30281
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0784218737874284e-06,
+      "loss": 0.9136,
+      "step": 30282
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0779014451429013e-06,
+      "loss": 0.8634,
+      "step": 30283
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.077381074108754e-06,
+      "loss": 0.8795,
+      "step": 30284
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0768607606887724e-06,
+      "loss": 0.8252,
+      "step": 30285
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.076340504886739e-06,
+      "loss": 0.8306,
+      "step": 30286
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0758203067064363e-06,
+      "loss": 0.7737,
+      "step": 30287
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.075300166151649e-06,
+      "loss": 0.9281,
+      "step": 30288
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.07478008322616e-06,
+      "loss": 0.7249,
+      "step": 30289
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0742600579337467e-06,
+      "loss": 0.9264,
+      "step": 30290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0737400902781966e-06,
+      "loss": 0.7284,
+      "step": 30291
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.073220180263289e-06,
+      "loss": 0.8276,
+      "step": 30292
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072700327892805e-06,
+      "loss": 0.741,
+      "step": 30293
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072180533170527e-06,
+      "loss": 0.854,
+      "step": 30294
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.07166079610023e-06,
+      "loss": 0.5743,
+      "step": 30295
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.071141116685693e-06,
+      "loss": 0.8677,
+      "step": 30296
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.070621494930701e-06,
+      "loss": 0.8617,
+      "step": 30297
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.07010193083903e-06,
+      "loss": 0.9851,
+      "step": 30298
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.069582424414459e-06,
+      "loss": 0.9783,
+      "step": 30299
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.069062975660765e-06,
+      "loss": 0.9274,
+      "step": 30300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0685435845817248e-06,
+      "loss": 0.9484,
+      "step": 30301
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0680242511811176e-06,
+      "loss": 0.7602,
+      "step": 30302
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.067504975462715e-06,
+      "loss": 0.8324,
+      "step": 30303
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0669857574303044e-06,
+      "loss": 0.8954,
+      "step": 30304
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0664665970876496e-06,
+      "loss": 0.6069,
+      "step": 30305
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0659474944385315e-06,
+      "loss": 0.8382,
+      "step": 30306
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.065428449486725e-06,
+      "loss": 0.8771,
+      "step": 30307
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0649094622360034e-06,
+      "loss": 0.4278,
+      "step": 30308
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.064390532690138e-06,
+      "loss": 1.1406,
+      "step": 30309
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0638716608529096e-06,
+      "loss": 0.7912,
+      "step": 30310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0633528467280874e-06,
+      "loss": 0.8107,
+      "step": 30311
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0628340903194443e-06,
+      "loss": 0.8433,
+      "step": 30312
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0623153916307526e-06,
+      "loss": 0.8943,
+      "step": 30313
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0617967506657854e-06,
+      "loss": 0.7265,
+      "step": 30314
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0612781674283142e-06,
+      "loss": 0.6758,
+      "step": 30315
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.06075964192211e-06,
+      "loss": 0.8639,
+      "step": 30316
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.060241174150942e-06,
+      "loss": 0.877,
+      "step": 30317
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0597227641185837e-06,
+      "loss": 0.7521,
+      "step": 30318
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.059204411828801e-06,
+      "loss": 0.7142,
+      "step": 30319
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.058686117285367e-06,
+      "loss": 0.9123,
+      "step": 30320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.058167880492047e-06,
+      "loss": 1.0219,
+      "step": 30321
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0576497014526144e-06,
+      "loss": 0.9181,
+      "step": 30322
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0571315801708346e-06,
+      "loss": 0.6747,
+      "step": 30323
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0566135166504788e-06,
+      "loss": 0.9196,
+      "step": 30324
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0560955108953097e-06,
+      "loss": 0.7689,
+      "step": 30325
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.055577562909096e-06,
+      "loss": 0.7653,
+      "step": 30326
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0550596726956007e-06,
+      "loss": 1.0253,
+      "step": 30327
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.054541840258597e-06,
+      "loss": 0.6299,
+      "step": 30328
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0540240656018475e-06,
+      "loss": 0.6079,
+      "step": 30329
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0535063487291176e-06,
+      "loss": 0.8272,
+      "step": 30330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.052988689644172e-06,
+      "loss": 0.989,
+      "step": 30331
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.052471088350775e-06,
+      "loss": 0.9183,
+      "step": 30332
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0519535448526907e-06,
+      "loss": 0.7282,
+      "step": 30333
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.051436059153683e-06,
+      "loss": 0.8006,
+      "step": 30334
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0509186312575146e-06,
+      "loss": 0.6647,
+      "step": 30335
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.050401261167948e-06,
+      "loss": 0.9478,
+      "step": 30336
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.049883948888748e-06,
+      "loss": 0.7881,
+      "step": 30337
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0493666944236735e-06,
+      "loss": 0.8192,
+      "step": 30338
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.048849497776485e-06,
+      "loss": 0.7151,
+      "step": 30339
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0483323589509484e-06,
+      "loss": 0.9179,
+      "step": 30340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.047815277950822e-06,
+      "loss": 0.9756,
+      "step": 30341
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.047298254779866e-06,
+      "loss": 1.0537,
+      "step": 30342
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0467812894418406e-06,
+      "loss": 0.549,
+      "step": 30343
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.046264381940507e-06,
+      "loss": 0.8172,
+      "step": 30344
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.045747532279616e-06,
+      "loss": 0.9263,
+      "step": 30345
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0452307404629357e-06,
+      "loss": 1.0145,
+      "step": 30346
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0447140064942216e-06,
+      "loss": 0.7622,
+      "step": 30347
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.04419733037723e-06,
+      "loss": 0.5853,
+      "step": 30348
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0436807121157186e-06,
+      "loss": 0.9049,
+      "step": 30349
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0431641517134447e-06,
+      "loss": 0.6764,
+      "step": 30350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.042647649174164e-06,
+      "loss": 0.8449,
+      "step": 30351
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0421312045016316e-06,
+      "loss": 0.6928,
+      "step": 30352
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0416148176996076e-06,
+      "loss": 0.387,
+      "step": 30353
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0410984887718468e-06,
+      "loss": 0.6956,
+      "step": 30354
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.040582217722098e-06,
+      "loss": 0.8519,
+      "step": 30355
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.040066004554121e-06,
+      "loss": 0.9786,
+      "step": 30356
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.039549849271664e-06,
+      "loss": 1.0268,
+      "step": 30357
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0390337518784875e-06,
+      "loss": 0.5762,
+      "step": 30358
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0385177123783407e-06,
+      "loss": 0.779,
+      "step": 30359
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.038001730774978e-06,
+      "loss": 0.7759,
+      "step": 30360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0374858070721514e-06,
+      "loss": 0.9499,
+      "step": 30361
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.036969941273611e-06,
+      "loss": 0.9526,
+      "step": 30362
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.036454133383111e-06,
+      "loss": 0.8641,
+      "step": 30363
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0359383834044e-06,
+      "loss": 0.6749,
+      "step": 30364
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0354226913412288e-06,
+      "loss": 0.6947,
+      "step": 30365
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0349070571973496e-06,
+      "loss": 0.7742,
+      "step": 30366
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.03439148097651e-06,
+      "loss": 0.9648,
+      "step": 30367
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.03387596268246e-06,
+      "loss": 0.8097,
+      "step": 30368
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0333605023189485e-06,
+      "loss": 0.8991,
+      "step": 30369
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.032845099889722e-06,
+      "loss": 0.8776,
+      "step": 30370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0323297553985333e-06,
+      "loss": 0.879,
+      "step": 30371
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.031814468849127e-06,
+      "loss": 0.8151,
+      "step": 30372
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0312992402452513e-06,
+      "loss": 0.6566,
+      "step": 30373
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0307840695906545e-06,
+      "loss": 0.8121,
+      "step": 30374
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0302689568890753e-06,
+      "loss": 0.8199,
+      "step": 30375
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.029753902144269e-06,
+      "loss": 0.879,
+      "step": 30376
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.029238905359977e-06,
+      "loss": 1.148,
+      "step": 30377
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0287239665399438e-06,
+      "loss": 0.702,
+      "step": 30378
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0282090856879167e-06,
+      "loss": 1.0819,
+      "step": 30379
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.027694262807638e-06,
+      "loss": 0.6067,
+      "step": 30380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0271794979028515e-06,
+      "loss": 0.9947,
+      "step": 30381
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0266647909772973e-06,
+      "loss": 0.7779,
+      "step": 30382
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.026150142034726e-06,
+      "loss": 0.7463,
+      "step": 30383
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0256355510788796e-06,
+      "loss": 0.8329,
+      "step": 30384
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.025121018113494e-06,
+      "loss": 0.8356,
+      "step": 30385
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0246065431423136e-06,
+      "loss": 0.5559,
+      "step": 30386
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0240921261690795e-06,
+      "loss": 0.9142,
+      "step": 30387
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.023577767197531e-06,
+      "loss": 1.0277,
+      "step": 30388
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.023063466231414e-06,
+      "loss": 0.6631,
+      "step": 30389
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.022549223274465e-06,
+      "loss": 0.817,
+      "step": 30390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0220350383304234e-06,
+      "loss": 1.0315,
+      "step": 30391
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0215209114030286e-06,
+      "loss": 0.8203,
+      "step": 30392
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0210068424960206e-06,
+      "loss": 0.7584,
+      "step": 30393
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0204928316131366e-06,
+      "loss": 0.6963,
+      "step": 30394
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.019978878758115e-06,
+      "loss": 0.4346,
+      "step": 30395
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0194649839346937e-06,
+      "loss": 1.0091,
+      "step": 30396
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0189511471466083e-06,
+      "loss": 0.8781,
+      "step": 30397
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0184373683975965e-06,
+      "loss": 1.03,
+      "step": 30398
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0179236476913954e-06,
+      "loss": 1.0446,
+      "step": 30399
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0174099850317363e-06,
+      "loss": 1.102,
+      "step": 30400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.016896380422361e-06,
+      "loss": 0.729,
+      "step": 30401
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0163828338670023e-06,
+      "loss": 0.8023,
+      "step": 30402
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0158693453693933e-06,
+      "loss": 0.5642,
+      "step": 30403
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0153559149332725e-06,
+      "loss": 1.0012,
+      "step": 30404
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0148425425623673e-06,
+      "loss": 0.7734,
+      "step": 30405
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.014329228260411e-06,
+      "loss": 0.8039,
+      "step": 30406
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0138159720311425e-06,
+      "loss": 0.8732,
+      "step": 30407
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0133027738782906e-06,
+      "loss": 0.6803,
+      "step": 30408
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0127896338055876e-06,
+      "loss": 0.6466,
+      "step": 30409
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0122765518167654e-06,
+      "loss": 0.739,
+      "step": 30410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0117635279155557e-06,
+      "loss": 0.8174,
+      "step": 30411
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0112505621056887e-06,
+      "loss": 0.7458,
+      "step": 30412
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.010737654390892e-06,
+      "loss": 0.9071,
+      "step": 30413
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.010224804774903e-06,
+      "loss": 0.8697,
+      "step": 30414
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0097120132614446e-06,
+      "loss": 1.0067,
+      "step": 30415
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.009199279854247e-06,
+      "loss": 0.6682,
+      "step": 30416
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.00868660455704e-06,
+      "loss": 0.6362,
+      "step": 30417
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0081739873735486e-06,
+      "loss": 0.8154,
+      "step": 30418
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0076614283075057e-06,
+      "loss": 0.7498,
+      "step": 30419
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0071489273626376e-06,
+      "loss": 0.664,
+      "step": 30420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.006636484542669e-06,
+      "loss": 0.9,
+      "step": 30421
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0061240998513274e-06,
+      "loss": 0.5127,
+      "step": 30422
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0056117732923396e-06,
+      "loss": 0.9592,
+      "step": 30423
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.005099504869431e-06,
+      "loss": 0.9351,
+      "step": 30424
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0045872945863265e-06,
+      "loss": 0.8479,
+      "step": 30425
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.004075142446751e-06,
+      "loss": 0.6461,
+      "step": 30426
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.003563048454429e-06,
+      "loss": 0.7754,
+      "step": 30427
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0030510126130842e-06,
+      "loss": 0.9104,
+      "step": 30428
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.002539034926441e-06,
+      "loss": 0.8113,
+      "step": 30429
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0020271153982205e-06,
+      "loss": 0.9948,
+      "step": 30430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001515254032146e-06,
+      "loss": 0.5623,
+      "step": 30431
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001003450831943e-06,
+      "loss": 0.575,
+      "step": 30432
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0004917058013305e-06,
+      "loss": 0.779,
+      "step": 30433
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9999800189440344e-06,
+      "loss": 0.9516,
+      "step": 30434
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.999468390263769e-06,
+      "loss": 0.853,
+      "step": 30435
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.998956819764255e-06,
+      "loss": 0.8787,
+      "step": 30436
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9984453074492183e-06,
+      "loss": 0.9292,
+      "step": 30437
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.997933853322376e-06,
+      "loss": 0.9502,
+      "step": 30438
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9974224573874466e-06,
+      "loss": 0.7932,
+      "step": 30439
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9969111196481506e-06,
+      "loss": 1.0432,
+      "step": 30440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9963998401082054e-06,
+      "loss": 0.9314,
+      "step": 30441
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9958886187713288e-06,
+      "loss": 0.5994,
+      "step": 30442
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9953774556412366e-06,
+      "loss": 0.4967,
+      "step": 30443
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.994866350721655e-06,
+      "loss": 0.7709,
+      "step": 30444
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.99435530401629e-06,
+      "loss": 0.9582,
+      "step": 30445
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.993844315528862e-06,
+      "loss": 0.8028,
+      "step": 30446
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9933333852630888e-06,
+      "loss": 0.9401,
+      "step": 30447
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.992822513222683e-06,
+      "loss": 0.9523,
+      "step": 30448
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9923116994113587e-06,
+      "loss": 0.735,
+      "step": 30449
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9918009438328365e-06,
+      "loss": 1.1419,
+      "step": 30450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9912902464908267e-06,
+      "loss": 0.8095,
+      "step": 30451
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.990779607389043e-06,
+      "loss": 0.813,
+      "step": 30452
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9902690265311997e-06,
+      "loss": 0.6614,
+      "step": 30453
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9897585039210098e-06,
+      "loss": 1.0303,
+      "step": 30454
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.989248039562185e-06,
+      "loss": 0.7081,
+      "step": 30455
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9887376334584385e-06,
+      "loss": 0.7983,
+      "step": 30456
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.988227285613481e-06,
+      "loss": 0.5863,
+      "step": 30457
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.987716996031025e-06,
+      "loss": 0.8149,
+      "step": 30458
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.987206764714781e-06,
+      "loss": 0.7798,
+      "step": 30459
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.986696591668459e-06,
+      "loss": 0.7365,
+      "step": 30460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9861864768957664e-06,
+      "loss": 0.9581,
+      "step": 30461
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9856764204004174e-06,
+      "loss": 1.0631,
+      "step": 30462
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.985166422186121e-06,
+      "loss": 0.8272,
+      "step": 30463
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9846564822565863e-06,
+      "loss": 0.7438,
+      "step": 30464
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9841466006155162e-06,
+      "loss": 0.6981,
+      "step": 30465
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9836367772666234e-06,
+      "loss": 0.7465,
+      "step": 30466
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9831270122136114e-06,
+      "loss": 0.6148,
+      "step": 30467
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.982617305460193e-06,
+      "loss": 0.7147,
+      "step": 30468
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9821076570100706e-06,
+      "loss": 0.6246,
+      "step": 30469
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.981598066866952e-06,
+      "loss": 0.6061,
+      "step": 30470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9810885350345433e-06,
+      "loss": 0.7247,
+      "step": 30471
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.980579061516549e-06,
+      "loss": 0.8549,
+      "step": 30472
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.980069646316675e-06,
+      "loss": 0.832,
+      "step": 30473
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9795602894386245e-06,
+      "loss": 0.9806,
+      "step": 30474
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.979050990886102e-06,
+      "loss": 0.7354,
+      "step": 30475
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.978541750662811e-06,
+      "loss": 0.9075,
+      "step": 30476
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9780325687724554e-06,
+      "loss": 0.8424,
+      "step": 30477
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9775234452187386e-06,
+      "loss": 0.926,
+      "step": 30478
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9770143800053587e-06,
+      "loss": 1.181,
+      "step": 30479
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.976505373136025e-06,
+      "loss": 0.7385,
+      "step": 30480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.975996424614434e-06,
+      "loss": 0.7512,
+      "step": 30481
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.975487534444288e-06,
+      "loss": 0.8862,
+      "step": 30482
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.974978702629289e-06,
+      "loss": 0.8305,
+      "step": 30483
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9744699291731373e-06,
+      "loss": 0.7068,
+      "step": 30484
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9739612140795273e-06,
+      "loss": 0.7538,
+      "step": 30485
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9734525573521645e-06,
+      "loss": 1.1536,
+      "step": 30486
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9729439589947464e-06,
+      "loss": 0.8293,
+      "step": 30487
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9724354190109706e-06,
+      "loss": 0.629,
+      "step": 30488
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9719269374045356e-06,
+      "loss": 0.7786,
+      "step": 30489
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.971418514179141e-06,
+      "loss": 0.7039,
+      "step": 30490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.970910149338482e-06,
+      "loss": 0.8287,
+      "step": 30491
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9704018428862526e-06,
+      "loss": 0.8212,
+      "step": 30492
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9698935948261555e-06,
+      "loss": 0.8557,
+      "step": 30493
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9693854051618867e-06,
+      "loss": 0.8793,
+      "step": 30494
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.968877273897136e-06,
+      "loss": 1.1633,
+      "step": 30495
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9683692010356025e-06,
+      "loss": 0.7925,
+      "step": 30496
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.967861186580977e-06,
+      "loss": 0.7483,
+      "step": 30497
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.967353230536959e-06,
+      "loss": 0.6485,
+      "step": 30498
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9668453329072414e-06,
+      "loss": 0.8459,
+      "step": 30499
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9663374936955158e-06,
+      "loss": 1.036,
+      "step": 30500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9658297129054758e-06,
+      "loss": 0.8277,
+      "step": 30501
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.965321990540815e-06,
+      "loss": 0.5723,
+      "step": 30502
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9648143266052244e-06,
+      "loss": 0.6891,
+      "step": 30503
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.964306721102397e-06,
+      "loss": 0.757,
+      "step": 30504
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.963799174036023e-06,
+      "loss": 0.9609,
+      "step": 30505
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.963291685409794e-06,
+      "loss": 0.667,
+      "step": 30506
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9627842552274002e-06,
+      "loss": 0.576,
+      "step": 30507
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9622768834925322e-06,
+      "loss": 0.5854,
+      "step": 30508
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.961769570208878e-06,
+      "loss": 0.9144,
+      "step": 30509
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9612623153801267e-06,
+      "loss": 0.9012,
+      "step": 30510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9607551190099707e-06,
+      "loss": 0.9263,
+      "step": 30511
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.960247981102097e-06,
+      "loss": 0.7076,
+      "step": 30512
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.959740901660191e-06,
+      "loss": 1.1249,
+      "step": 30513
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9592338806879465e-06,
+      "loss": 0.9487,
+      "step": 30514
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.95872691818904e-06,
+      "loss": 0.8959,
+      "step": 30515
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.958220014167167e-06,
+      "loss": 0.8446,
+      "step": 30516
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.957713168626011e-06,
+      "loss": 0.8944,
+      "step": 30517
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9572063815692576e-06,
+      "loss": 0.7989,
+      "step": 30518
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.956699653000593e-06,
+      "loss": 0.7717,
+      "step": 30519
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9561929829237015e-06,
+      "loss": 0.9423,
+      "step": 30520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.955686371342268e-06,
+      "loss": 1.0186,
+      "step": 30521
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9551798182599745e-06,
+      "loss": 0.6716,
+      "step": 30522
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9546733236805084e-06,
+      "loss": 0.7837,
+      "step": 30523
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.954166887607555e-06,
+      "loss": 0.8328,
+      "step": 30524
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.953660510044789e-06,
+      "loss": 0.8239,
+      "step": 30525
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9531541909958984e-06,
+      "loss": 0.7207,
+      "step": 30526
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.952647930464564e-06,
+      "loss": 0.8294,
+      "step": 30527
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9521417284544654e-06,
+      "loss": 0.7229,
+      "step": 30528
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.951635584969288e-06,
+      "loss": 0.8161,
+      "step": 30529
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9511295000127107e-06,
+      "loss": 0.9207,
+      "step": 30530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9506234735884134e-06,
+      "loss": 0.7766,
+      "step": 30531
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9501175057000753e-06,
+      "loss": 0.7857,
+      "step": 30532
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9496115963513773e-06,
+      "loss": 0.6885,
+      "step": 30533
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.949105745545997e-06,
+      "loss": 0.7532,
+      "step": 30534
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9485999532876155e-06,
+      "loss": 0.92,
+      "step": 30535
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.948094219579908e-06,
+      "loss": 0.6585,
+      "step": 30536
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.947588544426553e-06,
+      "loss": 1.1189,
+      "step": 30537
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9470829278312285e-06,
+      "loss": 1.0835,
+      "step": 30538
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.946577369797612e-06,
+      "loss": 0.806,
+      "step": 30539
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.946071870329377e-06,
+      "loss": 0.8358,
+      "step": 30540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9455664294302023e-06,
+      "loss": 0.8058,
+      "step": 30541
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9450610471037646e-06,
+      "loss": 0.8519,
+      "step": 30542
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.944555723353737e-06,
+      "loss": 0.7489,
+      "step": 30543
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.944050458183797e-06,
+      "loss": 0.983,
+      "step": 30544
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.943545251597615e-06,
+      "loss": 0.8942,
+      "step": 30545
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9430401035988635e-06,
+      "loss": 0.7263,
+      "step": 30546
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9425350141912213e-06,
+      "loss": 0.6149,
+      "step": 30547
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9420299833783597e-06,
+      "loss": 0.9444,
+      "step": 30548
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9415250111639506e-06,
+      "loss": 0.8175,
+      "step": 30549
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.941020097551666e-06,
+      "loss": 0.8277,
+      "step": 30550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.940515242545179e-06,
+      "loss": 0.3924,
+      "step": 30551
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9400104461481597e-06,
+      "loss": 0.9392,
+      "step": 30552
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9395057083642764e-06,
+      "loss": 0.7011,
+      "step": 30553
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9390010291972094e-06,
+      "loss": 0.9116,
+      "step": 30554
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9384964086506185e-06,
+      "loss": 0.8279,
+      "step": 30555
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9379918467281766e-06,
+      "loss": 0.6752,
+      "step": 30556
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.937487343433554e-06,
+      "loss": 0.8752,
+      "step": 30557
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9369828987704153e-06,
+      "loss": 0.8475,
+      "step": 30558
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.936478512742436e-06,
+      "loss": 0.8281,
+      "step": 30559
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9359741853532797e-06,
+      "loss": 0.7701,
+      "step": 30560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.935469916606615e-06,
+      "loss": 0.7727,
+      "step": 30561
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9349657065061077e-06,
+      "loss": 0.5643,
+      "step": 30562
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.934461555055426e-06,
+      "loss": 1.0019,
+      "step": 30563
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.933957462258236e-06,
+      "loss": 0.9167,
+      "step": 30564
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.933453428118203e-06,
+      "loss": 0.9146,
+      "step": 30565
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9329494526389915e-06,
+      "loss": 0.7075,
+      "step": 30566
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9324455358242677e-06,
+      "loss": 0.748,
+      "step": 30567
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9319416776776958e-06,
+      "loss": 0.8053,
+      "step": 30568
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.931437878202941e-06,
+      "loss": 1.1861,
+      "step": 30569
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.930934137403665e-06,
+      "loss": 0.7499,
+      "step": 30570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9304304552835297e-06,
+      "loss": 0.57,
+      "step": 30571
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.929926831846203e-06,
+      "loss": 0.5475,
+      "step": 30572
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.929423267095344e-06,
+      "loss": 0.9957,
+      "step": 30573
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9289197610346177e-06,
+      "loss": 0.5093,
+      "step": 30574
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9284163136676816e-06,
+      "loss": 0.6268,
+      "step": 30575
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.927912924998194e-06,
+      "loss": 0.8418,
+      "step": 30576
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9274095950298246e-06,
+      "loss": 0.8828,
+      "step": 30577
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9269063237662276e-06,
+      "loss": 0.6058,
+      "step": 30578
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9264031112110648e-06,
+      "loss": 0.8955,
+      "step": 30579
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.925899957367996e-06,
+      "loss": 0.7524,
+      "step": 30580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.925396862240678e-06,
+      "loss": 0.8343,
+      "step": 30581
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9248938258327722e-06,
+      "loss": 0.8035,
+      "step": 30582
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9243908481479313e-06,
+      "loss": 0.9547,
+      "step": 30583
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9238879291898227e-06,
+      "loss": 0.6116,
+      "step": 30584
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.923385068962095e-06,
+      "loss": 0.7386,
+      "step": 30585
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.922882267468408e-06,
+      "loss": 0.8625,
+      "step": 30586
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9223795247124177e-06,
+      "loss": 0.832,
+      "step": 30587
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9218768406977807e-06,
+      "loss": 0.5575,
+      "step": 30588
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9213742154281503e-06,
+      "loss": 0.3649,
+      "step": 30589
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9208716489071864e-06,
+      "loss": 0.8159,
+      "step": 30590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9203691411385408e-06,
+      "loss": 0.8768,
+      "step": 30591
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.919866692125868e-06,
+      "loss": 0.6277,
+      "step": 30592
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9193643018728215e-06,
+      "loss": 0.4828,
+      "step": 30593
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9188619703830557e-06,
+      "loss": 0.9402,
+      "step": 30594
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9183596976602227e-06,
+      "loss": 0.9005,
+      "step": 30595
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9178574837079755e-06,
+      "loss": 0.745,
+      "step": 30596
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9173553285299663e-06,
+      "loss": 0.8621,
+      "step": 30597
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.916853232129846e-06,
+      "loss": 0.9298,
+      "step": 30598
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9163511945112667e-06,
+      "loss": 1.0378,
+      "step": 30599
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9158492156778807e-06,
+      "loss": 0.7009,
+      "step": 30600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.915347295633333e-06,
+      "loss": 0.6416,
+      "step": 30601
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9148454343812806e-06,
+      "loss": 0.9093,
+      "step": 30602
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9143436319253695e-06,
+      "loss": 1.0035,
+      "step": 30603
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9138418882692524e-06,
+      "loss": 0.7943,
+      "step": 30604
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9133402034165727e-06,
+      "loss": 0.7512,
+      "step": 30605
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9128385773709803e-06,
+      "loss": 0.6118,
+      "step": 30606
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9123370101361217e-06,
+      "loss": 0.9608,
+      "step": 30607
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.911835501715649e-06,
+      "loss": 0.7586,
+      "step": 30608
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9113340521132073e-06,
+      "loss": 0.6981,
+      "step": 30609
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.910832661332441e-06,
+      "loss": 0.7702,
+      "step": 30610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9103313293769997e-06,
+      "loss": 0.4926,
+      "step": 30611
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9098300562505266e-06,
+      "loss": 0.6275,
+      "step": 30612
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9093288419566682e-06,
+      "loss": 0.804,
+      "step": 30613
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9088276864990673e-06,
+      "loss": 0.9236,
+      "step": 30614
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.908326589881372e-06,
+      "loss": 0.6395,
+      "step": 30615
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.907825552107223e-06,
+      "loss": 0.9077,
+      "step": 30616
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9073245731802647e-06,
+      "loss": 1.1522,
+      "step": 30617
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9068236531041417e-06,
+      "loss": 0.8947,
+      "step": 30618
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9063227918824934e-06,
+      "loss": 0.7309,
+      "step": 30619
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9058219895189668e-06,
+      "loss": 0.886,
+      "step": 30620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9053212460172022e-06,
+      "loss": 1.1418,
+      "step": 30621
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.904820561380839e-06,
+      "loss": 0.9072,
+      "step": 30622
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9043199356135232e-06,
+      "loss": 0.8357,
+      "step": 30623
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9038193687188888e-06,
+      "loss": 0.8064,
+      "step": 30624
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.903318860700577e-06,
+      "loss": 0.5641,
+      "step": 30625
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9028184115622317e-06,
+      "loss": 1.0566,
+      "step": 30626
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.902318021307491e-06,
+      "loss": 0.8866,
+      "step": 30627
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9018176899399921e-06,
+      "loss": 1.1137,
+      "step": 30628
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9013174174633741e-06,
+      "loss": 0.8357,
+      "step": 30629
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9008172038812744e-06,
+      "loss": 0.8189,
+      "step": 30630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9003170491973322e-06,
+      "loss": 0.7919,
+      "step": 30631
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8998169534151823e-06,
+      "loss": 0.5575,
+      "step": 30632
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8993169165384672e-06,
+      "loss": 0.767,
+      "step": 30633
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8988169385708165e-06,
+      "loss": 0.6561,
+      "step": 30634
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8983170195158684e-06,
+      "loss": 1.0471,
+      "step": 30635
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8978171593772587e-06,
+      "loss": 0.926,
+      "step": 30636
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8973173581586202e-06,
+      "loss": 0.6511,
+      "step": 30637
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.896817615863592e-06,
+      "loss": 0.8312,
+      "step": 30638
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8963179324958059e-06,
+      "loss": 0.7581,
+      "step": 30639
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8958183080588955e-06,
+      "loss": 0.9159,
+      "step": 30640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8953187425564946e-06,
+      "loss": 0.7248,
+      "step": 30641
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8948192359922356e-06,
+      "loss": 0.93,
+      "step": 30642
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8943197883697506e-06,
+      "loss": 0.6531,
+      "step": 30643
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.893820399692673e-06,
+      "loss": 0.7889,
+      "step": 30644
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8933210699646342e-06,
+      "loss": 0.8945,
+      "step": 30645
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8928217991892639e-06,
+      "loss": 0.7947,
+      "step": 30646
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8923225873701934e-06,
+      "loss": 0.7477,
+      "step": 30647
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8918234345110543e-06,
+      "loss": 1.0189,
+      "step": 30648
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8913243406154747e-06,
+      "loss": 0.5575,
+      "step": 30649
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8908253056870828e-06,
+      "loss": 0.6648,
+      "step": 30650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8903263297295127e-06,
+      "loss": 0.7804,
+      "step": 30651
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.88982741274639e-06,
+      "loss": 0.9046,
+      "step": 30652
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8893285547413454e-06,
+      "loss": 0.8201,
+      "step": 30653
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8888297557180013e-06,
+      "loss": 0.7397,
+      "step": 30654
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8883310156799861e-06,
+      "loss": 0.7184,
+      "step": 30655
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8878323346309302e-06,
+      "loss": 0.9153,
+      "step": 30656
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8873337125744595e-06,
+      "loss": 1.0799,
+      "step": 30657
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.886835149514198e-06,
+      "loss": 0.7421,
+      "step": 30658
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8863366454537724e-06,
+      "loss": 0.8265,
+      "step": 30659
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.885838200396808e-06,
+      "loss": 0.9576,
+      "step": 30660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8853398143469292e-06,
+      "loss": 0.5557,
+      "step": 30661
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.884841487307758e-06,
+      "loss": 1.0109,
+      "step": 30662
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8843432192829258e-06,
+      "loss": 0.9606,
+      "step": 30663
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8838450102760475e-06,
+      "loss": 0.8827,
+      "step": 30664
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8833468602907502e-06,
+      "loss": 0.8513,
+      "step": 30665
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8828487693306562e-06,
+      "loss": 0.7813,
+      "step": 30666
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8823507373993865e-06,
+      "loss": 0.8552,
+      "step": 30667
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8818527645005625e-06,
+      "loss": 0.9094,
+      "step": 30668
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8813548506378076e-06,
+      "loss": 0.9827,
+      "step": 30669
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8808569958147426e-06,
+      "loss": 0.8926,
+      "step": 30670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8803592000349867e-06,
+      "loss": 0.9314,
+      "step": 30671
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8798614633021617e-06,
+      "loss": 0.6184,
+      "step": 30672
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8793637856198843e-06,
+      "loss": 0.7408,
+      "step": 30673
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8788661669917763e-06,
+      "loss": 0.8466,
+      "step": 30674
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8783686074214546e-06,
+      "loss": 0.7389,
+      "step": 30675
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8778711069125387e-06,
+      "loss": 0.7294,
+      "step": 30676
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8773736654686458e-06,
+      "loss": 1.0276,
+      "step": 30677
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8768762830933939e-06,
+      "loss": 0.6961,
+      "step": 30678
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8763789597903992e-06,
+      "loss": 0.8824,
+      "step": 30679
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8758816955632764e-06,
+      "loss": 0.946,
+      "step": 30680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.875384490415647e-06,
+      "loss": 0.9039,
+      "step": 30681
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.874887344351124e-06,
+      "loss": 0.8366,
+      "step": 30682
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8743902573733242e-06,
+      "loss": 0.7731,
+      "step": 30683
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8738932294858592e-06,
+      "loss": 0.574,
+      "step": 30684
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8733962606923439e-06,
+      "loss": 0.7141,
+      "step": 30685
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8728993509963922e-06,
+      "loss": 0.8673,
+      "step": 30686
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.87240250040162e-06,
+      "loss": 0.8484,
+      "step": 30687
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.871905708911641e-06,
+      "loss": 0.786,
+      "step": 30688
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8714089765300647e-06,
+      "loss": 0.9467,
+      "step": 30689
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8709123032605058e-06,
+      "loss": 0.9563,
+      "step": 30690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.870415689106575e-06,
+      "loss": 0.7256,
+      "step": 30691
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8699191340718848e-06,
+      "loss": 0.8798,
+      "step": 30692
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8694226381600443e-06,
+      "loss": 0.5441,
+      "step": 30693
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8689262013746667e-06,
+      "loss": 0.9393,
+      "step": 30694
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.86842982371936e-06,
+      "loss": 0.8112,
+      "step": 30695
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8679335051977344e-06,
+      "loss": 0.5717,
+      "step": 30696
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8674372458133994e-06,
+      "loss": 0.9329,
+      "step": 30697
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.866941045569961e-06,
+      "loss": 0.9109,
+      "step": 30698
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8664449044710343e-06,
+      "loss": 0.8682,
+      "step": 30699
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8659488225202227e-06,
+      "loss": 0.7857,
+      "step": 30700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8654527997211346e-06,
+      "loss": 0.7168,
+      "step": 30701
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8649568360773763e-06,
+      "loss": 0.9246,
+      "step": 30702
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8644609315925588e-06,
+      "loss": 0.7393,
+      "step": 30703
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8639650862702796e-06,
+      "loss": 0.7919,
+      "step": 30704
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8634693001141513e-06,
+      "loss": 1.007,
+      "step": 30705
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8629735731277787e-06,
+      "loss": 0.6476,
+      "step": 30706
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8624779053147657e-06,
+      "loss": 1.0888,
+      "step": 30707
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8619822966787172e-06,
+      "loss": 0.4785,
+      "step": 30708
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8614867472232378e-06,
+      "loss": 0.6878,
+      "step": 30709
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8609912569519296e-06,
+      "loss": 0.7484,
+      "step": 30710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8604958258683936e-06,
+      "loss": 0.8316,
+      "step": 30711
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8600004539762396e-06,
+      "loss": 0.4361,
+      "step": 30712
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8595051412790688e-06,
+      "loss": 0.7576,
+      "step": 30713
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8590098877804774e-06,
+      "loss": 0.8659,
+      "step": 30714
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8585146934840703e-06,
+      "loss": 0.7705,
+      "step": 30715
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8580195583934458e-06,
+      "loss": 0.4743,
+      "step": 30716
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8575244825122096e-06,
+      "loss": 0.5907,
+      "step": 30717
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8570294658439604e-06,
+      "loss": 0.6215,
+      "step": 30718
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.856534508392297e-06,
+      "loss": 0.9905,
+      "step": 30719
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.856039610160818e-06,
+      "loss": 1.0224,
+      "step": 30720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8555447711531238e-06,
+      "loss": 0.8174,
+      "step": 30721
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8550499913728127e-06,
+      "loss": 0.9695,
+      "step": 30722
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8545552708234826e-06,
+      "loss": 0.8554,
+      "step": 30723
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.854060609508731e-06,
+      "loss": 0.5859,
+      "step": 30724
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.853566007432155e-06,
+      "loss": 0.7902,
+      "step": 30725
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8530714645973513e-06,
+      "loss": 0.7998,
+      "step": 30726
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8525769810079175e-06,
+      "loss": 0.5787,
+      "step": 30727
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8520825566674483e-06,
+      "loss": 0.4717,
+      "step": 30728
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8515881915795364e-06,
+      "loss": 0.9737,
+      "step": 30729
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8510938857477834e-06,
+      "loss": 0.823,
+      "step": 30730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8505996391757807e-06,
+      "loss": 0.8946,
+      "step": 30731
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8501054518671212e-06,
+      "loss": 0.6223,
+      "step": 30732
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8496113238254033e-06,
+      "loss": 0.6736,
+      "step": 30733
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8491172550542125e-06,
+      "loss": 0.894,
+      "step": 30734
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8486232455571473e-06,
+      "loss": 0.5529,
+      "step": 30735
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8481292953378006e-06,
+      "loss": 0.6523,
+      "step": 30736
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8476354043997612e-06,
+      "loss": 0.8865,
+      "step": 30737
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8471415727466234e-06,
+      "loss": 1.2014,
+      "step": 30738
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8466478003819778e-06,
+      "loss": 0.9785,
+      "step": 30739
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8461540873094141e-06,
+      "loss": 0.9033,
+      "step": 30740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8456604335325213e-06,
+      "loss": 0.8096,
+      "step": 30741
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8451668390548926e-06,
+      "loss": 0.7845,
+      "step": 30742
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.84467330388012e-06,
+      "loss": 0.5039,
+      "step": 30743
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8441798280117862e-06,
+      "loss": 0.92,
+      "step": 30744
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8436864114534826e-06,
+      "loss": 1.0182,
+      "step": 30745
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8431930542087962e-06,
+      "loss": 0.7357,
+      "step": 30746
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8426997562813131e-06,
+      "loss": 0.8146,
+      "step": 30747
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8422065176746263e-06,
+      "loss": 0.8561,
+      "step": 30748
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8417133383923202e-06,
+      "loss": 0.8389,
+      "step": 30749
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8412202184379801e-06,
+      "loss": 0.8589,
+      "step": 30750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8407271578151919e-06,
+      "loss": 0.7562,
+      "step": 30751
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8402341565275427e-06,
+      "loss": 0.7055,
+      "step": 30752
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8397412145786164e-06,
+      "loss": 0.7773,
+      "step": 30753
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8392483319719989e-06,
+      "loss": 0.6434,
+      "step": 30754
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8387555087112729e-06,
+      "loss": 0.3906,
+      "step": 30755
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8382627448000235e-06,
+      "loss": 1.0093,
+      "step": 30756
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8377700402418342e-06,
+      "loss": 0.7643,
+      "step": 30757
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8372773950402866e-06,
+      "loss": 0.9401,
+      "step": 30758
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8367848091989625e-06,
+      "loss": 0.9295,
+      "step": 30759
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.836292282721448e-06,
+      "loss": 0.8902,
+      "step": 30760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8357998156113233e-06,
+      "loss": 0.7971,
+      "step": 30761
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8353074078721677e-06,
+      "loss": 0.7677,
+      "step": 30762
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8348150595075664e-06,
+      "loss": 1.0456,
+      "step": 30763
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8343227705210942e-06,
+      "loss": 0.8916,
+      "step": 30764
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8338305409163314e-06,
+      "loss": 0.6091,
+      "step": 30765
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8333383706968611e-06,
+      "loss": 0.8505,
+      "step": 30766
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8328462598662621e-06,
+      "loss": 0.7672,
+      "step": 30767
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.832354208428112e-06,
+      "loss": 0.7564,
+      "step": 30768
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.831862216385989e-06,
+      "loss": 0.7348,
+      "step": 30769
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.831370283743471e-06,
+      "loss": 0.9069,
+      "step": 30770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8308784105041355e-06,
+      "loss": 0.986,
+      "step": 30771
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8303865966715563e-06,
+      "loss": 0.7365,
+      "step": 30772
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.829894842249318e-06,
+      "loss": 0.8759,
+      "step": 30773
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.82940314724099e-06,
+      "loss": 0.8148,
+      "step": 30774
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8289115116501487e-06,
+      "loss": 0.9344,
+      "step": 30775
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8284199354803712e-06,
+      "loss": 0.8132,
+      "step": 30776
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8279284187352276e-06,
+      "loss": 0.792,
+      "step": 30777
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8274369614182996e-06,
+      "loss": 0.8386,
+      "step": 30778
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.826945563533158e-06,
+      "loss": 0.5914,
+      "step": 30779
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.826454225083375e-06,
+      "loss": 0.7495,
+      "step": 30780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8259629460725236e-06,
+      "loss": 0.775,
+      "step": 30781
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8254717265041789e-06,
+      "loss": 1.0429,
+      "step": 30782
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.824980566381911e-06,
+      "loss": 0.8687,
+      "step": 30783
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8244894657092925e-06,
+      "loss": 0.5751,
+      "step": 30784
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8239984244898945e-06,
+      "loss": 0.6126,
+      "step": 30785
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.823507442727287e-06,
+      "loss": 0.69,
+      "step": 30786
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8230165204250428e-06,
+      "loss": 0.5709,
+      "step": 30787
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.822525657586729e-06,
+      "loss": 1.0931,
+      "step": 30788
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8220348542159173e-06,
+      "loss": 0.9109,
+      "step": 30789
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8215441103161747e-06,
+      "loss": 0.9884,
+      "step": 30790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8210534258910729e-06,
+      "loss": 0.6476,
+      "step": 30791
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8205628009441789e-06,
+      "loss": 0.674,
+      "step": 30792
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8200722354790633e-06,
+      "loss": 0.7413,
+      "step": 30793
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8195817294992878e-06,
+      "loss": 0.987,
+      "step": 30794
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8190912830084207e-06,
+      "loss": 0.797,
+      "step": 30795
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8186008960100322e-06,
+      "loss": 0.8868,
+      "step": 30796
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8181105685076872e-06,
+      "loss": 0.7903,
+      "step": 30797
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8176203005049497e-06,
+      "loss": 1.0121,
+      "step": 30798
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.817130092005387e-06,
+      "loss": 0.5839,
+      "step": 30799
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8166399430125626e-06,
+      "loss": 0.9454,
+      "step": 30800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8161498535300404e-06,
+      "loss": 1.1747,
+      "step": 30801
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8156598235613843e-06,
+      "loss": 0.6726,
+      "step": 30802
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8151698531101624e-06,
+      "loss": 0.8651,
+      "step": 30803
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.814679942179932e-06,
+      "loss": 0.9606,
+      "step": 30804
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8141900907742592e-06,
+      "loss": 1.1486,
+      "step": 30805
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8137002988967046e-06,
+      "loss": 0.8225,
+      "step": 30806
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8132105665508292e-06,
+      "loss": 0.7231,
+      "step": 30807
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.812720893740194e-06,
+      "loss": 0.5527,
+      "step": 30808
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8122312804683651e-06,
+      "loss": 0.8386,
+      "step": 30809
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.811741726738898e-06,
+      "loss": 0.8777,
+      "step": 30810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.811252232555355e-06,
+      "loss": 0.6395,
+      "step": 30811
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.810762797921296e-06,
+      "loss": 0.7107,
+      "step": 30812
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8102734228402785e-06,
+      "loss": 0.6175,
+      "step": 30813
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8097841073158618e-06,
+      "loss": 0.6656,
+      "step": 30814
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8092948513516051e-06,
+      "loss": 0.8456,
+      "step": 30815
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8088056549510657e-06,
+      "loss": 0.821,
+      "step": 30816
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8083165181178008e-06,
+      "loss": 0.8571,
+      "step": 30817
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8078274408553676e-06,
+      "loss": 0.9032,
+      "step": 30818
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8073384231673241e-06,
+      "loss": 1.0818,
+      "step": 30819
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8068494650572245e-06,
+      "loss": 0.9785,
+      "step": 30820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8063605665286232e-06,
+      "loss": 0.7215,
+      "step": 30821
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8058717275850801e-06,
+      "loss": 1.2181,
+      "step": 30822
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.80538294823015e-06,
+      "loss": 0.801,
+      "step": 30823
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8048942284673831e-06,
+      "loss": 0.9444,
+      "step": 30824
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8044055683003358e-06,
+      "loss": 0.7645,
+      "step": 30825
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8039169677325573e-06,
+      "loss": 0.4796,
+      "step": 30826
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8034284267676083e-06,
+      "loss": 0.6288,
+      "step": 30827
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.802939945409038e-06,
+      "loss": 0.8914,
+      "step": 30828
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.802451523660399e-06,
+      "loss": 0.8684,
+      "step": 30829
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.801963161525242e-06,
+      "loss": 0.5607,
+      "step": 30830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8014748590071195e-06,
+      "loss": 0.7283,
+      "step": 30831
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8009866161095823e-06,
+      "loss": 1.0918,
+      "step": 30832
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8004984328361818e-06,
+      "loss": 0.9721,
+      "step": 30833
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8000103091904663e-06,
+      "loss": 0.8199,
+      "step": 30834
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7995222451759863e-06,
+      "loss": 0.8682,
+      "step": 30835
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7990342407962914e-06,
+      "loss": 0.7088,
+      "step": 30836
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7985462960549305e-06,
+      "loss": 1.2885,
+      "step": 30837
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7980584109554512e-06,
+      "loss": 0.7767,
+      "step": 30838
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7975705855014004e-06,
+      "loss": 0.8707,
+      "step": 30839
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7970828196963286e-06,
+      "loss": 0.6498,
+      "step": 30840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.796595113543782e-06,
+      "loss": 0.8802,
+      "step": 30841
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7961074670473067e-06,
+      "loss": 0.7163,
+      "step": 30842
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7956198802104508e-06,
+      "loss": 1.0008,
+      "step": 30843
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7951323530367537e-06,
+      "loss": 0.6247,
+      "step": 30844
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.794644885529767e-06,
+      "loss": 0.8068,
+      "step": 30845
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7941574776930349e-06,
+      "loss": 0.7895,
+      "step": 30846
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7936701295301006e-06,
+      "loss": 1.0746,
+      "step": 30847
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7931828410445084e-06,
+      "loss": 1.0387,
+      "step": 30848
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.792695612239802e-06,
+      "loss": 0.7169,
+      "step": 30849
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7922084431195242e-06,
+      "loss": 0.8492,
+      "step": 30850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7917213336872153e-06,
+      "loss": 0.6366,
+      "step": 30851
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7912342839464236e-06,
+      "loss": 0.6485,
+      "step": 30852
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7907472939006897e-06,
+      "loss": 0.8821,
+      "step": 30853
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.790260363553551e-06,
+      "loss": 0.6539,
+      "step": 30854
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7897734929085508e-06,
+      "loss": 1.0779,
+      "step": 30855
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.789286681969229e-06,
+      "loss": 0.8137,
+      "step": 30856
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7887999307391236e-06,
+      "loss": 0.9037,
+      "step": 30857
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7883132392217794e-06,
+      "loss": 0.716,
+      "step": 30858
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.787826607420734e-06,
+      "loss": 0.8092,
+      "step": 30859
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7873400353395244e-06,
+      "loss": 0.8744,
+      "step": 30860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7868535229816908e-06,
+      "loss": 0.7693,
+      "step": 30861
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7863670703507707e-06,
+      "loss": 0.8545,
+      "step": 30862
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7858806774503e-06,
+      "loss": 0.8918,
+      "step": 30863
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7853943442838184e-06,
+      "loss": 0.7742,
+      "step": 30864
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7849080708548615e-06,
+      "loss": 0.6727,
+      "step": 30865
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7844218571669648e-06,
+      "loss": 0.7533,
+      "step": 30866
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7839357032236648e-06,
+      "loss": 0.8529,
+      "step": 30867
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7834496090284959e-06,
+      "loss": 0.7146,
+      "step": 30868
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7829635745849926e-06,
+      "loss": 0.959,
+      "step": 30869
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7824775998966926e-06,
+      "loss": 1.0342,
+      "step": 30870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7819916849671282e-06,
+      "loss": 0.7214,
+      "step": 30871
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7815058297998334e-06,
+      "loss": 0.8697,
+      "step": 30872
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7810200343983418e-06,
+      "loss": 0.6261,
+      "step": 30873
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7805342987661845e-06,
+      "loss": 0.8747,
+      "step": 30874
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7800486229068903e-06,
+      "loss": 1.129,
+      "step": 30875
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7795630068239989e-06,
+      "loss": 0.7129,
+      "step": 30876
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7790774505210383e-06,
+      "loss": 0.7742,
+      "step": 30877
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7785919540015385e-06,
+      "loss": 0.9097,
+      "step": 30878
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7781065172690314e-06,
+      "loss": 0.8,
+      "step": 30879
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7776211403270472e-06,
+      "loss": 0.8916,
+      "step": 30880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7771358231791148e-06,
+      "loss": 1.003,
+      "step": 30881
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7766505658287614e-06,
+      "loss": 0.786,
+      "step": 30882
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7761653682795232e-06,
+      "loss": 0.6437,
+      "step": 30883
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7756802305349209e-06,
+      "loss": 0.7946,
+      "step": 30884
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7751951525984857e-06,
+      "loss": 0.7206,
+      "step": 30885
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.774710134473745e-06,
+      "loss": 0.8303,
+      "step": 30886
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7742251761642227e-06,
+      "loss": 0.8685,
+      "step": 30887
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7737402776734502e-06,
+      "loss": 0.5746,
+      "step": 30888
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7732554390049529e-06,
+      "loss": 0.7207,
+      "step": 30889
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7727706601622552e-06,
+      "loss": 0.8101,
+      "step": 30890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7722859411488824e-06,
+      "loss": 0.6366,
+      "step": 30891
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7718012819683605e-06,
+      "loss": 0.8885,
+      "step": 30892
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7713166826242133e-06,
+      "loss": 0.6645,
+      "step": 30893
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.770832143119965e-06,
+      "loss": 0.9316,
+      "step": 30894
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7703476634591389e-06,
+      "loss": 0.9495,
+      "step": 30895
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.769863243645259e-06,
+      "loss": 0.7356,
+      "step": 30896
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.769378883681847e-06,
+      "loss": 0.9043,
+      "step": 30897
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7688945835724258e-06,
+      "loss": 0.6559,
+      "step": 30898
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.768410343320518e-06,
+      "loss": 0.9723,
+      "step": 30899
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7679261629296408e-06,
+      "loss": 0.7579,
+      "step": 30900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7674420424033211e-06,
+      "loss": 0.7993,
+      "step": 30901
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7669579817450776e-06,
+      "loss": 0.9759,
+      "step": 30902
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.766473980958433e-06,
+      "loss": 0.8089,
+      "step": 30903
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.765990040046901e-06,
+      "loss": 0.842,
+      "step": 30904
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7655061590140011e-06,
+      "loss": 1.0535,
+      "step": 30905
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7650223378632569e-06,
+      "loss": 0.708,
+      "step": 30906
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7645385765981848e-06,
+      "loss": 0.8958,
+      "step": 30907
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.764054875222303e-06,
+      "loss": 0.6961,
+      "step": 30908
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7635712337391287e-06,
+      "loss": 0.9021,
+      "step": 30909
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.763087652152179e-06,
+      "loss": 0.8358,
+      "step": 30910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7626041304649711e-06,
+      "loss": 0.8458,
+      "step": 30911
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7621206686810177e-06,
+      "loss": 0.9853,
+      "step": 30912
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7616372668038428e-06,
+      "loss": 0.9675,
+      "step": 30913
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7611539248369536e-06,
+      "loss": 0.7399,
+      "step": 30914
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7606706427838682e-06,
+      "loss": 0.8457,
+      "step": 30915
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7601874206481007e-06,
+      "loss": 0.8968,
+      "step": 30916
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.759704258433166e-06,
+      "loss": 0.6799,
+      "step": 30917
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7592211561425742e-06,
+      "loss": 0.7122,
+      "step": 30918
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7587381137798432e-06,
+      "loss": 0.6209,
+      "step": 30919
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7582551313484841e-06,
+      "loss": 0.6697,
+      "step": 30920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7577722088520087e-06,
+      "loss": 0.6515,
+      "step": 30921
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.75728934629393e-06,
+      "loss": 0.8035,
+      "step": 30922
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.756806543677757e-06,
+      "loss": 1.167,
+      "step": 30923
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7563238010070038e-06,
+      "loss": 0.7672,
+      "step": 30924
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7558411182851787e-06,
+      "loss": 0.7504,
+      "step": 30925
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.755358495515792e-06,
+      "loss": 0.5086,
+      "step": 30926
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7548759327023545e-06,
+      "loss": 0.6255,
+      "step": 30927
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7543934298483745e-06,
+      "loss": 0.9939,
+      "step": 30928
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7539109869573612e-06,
+      "loss": 1.0913,
+      "step": 30929
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7534286040328208e-06,
+      "loss": 0.6512,
+      "step": 30930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7529462810782648e-06,
+      "loss": 0.8212,
+      "step": 30931
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7524640180971997e-06,
+      "loss": 0.6282,
+      "step": 30932
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7519818150931356e-06,
+      "loss": 0.6477,
+      "step": 30933
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.751499672069572e-06,
+      "loss": 0.7681,
+      "step": 30934
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7510175890300197e-06,
+      "loss": 0.9025,
+      "step": 30935
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.75053556597798e-06,
+      "loss": 1.0502,
+      "step": 30936
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7500536029169646e-06,
+      "loss": 0.7145,
+      "step": 30937
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7495716998504752e-06,
+      "loss": 0.8779,
+      "step": 30938
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.749089856782018e-06,
+      "loss": 0.996,
+      "step": 30939
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7486080737150945e-06,
+      "loss": 0.768,
+      "step": 30940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7481263506532097e-06,
+      "loss": 0.7701,
+      "step": 30941
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7476446875998654e-06,
+      "loss": 0.825,
+      "step": 30942
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7471630845585662e-06,
+      "loss": 0.9873,
+      "step": 30943
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7466815415328142e-06,
+      "loss": 0.9519,
+      "step": 30944
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7462000585261096e-06,
+      "loss": 0.6485,
+      "step": 30945
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7457186355419554e-06,
+      "loss": 0.9808,
+      "step": 30946
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7452372725838506e-06,
+      "loss": 0.729,
+      "step": 30947
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7447559696552963e-06,
+      "loss": 0.4925,
+      "step": 30948
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7442747267597949e-06,
+      "loss": 1.3641,
+      "step": 30949
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7437935439008446e-06,
+      "loss": 0.5652,
+      "step": 30950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.743312421081944e-06,
+      "loss": 0.9411,
+      "step": 30951
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7428313583065926e-06,
+      "loss": 0.6658,
+      "step": 30952
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7423503555782906e-06,
+      "loss": 0.8186,
+      "step": 30953
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7418694129005299e-06,
+      "loss": 0.9385,
+      "step": 30954
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.741388530276814e-06,
+      "loss": 0.5164,
+      "step": 30955
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7409077077106374e-06,
+      "loss": 0.7869,
+      "step": 30956
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7404269452054967e-06,
+      "loss": 0.7697,
+      "step": 30957
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7399462427648895e-06,
+      "loss": 0.7408,
+      "step": 30958
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7394656003923105e-06,
+      "loss": 0.4946,
+      "step": 30959
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7389850180912537e-06,
+      "loss": 0.9083,
+      "step": 30960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7385044958652141e-06,
+      "loss": 1.0266,
+      "step": 30961
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.738024033717689e-06,
+      "loss": 0.6946,
+      "step": 30962
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7375436316521733e-06,
+      "loss": 1.0568,
+      "step": 30963
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7370632896721551e-06,
+      "loss": 0.9388,
+      "step": 30964
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7365830077811307e-06,
+      "loss": 0.8748,
+      "step": 30965
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7361027859825896e-06,
+      "loss": 0.8569,
+      "step": 30966
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.735622624280029e-06,
+      "loss": 0.7248,
+      "step": 30967
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7351425226769381e-06,
+      "loss": 0.8689,
+      "step": 30968
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7346624811768086e-06,
+      "loss": 0.8311,
+      "step": 30969
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7341824997831326e-06,
+      "loss": 0.822,
+      "step": 30970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7337025784993988e-06,
+      "loss": 0.8274,
+      "step": 30971
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7332227173290972e-06,
+      "loss": 0.6952,
+      "step": 30972
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.732742916275718e-06,
+      "loss": 0.7299,
+      "step": 30973
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.732263175342751e-06,
+      "loss": 0.6017,
+      "step": 30974
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7317834945336843e-06,
+      "loss": 0.6481,
+      "step": 30975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7313038738520061e-06,
+      "loss": 0.8897,
+      "step": 30976
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7308243133012037e-06,
+      "loss": 0.8046,
+      "step": 30977
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7303448128847656e-06,
+      "loss": 1.0893,
+      "step": 30978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7298653726061754e-06,
+      "loss": 0.79,
+      "step": 30979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.729385992468926e-06,
+      "loss": 0.7704,
+      "step": 30980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7289066724764992e-06,
+      "loss": 0.6283,
+      "step": 30981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7284274126323831e-06,
+      "loss": 0.8249,
+      "step": 30982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7279482129400626e-06,
+      "loss": 0.9173,
+      "step": 30983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7274690734030165e-06,
+      "loss": 0.5462,
+      "step": 30984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.726989994024736e-06,
+      "loss": 0.6029,
+      "step": 30985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.726510974808704e-06,
+      "loss": 0.7017,
+      "step": 30986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.726032015758402e-06,
+      "loss": 0.8034,
+      "step": 30987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7255531168773144e-06,
+      "loss": 0.4656,
+      "step": 30988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7250742781689234e-06,
+      "loss": 0.9545,
+      "step": 30989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.724595499636711e-06,
+      "loss": 0.7055,
+      "step": 30990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7241167812841564e-06,
+      "loss": 0.6004,
+      "step": 30991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7236381231147459e-06,
+      "loss": 0.7632,
+      "step": 30992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.723159525131961e-06,
+      "loss": 0.9146,
+      "step": 30993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.722680987339277e-06,
+      "loss": 0.9074,
+      "step": 30994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7222025097401752e-06,
+      "loss": 0.9187,
+      "step": 30995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7217240923381363e-06,
+      "loss": 1.0139,
+      "step": 30996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7212457351366373e-06,
+      "loss": 0.7446,
+      "step": 30997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7207674381391605e-06,
+      "loss": 0.7922,
+      "step": 30998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7202892013491823e-06,
+      "loss": 0.9556,
+      "step": 30999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7198110247701804e-06,
+      "loss": 0.8345,
+      "step": 31000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.719332908405632e-06,
+      "loss": 0.9636,
+      "step": 31001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7188548522590154e-06,
+      "loss": 0.8525,
+      "step": 31002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7183768563338043e-06,
+      "loss": 0.749,
+      "step": 31003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7178989206334773e-06,
+      "loss": 0.6705,
+      "step": 31004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7174210451615091e-06,
+      "loss": 0.8595,
+      "step": 31005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7169432299213739e-06,
+      "loss": 0.7701,
+      "step": 31006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7164654749165488e-06,
+      "loss": 0.6917,
+      "step": 31007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7159877801505054e-06,
+      "loss": 1.0087,
+      "step": 31008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7155101456267175e-06,
+      "loss": 0.9493,
+      "step": 31009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7150325713486616e-06,
+      "loss": 0.693,
+      "step": 31010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.714555057319809e-06,
+      "loss": 0.933,
+      "step": 31011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7140776035436312e-06,
+      "loss": 0.5416,
+      "step": 31012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7136002100236048e-06,
+      "loss": 1.0186,
+      "step": 31013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7131228767631946e-06,
+      "loss": 0.5679,
+      "step": 31014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7126456037658735e-06,
+      "loss": 0.9867,
+      "step": 31015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.712168391035115e-06,
+      "loss": 1.0352,
+      "step": 31016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.711691238574389e-06,
+      "loss": 0.9765,
+      "step": 31017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7112141463871657e-06,
+      "loss": 0.8042,
+      "step": 31018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7107371144769125e-06,
+      "loss": 0.6658,
+      "step": 31019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7102601428470988e-06,
+      "loss": 0.7653,
+      "step": 31020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7097832315011953e-06,
+      "loss": 0.9847,
+      "step": 31021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7093063804426657e-06,
+      "loss": 0.9665,
+      "step": 31022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7088295896749863e-06,
+      "loss": 0.9799,
+      "step": 31023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7083528592016152e-06,
+      "loss": 0.8865,
+      "step": 31024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7078761890260244e-06,
+      "loss": 0.6734,
+      "step": 31025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7073995791516774e-06,
+      "loss": 0.8463,
+      "step": 31026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7069230295820394e-06,
+      "loss": 0.8662,
+      "step": 31027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.706446540320581e-06,
+      "loss": 0.7855,
+      "step": 31028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7059701113707638e-06,
+      "loss": 0.9548,
+      "step": 31029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.705493742736053e-06,
+      "loss": 0.8888,
+      "step": 31030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7050174344199133e-06,
+      "loss": 0.8763,
+      "step": 31031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7045411864258078e-06,
+      "loss": 0.7667,
+      "step": 31032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.704064998757199e-06,
+      "loss": 0.7016,
+      "step": 31033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.703588871417552e-06,
+      "loss": 0.6581,
+      "step": 31034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7031128044103272e-06,
+      "loss": 0.6682,
+      "step": 31035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.702636797738988e-06,
+      "loss": 0.6914,
+      "step": 31036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7021608514069955e-06,
+      "loss": 0.9162,
+      "step": 31037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7016849654178102e-06,
+      "loss": 0.521,
+      "step": 31038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7012091397748942e-06,
+      "loss": 1.1,
+      "step": 31039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7007333744817034e-06,
+      "loss": 0.8454,
+      "step": 31040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.700257669541704e-06,
+      "loss": 0.5133,
+      "step": 31041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.699782024958353e-06,
+      "loss": 0.9619,
+      "step": 31042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.69930644073511e-06,
+      "loss": 0.8764,
+      "step": 31043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6988309168754313e-06,
+      "loss": 0.4989,
+      "step": 31044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.698355453382773e-06,
+      "loss": 0.756,
+      "step": 31045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6978800502605974e-06,
+      "loss": 0.3361,
+      "step": 31046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6974047075123612e-06,
+      "loss": 0.7486,
+      "step": 31047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6969294251415192e-06,
+      "loss": 0.8262,
+      "step": 31048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6964542031515285e-06,
+      "loss": 0.8874,
+      "step": 31049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6959790415458454e-06,
+      "loss": 0.8652,
+      "step": 31050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.695503940327925e-06,
+      "loss": 0.7364,
+      "step": 31051
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6950288995012188e-06,
+      "loss": 0.8429,
+      "step": 31052
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6945539190691895e-06,
+      "loss": 0.8205,
+      "step": 31053
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6940789990352846e-06,
+      "loss": 1.0395,
+      "step": 31054
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.69360413940296e-06,
+      "loss": 1.0555,
+      "step": 31055
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6931293401756677e-06,
+      "loss": 0.9462,
+      "step": 31056
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6926546013568612e-06,
+      "loss": 0.7678,
+      "step": 31057
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6921799229499914e-06,
+      "loss": 0.8365,
+      "step": 31058
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.691705304958513e-06,
+      "loss": 1.0907,
+      "step": 31059
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.691230747385878e-06,
+      "loss": 0.695,
+      "step": 31060
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6907562502355345e-06,
+      "loss": 0.8308,
+      "step": 31061
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6902818135109344e-06,
+      "loss": 0.9389,
+      "step": 31062
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.689807437215527e-06,
+      "loss": 0.8273,
+      "step": 31063
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.689333121352763e-06,
+      "loss": 0.5995,
+      "step": 31064
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6888588659260929e-06,
+      "loss": 0.9253,
+      "step": 31065
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6883846709389628e-06,
+      "loss": 1.0994,
+      "step": 31066
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6879105363948222e-06,
+      "loss": 0.9175,
+      "step": 31067
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6874364622971195e-06,
+      "loss": 0.8582,
+      "step": 31068
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6869624486493018e-06,
+      "loss": 0.9713,
+      "step": 31069
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6864884954548145e-06,
+      "loss": 0.8136,
+      "step": 31070
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6860146027171077e-06,
+      "loss": 0.8683,
+      "step": 31071
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6855407704396265e-06,
+      "loss": 0.6406,
+      "step": 31072
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6850669986258183e-06,
+      "loss": 0.7384,
+      "step": 31073
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6845932872791237e-06,
+      "loss": 0.8129,
+      "step": 31074
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.684119636402991e-06,
+      "loss": 0.7828,
+      "step": 31075
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6836460460008598e-06,
+      "loss": 0.8693,
+      "step": 31076
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6831725160761814e-06,
+      "loss": 0.6409,
+      "step": 31077
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6826990466323966e-06,
+      "loss": 0.6816,
+      "step": 31078
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6822256376729472e-06,
+      "loss": 0.7975,
+      "step": 31079
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6817522892012762e-06,
+      "loss": 0.8196,
+      "step": 31080
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6812790012208268e-06,
+      "loss": 0.7383,
+      "step": 31081
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6808057737350403e-06,
+      "loss": 0.4694,
+      "step": 31082
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.680332606747358e-06,
+      "loss": 0.8459,
+      "step": 31083
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6798595002612206e-06,
+      "loss": 0.9195,
+      "step": 31084
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6793864542800697e-06,
+      "loss": 0.7323,
+      "step": 31085
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6789134688073438e-06,
+      "loss": 0.6251,
+      "step": 31086
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6784405438464835e-06,
+      "loss": 1.0299,
+      "step": 31087
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6779676794009247e-06,
+      "loss": 1.0958,
+      "step": 31088
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6774948754741116e-06,
+      "loss": 0.677,
+      "step": 31089
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6770221320694802e-06,
+      "loss": 0.7375,
+      "step": 31090
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6765494491904688e-06,
+      "loss": 0.968,
+      "step": 31091
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6760768268405126e-06,
+      "loss": 0.8619,
+      "step": 31092
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6756042650230541e-06,
+      "loss": 0.9528,
+      "step": 31093
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.67513176374152e-06,
+      "loss": 1.0334,
+      "step": 31094
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6746593229993545e-06,
+      "loss": 0.4473,
+      "step": 31095
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6741869427999913e-06,
+      "loss": 0.9129,
+      "step": 31096
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6737146231468649e-06,
+      "loss": 0.778,
+      "step": 31097
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6732423640434103e-06,
+      "loss": 0.8549,
+      "step": 31098
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6727701654930617e-06,
+      "loss": 0.7406,
+      "step": 31099
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.672298027499254e-06,
+      "loss": 0.9729,
+      "step": 31100
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6718259500654165e-06,
+      "loss": 0.7815,
+      "step": 31101
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.671353933194989e-06,
+      "loss": 0.7084,
+      "step": 31102
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.670881976891402e-06,
+      "loss": 0.5599,
+      "step": 31103
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6704100811580836e-06,
+      "loss": 0.7978,
+      "step": 31104
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6699382459984691e-06,
+      "loss": 0.7008,
+      "step": 31105
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6694664714159858e-06,
+      "loss": 0.9417,
+      "step": 31106
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6689947574140697e-06,
+      "loss": 0.874,
+      "step": 31107
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6685231039961491e-06,
+      "loss": 0.9206,
+      "step": 31108
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6680515111656537e-06,
+      "loss": 0.8846,
+      "step": 31109
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6675799789260128e-06,
+      "loss": 0.7812,
+      "step": 31110
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6671085072806558e-06,
+      "loss": 0.8646,
+      "step": 31111
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6666370962330102e-06,
+      "loss": 0.9989,
+      "step": 31112
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6661657457865065e-06,
+      "loss": 0.5181,
+      "step": 31113
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6656944559445697e-06,
+      "loss": 0.7455,
+      "step": 31114
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6652232267106294e-06,
+      "loss": 0.884,
+      "step": 31115
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6647520580881106e-06,
+      "loss": 0.8132,
+      "step": 31116
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6642809500804402e-06,
+      "loss": 1.0263,
+      "step": 31117
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6638099026910448e-06,
+      "loss": 0.8423,
+      "step": 31118
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.663338915923347e-06,
+      "loss": 0.9075,
+      "step": 31119
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6628679897807764e-06,
+      "loss": 1.0171,
+      "step": 31120
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6623971242667568e-06,
+      "loss": 0.7555,
+      "step": 31121
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.66192631938471e-06,
+      "loss": 0.8799,
+      "step": 31122
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6614555751380634e-06,
+      "loss": 0.9238,
+      "step": 31123
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.660984891530234e-06,
+      "loss": 1.0737,
+      "step": 31124
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6605142685646503e-06,
+      "loss": 0.6697,
+      "step": 31125
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6600437062447327e-06,
+      "loss": 0.7403,
+      "step": 31126
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6595732045739033e-06,
+      "loss": 0.8904,
+      "step": 31127
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6591027635555845e-06,
+      "loss": 0.7695,
+      "step": 31128
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.658632383193196e-06,
+      "loss": 0.5542,
+      "step": 31129
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6581620634901595e-06,
+      "loss": 0.9253,
+      "step": 31130
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6576918044498913e-06,
+      "loss": 0.8459,
+      "step": 31131
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6572216060758184e-06,
+      "loss": 0.7276,
+      "step": 31132
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6567514683713582e-06,
+      "loss": 0.6513,
+      "step": 31133
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6562813913399256e-06,
+      "loss": 0.7662,
+      "step": 31134
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6558113749849414e-06,
+      "loss": 1.1044,
+      "step": 31135
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6553414193098238e-06,
+      "loss": 0.6836,
+      "step": 31136
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6548715243179868e-06,
+      "loss": 0.9917,
+      "step": 31137
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6544016900128535e-06,
+      "loss": 0.7425,
+      "step": 31138
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6539319163978384e-06,
+      "loss": 1.0158,
+      "step": 31139
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6534622034763558e-06,
+      "loss": 0.996,
+      "step": 31140
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652992551251824e-06,
+      "loss": 0.8881,
+      "step": 31141
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6525229597276583e-06,
+      "loss": 0.9331,
+      "step": 31142
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.652053428907271e-06,
+      "loss": 0.6634,
+      "step": 31143
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6515839587940786e-06,
+      "loss": 0.8575,
+      "step": 31144
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.651114549391495e-06,
+      "loss": 0.7298,
+      "step": 31145
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6506452007029327e-06,
+      "loss": 0.7049,
+      "step": 31146
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6501759127318061e-06,
+      "loss": 0.8495,
+      "step": 31147
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.649706685481528e-06,
+      "loss": 0.6769,
+      "step": 31148
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6492375189555078e-06,
+      "loss": 0.7604,
+      "step": 31149
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6487684131571613e-06,
+      "loss": 0.8396,
+      "step": 31150
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6482993680898984e-06,
+      "loss": 0.6222,
+      "step": 31151
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6478303837571307e-06,
+      "loss": 1.0716,
+      "step": 31152
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.64736146016227e-06,
+      "loss": 0.7859,
+      "step": 31153
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6468925973087213e-06,
+      "loss": 0.7976,
+      "step": 31154
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6464237951998952e-06,
+      "loss": 1.0065,
+      "step": 31155
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6459550538392054e-06,
+      "loss": 0.7889,
+      "step": 31156
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6454863732300586e-06,
+      "loss": 0.8679,
+      "step": 31157
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6450177533758616e-06,
+      "loss": 0.6643,
+      "step": 31158
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.644549194280024e-06,
+      "loss": 0.8817,
+      "step": 31159
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.644080695945952e-06,
+      "loss": 0.9444,
+      "step": 31160
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6436122583770542e-06,
+      "loss": 0.6244,
+      "step": 31161
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.643143881576732e-06,
+      "loss": 0.7103,
+      "step": 31162
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6426755655484017e-06,
+      "loss": 1.0312,
+      "step": 31163
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6422073102954595e-06,
+      "loss": 0.8683,
+      "step": 31164
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6417391158213147e-06,
+      "loss": 0.9077,
+      "step": 31165
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6412709821293704e-06,
+      "loss": 0.8431,
+      "step": 31166
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6408029092230294e-06,
+      "loss": 1.0476,
+      "step": 31167
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6403348971057009e-06,
+      "loss": 0.6058,
+      "step": 31168
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6398669457807848e-06,
+      "loss": 0.7217,
+      "step": 31169
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6393990552516848e-06,
+      "loss": 0.9587,
+      "step": 31170
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6389312255218038e-06,
+      "loss": 0.7174,
+      "step": 31171
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6384634565945422e-06,
+      "loss": 0.787,
+      "step": 31172
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6379957484733034e-06,
+      "loss": 0.7856,
+      "step": 31173
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6375281011614885e-06,
+      "loss": 1.0402,
+      "step": 31174
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6370605146624974e-06,
+      "loss": 0.7627,
+      "step": 31175
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6365929889797316e-06,
+      "loss": 0.7627,
+      "step": 31176
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6361255241165896e-06,
+      "loss": 0.6035,
+      "step": 31177
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.635658120076472e-06,
+      "loss": 0.6421,
+      "step": 31178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6351907768627772e-06,
+      "loss": 1.2838,
+      "step": 31179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6347234944789015e-06,
+      "loss": 0.8058,
+      "step": 31180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6342562729282485e-06,
+      "loss": 0.7464,
+      "step": 31181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6337891122142123e-06,
+      "loss": 0.7555,
+      "step": 31182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6333220123401927e-06,
+      "loss": 0.6805,
+      "step": 31183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6328549733095832e-06,
+      "loss": 1.1053,
+      "step": 31184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6323879951257783e-06,
+      "loss": 0.6203,
+      "step": 31185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.631921077792179e-06,
+      "loss": 0.938,
+      "step": 31186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6314542213121797e-06,
+      "loss": 0.8372,
+      "step": 31187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6309874256891745e-06,
+      "loss": 0.6323,
+      "step": 31188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.630520690926558e-06,
+      "loss": 0.7552,
+      "step": 31189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6300540170277247e-06,
+      "loss": 0.9292,
+      "step": 31190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.629587403996068e-06,
+      "loss": 0.9028,
+      "step": 31191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6291208518349788e-06,
+      "loss": 0.6384,
+      "step": 31192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6286543605478567e-06,
+      "loss": 0.8649,
+      "step": 31193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6281879301380875e-06,
+      "loss": 0.8577,
+      "step": 31194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6277215606090658e-06,
+      "loss": 1.0099,
+      "step": 31195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6272552519641827e-06,
+      "loss": 0.7987,
+      "step": 31196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.626789004206829e-06,
+      "loss": 0.8917,
+      "step": 31197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6263228173403921e-06,
+      "loss": 0.8996,
+      "step": 31198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6258566913682695e-06,
+      "loss": 0.7452,
+      "step": 31199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6253906262938457e-06,
+      "loss": 0.9187,
+      "step": 31200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6249246221205117e-06,
+      "loss": 0.6247,
+      "step": 31201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.624458678851656e-06,
+      "loss": 0.9809,
+      "step": 31202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6239927964906664e-06,
+      "loss": 0.7898,
+      "step": 31203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6235269750409321e-06,
+      "loss": 0.8115,
+      "step": 31204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6230612145058389e-06,
+      "loss": 0.752,
+      "step": 31205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6225955148887751e-06,
+      "loss": 0.637,
+      "step": 31206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.622129876193128e-06,
+      "loss": 0.8817,
+      "step": 31207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6216642984222819e-06,
+      "loss": 0.854,
+      "step": 31208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6211987815796238e-06,
+      "loss": 0.8546,
+      "step": 31209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6207333256685354e-06,
+      "loss": 0.9827,
+      "step": 31210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6202679306924074e-06,
+      "loss": 0.8822,
+      "step": 31211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6198025966546216e-06,
+      "loss": 0.9125,
+      "step": 31212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6193373235585652e-06,
+      "loss": 0.9912,
+      "step": 31213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6188721114076145e-06,
+      "loss": 0.9834,
+      "step": 31214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6184069602051578e-06,
+      "loss": 0.7146,
+      "step": 31215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6179418699545735e-06,
+      "loss": 0.7331,
+      "step": 31216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.617476840659249e-06,
+      "loss": 0.9961,
+      "step": 31217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6170118723225648e-06,
+      "loss": 0.613,
+      "step": 31218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6165469649479004e-06,
+      "loss": 0.5568,
+      "step": 31219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6160821185386365e-06,
+      "loss": 1.0316,
+      "step": 31220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6156173330981562e-06,
+      "loss": 0.9069,
+      "step": 31221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6151526086298364e-06,
+      "loss": 0.7901,
+      "step": 31222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.614687945137059e-06,
+      "loss": 1.0093,
+      "step": 31223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6142233426232012e-06,
+      "loss": 0.4136,
+      "step": 31224
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6137588010916427e-06,
+      "loss": 0.6195,
+      "step": 31225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6132943205457607e-06,
+      "loss": 0.7092,
+      "step": 31226
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6128299009889348e-06,
+      "loss": 0.7039,
+      "step": 31227
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6123655424245388e-06,
+      "loss": 0.8447,
+      "step": 31228
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6119012448559546e-06,
+      "loss": 0.5986,
+      "step": 31229
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.611437008286555e-06,
+      "loss": 0.7385,
+      "step": 31230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6109728327197172e-06,
+      "loss": 0.8265,
+      "step": 31231
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6105087181588176e-06,
+      "loss": 0.8799,
+      "step": 31232
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.610044664607232e-06,
+      "loss": 0.739,
+      "step": 31233
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6095806720683282e-06,
+      "loss": 1.0809,
+      "step": 31234
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6091167405454877e-06,
+      "loss": 0.9392,
+      "step": 31235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6086528700420822e-06,
+      "loss": 0.7989,
+      "step": 31236
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6081890605614859e-06,
+      "loss": 0.6114,
+      "step": 31237
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.607725312107069e-06,
+      "loss": 0.7187,
+      "step": 31238
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.607261624682207e-06,
+      "loss": 0.7321,
+      "step": 31239
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6067979982902703e-06,
+      "loss": 0.8766,
+      "step": 31240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6063344329346275e-06,
+      "loss": 0.6623,
+      "step": 31241
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6058709286186559e-06,
+      "loss": 0.6633,
+      "step": 31242
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6054074853457246e-06,
+      "loss": 1.0401,
+      "step": 31243
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6049441031192003e-06,
+      "loss": 0.7207,
+      "step": 31244
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6044807819424545e-06,
+      "loss": 0.8088,
+      "step": 31245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6040175218188546e-06,
+      "loss": 0.8176,
+      "step": 31246
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6035543227517736e-06,
+      "loss": 0.8333,
+      "step": 31247
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6030911847445785e-06,
+      "loss": 0.9904,
+      "step": 31248
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6026281078006366e-06,
+      "loss": 0.8412,
+      "step": 31249
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6021650919233157e-06,
+      "loss": 0.7226,
+      "step": 31250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6017021371159835e-06,
+      "loss": 0.7457,
+      "step": 31251
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6012392433820046e-06,
+      "loss": 0.8212,
+      "step": 31252
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.600776410724747e-06,
+      "loss": 0.8065,
+      "step": 31253
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.600313639147577e-06,
+      "loss": 0.7207,
+      "step": 31254
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5998509286538578e-06,
+      "loss": 0.5951,
+      "step": 31255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.599388279246956e-06,
+      "loss": 0.8984,
+      "step": 31256
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5989256909302354e-06,
+      "loss": 0.7466,
+      "step": 31257
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5984631637070602e-06,
+      "loss": 0.9866,
+      "step": 31258
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5980006975807915e-06,
+      "loss": 0.8406,
+      "step": 31259
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5975382925547966e-06,
+      "loss": 0.8049,
+      "step": 31260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5970759486324373e-06,
+      "loss": 0.686,
+      "step": 31261
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5966136658170738e-06,
+      "loss": 0.8268,
+      "step": 31262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5961514441120719e-06,
+      "loss": 0.8269,
+      "step": 31263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5956892835207849e-06,
+      "loss": 0.8706,
+      "step": 31264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5952271840465806e-06,
+      "loss": 0.516,
+      "step": 31265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5947651456928182e-06,
+      "loss": 0.7366,
+      "step": 31266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5943031684628573e-06,
+      "loss": 0.8374,
+      "step": 31267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5938412523600566e-06,
+      "loss": 0.7071,
+      "step": 31268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5933793973877765e-06,
+      "loss": 0.5379,
+      "step": 31269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5929176035493744e-06,
+      "loss": 0.8738,
+      "step": 31270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5924558708482063e-06,
+      "loss": 0.8795,
+      "step": 31271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5919941992876352e-06,
+      "loss": 0.6283,
+      "step": 31272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5915325888710198e-06,
+      "loss": 0.9756,
+      "step": 31273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5910710396017093e-06,
+      "loss": 0.9429,
+      "step": 31274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5906095514830645e-06,
+      "loss": 0.6546,
+      "step": 31275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5901481245184402e-06,
+      "loss": 0.8584,
+      "step": 31276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.589686758711191e-06,
+      "loss": 0.5073,
+      "step": 31277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5892254540646756e-06,
+      "loss": 0.8088,
+      "step": 31278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5887642105822476e-06,
+      "loss": 0.773,
+      "step": 31279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5883030282672595e-06,
+      "loss": 0.6637,
+      "step": 31280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5878419071230656e-06,
+      "loss": 0.8051,
+      "step": 31281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5873808471530204e-06,
+      "loss": 0.8791,
+      "step": 31282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5869198483604752e-06,
+      "loss": 0.9066,
+      "step": 31283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5864589107487837e-06,
+      "loss": 0.7878,
+      "step": 31284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5859980343212965e-06,
+      "loss": 0.9368,
+      "step": 31285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5855372190813667e-06,
+      "loss": 0.7214,
+      "step": 31286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5850764650323437e-06,
+      "loss": 1.0338,
+      "step": 31287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5846157721775801e-06,
+      "loss": 0.811,
+      "step": 31288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5841551405204225e-06,
+      "loss": 0.9247,
+      "step": 31289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5836945700642248e-06,
+      "loss": 0.7739,
+      "step": 31290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5832340608123364e-06,
+      "loss": 1.0761,
+      "step": 31291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5827736127681038e-06,
+      "loss": 0.7474,
+      "step": 31292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5823132259348784e-06,
+      "loss": 1.1158,
+      "step": 31293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5818529003160033e-06,
+      "loss": 0.8676,
+      "step": 31294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5813926359148268e-06,
+      "loss": 0.9806,
+      "step": 31295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5809324327346997e-06,
+      "loss": 0.6297,
+      "step": 31296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.580472290778967e-06,
+      "loss": 0.6587,
+      "step": 31297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5800122100509752e-06,
+      "loss": 0.7895,
+      "step": 31298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.57955219055407e-06,
+      "loss": 0.8619,
+      "step": 31299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5790922322915958e-06,
+      "loss": 0.4514,
+      "step": 31300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5786323352668987e-06,
+      "loss": 0.9294,
+      "step": 31301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5781724994833192e-06,
+      "loss": 0.7389,
+      "step": 31302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5777127249442092e-06,
+      "loss": 1.0289,
+      "step": 31303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.577253011652905e-06,
+      "loss": 0.8962,
+      "step": 31304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5767933596127528e-06,
+      "loss": 0.9967,
+      "step": 31305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5763337688270952e-06,
+      "loss": 0.8763,
+      "step": 31306
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.575874239299271e-06,
+      "loss": 0.7452,
+      "step": 31307
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5754147710326261e-06,
+      "loss": 0.6218,
+      "step": 31308
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5749553640305015e-06,
+      "loss": 0.6503,
+      "step": 31309
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5744960182962376e-06,
+      "loss": 0.9239,
+      "step": 31310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.574036733833173e-06,
+      "loss": 0.8762,
+      "step": 31311
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5735775106446505e-06,
+      "loss": 0.7181,
+      "step": 31312
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5731183487340062e-06,
+      "loss": 0.6672,
+      "step": 31313
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.572659248104582e-06,
+      "loss": 0.8103,
+      "step": 31314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5722002087597166e-06,
+      "loss": 0.7306,
+      "step": 31315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5717412307027457e-06,
+      "loss": 0.9535,
+      "step": 31316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5712823139370092e-06,
+      "loss": 0.8846,
+      "step": 31317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5708234584658434e-06,
+      "loss": 1.0218,
+      "step": 31318
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5703646642925853e-06,
+      "loss": 0.89,
+      "step": 31319
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.56990593142057e-06,
+      "loss": 0.9756,
+      "step": 31320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5694472598531362e-06,
+      "loss": 0.7363,
+      "step": 31321
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.568988649593619e-06,
+      "loss": 0.832,
+      "step": 31322
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5685301006453546e-06,
+      "loss": 0.8322,
+      "step": 31323
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5680716130116724e-06,
+      "loss": 0.6943,
+      "step": 31324
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5676131866959087e-06,
+      "loss": 0.6746,
+      "step": 31325
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5671548217014e-06,
+      "loss": 0.8454,
+      "step": 31326
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5666965180314775e-06,
+      "loss": 1.0832,
+      "step": 31327
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5662382756894757e-06,
+      "loss": 0.8117,
+      "step": 31328
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5657800946787238e-06,
+      "loss": 0.6326,
+      "step": 31329
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5653219750025573e-06,
+      "loss": 0.951,
+      "step": 31330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5648639166643053e-06,
+      "loss": 0.8407,
+      "step": 31331
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5644059196673e-06,
+      "loss": 0.5801,
+      "step": 31332
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.563947984014871e-06,
+      "loss": 0.9558,
+      "step": 31333
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5634901097103495e-06,
+      "loss": 0.8852,
+      "step": 31334
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5630322967570655e-06,
+      "loss": 0.9313,
+      "step": 31335
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5625745451583462e-06,
+      "loss": 0.6964,
+      "step": 31336
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5621168549175226e-06,
+      "loss": 0.7016,
+      "step": 31337
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5616592260379193e-06,
+      "loss": 0.4827,
+      "step": 31338
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5612016585228695e-06,
+      "loss": 0.6536,
+      "step": 31339
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5607441523756994e-06,
+      "loss": 0.7298,
+      "step": 31340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5602867075997341e-06,
+      "loss": 1.0018,
+      "step": 31341
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.559829324198301e-06,
+      "loss": 0.7813,
+      "step": 31342
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5593720021747295e-06,
+      "loss": 0.7653,
+      "step": 31343
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5589147415323359e-06,
+      "loss": 0.7979,
+      "step": 31344
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5584575422744553e-06,
+      "loss": 0.7437,
+      "step": 31345
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5580004044044084e-06,
+      "loss": 0.714,
+      "step": 31346
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5575433279255204e-06,
+      "loss": 0.7405,
+      "step": 31347
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.557086312841114e-06,
+      "loss": 0.6678,
+      "step": 31348
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5566293591545144e-06,
+      "loss": 1.057,
+      "step": 31349
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5561724668690436e-06,
+      "loss": 0.7656,
+      "step": 31350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.555715635988021e-06,
+      "loss": 0.9369,
+      "step": 31351
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.555258866514775e-06,
+      "loss": 0.6701,
+      "step": 31352
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5548021584526274e-06,
+      "loss": 0.7896,
+      "step": 31353
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5543455118048934e-06,
+      "loss": 0.751,
+      "step": 31354
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5538889265748958e-06,
+      "loss": 0.6699,
+      "step": 31355
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5534324027659541e-06,
+      "loss": 0.8441,
+      "step": 31356
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5529759403813926e-06,
+      "loss": 0.7956,
+      "step": 31357
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5525195394245285e-06,
+      "loss": 0.9064,
+      "step": 31358
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.552063199898679e-06,
+      "loss": 0.5292,
+      "step": 31359
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.551606921807165e-06,
+      "loss": 0.5363,
+      "step": 31360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5511507051533037e-06,
+      "loss": 0.7797,
+      "step": 31361
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5506945499404125e-06,
+      "loss": 0.7035,
+      "step": 31362
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5502384561718087e-06,
+      "loss": 0.7131,
+      "step": 31363
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5497824238508097e-06,
+      "loss": 0.7899,
+      "step": 31364
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5493264529807305e-06,
+      "loss": 1.01,
+      "step": 31365
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5488705435648876e-06,
+      "loss": 0.673,
+      "step": 31366
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.548414695606597e-06,
+      "loss": 1.0572,
+      "step": 31367
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.547958909109174e-06,
+      "loss": 0.9736,
+      "step": 31368
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5475031840759292e-06,
+      "loss": 0.9211,
+      "step": 31369
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5470475205101821e-06,
+      "loss": 0.6461,
+      "step": 31370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5465919184152444e-06,
+      "loss": 0.736,
+      "step": 31371
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5461363777944293e-06,
+      "loss": 0.9436,
+      "step": 31372
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5456808986510507e-06,
+      "loss": 0.6149,
+      "step": 31373
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.545225480988416e-06,
+      "loss": 0.9395,
+      "step": 31374
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5447701248098423e-06,
+      "loss": 1.1285,
+      "step": 31375
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5443148301186384e-06,
+      "loss": 0.8445,
+      "step": 31376
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.543859596918117e-06,
+      "loss": 0.9759,
+      "step": 31377
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5434044252115875e-06,
+      "loss": 0.9231,
+      "step": 31378
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.54294931500236e-06,
+      "loss": 0.7079,
+      "step": 31379
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5424942662937436e-06,
+      "loss": 0.5109,
+      "step": 31380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.542039279089047e-06,
+      "loss": 0.6215,
+      "step": 31381
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5415843533915809e-06,
+      "loss": 0.5913,
+      "step": 31382
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5411294892046557e-06,
+      "loss": 0.9391,
+      "step": 31383
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5406746865315748e-06,
+      "loss": 0.8676,
+      "step": 31384
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.540219945375645e-06,
+      "loss": 0.8784,
+      "step": 31385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5397652657401763e-06,
+      "loss": 0.7454,
+      "step": 31386
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5393106476284724e-06,
+      "loss": 1.0829,
+      "step": 31387
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.538856091043842e-06,
+      "loss": 0.6267,
+      "step": 31388
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5384015959895904e-06,
+      "loss": 0.4337,
+      "step": 31389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5379471624690224e-06,
+      "loss": 0.6941,
+      "step": 31390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.537492790485442e-06,
+      "loss": 0.7426,
+      "step": 31391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5370384800421533e-06,
+      "loss": 1.0051,
+      "step": 31392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5365842311424607e-06,
+      "loss": 0.682,
+      "step": 31393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5361300437896677e-06,
+      "loss": 0.7854,
+      "step": 31394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5356759179870762e-06,
+      "loss": 0.9037,
+      "step": 31395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5352218537379903e-06,
+      "loss": 0.7959,
+      "step": 31396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5347678510457108e-06,
+      "loss": 0.6358,
+      "step": 31397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5343139099135397e-06,
+      "loss": 0.5715,
+      "step": 31398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.533860030344775e-06,
+      "loss": 0.6882,
+      "step": 31399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.533406212342723e-06,
+      "loss": 1.0193,
+      "step": 31400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.532952455910681e-06,
+      "loss": 0.8451,
+      "step": 31401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5324987610519494e-06,
+      "loss": 0.7325,
+      "step": 31402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5320451277698301e-06,
+      "loss": 0.7796,
+      "step": 31403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5315915560676154e-06,
+      "loss": 0.7917,
+      "step": 31404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5311380459486048e-06,
+      "loss": 0.7757,
+      "step": 31405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5306845974161021e-06,
+      "loss": 0.9985,
+      "step": 31406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5302312104734007e-06,
+      "loss": 1.0018,
+      "step": 31407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5297778851237988e-06,
+      "loss": 0.863,
+      "step": 31408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5293246213705936e-06,
+      "loss": 0.826,
+      "step": 31409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5288714192170796e-06,
+      "loss": 0.6969,
+      "step": 31410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5284182786665525e-06,
+      "loss": 0.716,
+      "step": 31411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5279651997223077e-06,
+      "loss": 0.8982,
+      "step": 31412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5275121823876437e-06,
+      "loss": 0.5173,
+      "step": 31413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5270592266658512e-06,
+      "loss": 0.7819,
+      "step": 31414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5266063325602242e-06,
+      "loss": 0.8679,
+      "step": 31415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.526153500074057e-06,
+      "loss": 0.8102,
+      "step": 31416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5257007292106396e-06,
+      "loss": 0.6719,
+      "step": 31417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5252480199732712e-06,
+      "loss": 0.6584,
+      "step": 31418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5247953723652397e-06,
+      "loss": 0.7674,
+      "step": 31419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5243427863898363e-06,
+      "loss": 0.5015,
+      "step": 31420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5238902620503548e-06,
+      "loss": 1.191,
+      "step": 31421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5234377993500837e-06,
+      "loss": 0.7846,
+      "step": 31422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.522985398292315e-06,
+      "loss": 0.9286,
+      "step": 31423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5225330588803377e-06,
+      "loss": 0.6792,
+      "step": 31424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.522080781117441e-06,
+      "loss": 0.8888,
+      "step": 31425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.521628565006914e-06,
+      "loss": 0.9365,
+      "step": 31426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5211764105520466e-06,
+      "loss": 0.913,
+      "step": 31427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5207243177561248e-06,
+      "loss": 0.8919,
+      "step": 31428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5202722866224385e-06,
+      "loss": 0.6415,
+      "step": 31429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5198203171542703e-06,
+      "loss": 0.8037,
+      "step": 31430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5193684093549132e-06,
+      "loss": 0.6564,
+      "step": 31431
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5189165632276516e-06,
+      "loss": 0.6762,
+      "step": 31432
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.518464778775771e-06,
+      "loss": 1.0862,
+      "step": 31433
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5180130560025564e-06,
+      "loss": 0.6251,
+      "step": 31434
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5175613949112889e-06,
+      "loss": 0.8325,
+      "step": 31435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5171097955052593e-06,
+      "loss": 1.0253,
+      "step": 31436
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5166582577877499e-06,
+      "loss": 0.9538,
+      "step": 31437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5162067817620428e-06,
+      "loss": 0.8195,
+      "step": 31438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5157553674314228e-06,
+      "loss": 1.1954,
+      "step": 31439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5153040147991716e-06,
+      "loss": 0.8407,
+      "step": 31440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5148527238685717e-06,
+      "loss": 1.1075,
+      "step": 31441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.514401494642903e-06,
+      "loss": 0.7739,
+      "step": 31442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.513950327125454e-06,
+      "loss": 0.699,
+      "step": 31443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5134992213194977e-06,
+      "loss": 0.9245,
+      "step": 31444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5130481772283179e-06,
+      "loss": 0.7076,
+      "step": 31445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5125971948551932e-06,
+      "loss": 0.9266,
+      "step": 31446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5121462742034055e-06,
+      "loss": 0.7012,
+      "step": 31447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5116954152762298e-06,
+      "loss": 0.7468,
+      "step": 31448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.511244618076949e-06,
+      "loss": 0.8709,
+      "step": 31449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5107938826088408e-06,
+      "loss": 0.5914,
+      "step": 31450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5103432088751824e-06,
+      "loss": 0.9157,
+      "step": 31451
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.509892596879251e-06,
+      "loss": 0.7544,
+      "step": 31452
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5094420466243232e-06,
+      "loss": 0.7884,
+      "step": 31453
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.508991558113676e-06,
+      "loss": 0.6957,
+      "step": 31454
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5085411313505849e-06,
+      "loss": 0.6897,
+      "step": 31455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.508090766338326e-06,
+      "loss": 1.0641,
+      "step": 31456
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5076404630801733e-06,
+      "loss": 0.9052,
+      "step": 31457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.507190221579402e-06,
+      "loss": 0.7694,
+      "step": 31458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5067400418392875e-06,
+      "loss": 0.6465,
+      "step": 31459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5062899238631002e-06,
+      "loss": 1.0426,
+      "step": 31460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5058398676541175e-06,
+      "loss": 0.7866,
+      "step": 31461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5053898732156104e-06,
+      "loss": 0.8264,
+      "step": 31462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5049399405508535e-06,
+      "loss": 0.8523,
+      "step": 31463
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5044900696631159e-06,
+      "loss": 1.0182,
+      "step": 31464
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5040402605556691e-06,
+      "loss": 0.9534,
+      "step": 31465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5035905132317818e-06,
+      "loss": 0.7681,
+      "step": 31466
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5031408276947312e-06,
+      "loss": 0.8524,
+      "step": 31467
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5026912039477837e-06,
+      "loss": 0.7261,
+      "step": 31468
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.50224164199421e-06,
+      "loss": 0.9255,
+      "step": 31469
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5017921418372772e-06,
+      "loss": 1.0248,
+      "step": 31470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5013427034802564e-06,
+      "loss": 0.596,
+      "step": 31471
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5008933269264159e-06,
+      "loss": 0.4777,
+      "step": 31472
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.500444012179022e-06,
+      "loss": 1.0255,
+      "step": 31473
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4999947592413444e-06,
+      "loss": 1.033,
+      "step": 31474
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4995455681166472e-06,
+      "loss": 0.6929,
+      "step": 31475
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4990964388081996e-06,
+      "loss": 0.9055,
+      "step": 31476
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4986473713192661e-06,
+      "loss": 0.9595,
+      "step": 31477
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4981983656531107e-06,
+      "loss": 0.8462,
+      "step": 31478
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.497749421813004e-06,
+      "loss": 0.6754,
+      "step": 31479
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4973005398022067e-06,
+      "loss": 0.8938,
+      "step": 31480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4968517196239851e-06,
+      "loss": 1.0039,
+      "step": 31481
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4964029612816012e-06,
+      "loss": 0.8721,
+      "step": 31482
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4959542647783221e-06,
+      "loss": 0.839,
+      "step": 31483
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4955056301174032e-06,
+      "loss": 1.0269,
+      "step": 31484
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4950570573021138e-06,
+      "loss": 0.5345,
+      "step": 31485
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.494608546335714e-06,
+      "loss": 0.6889,
+      "step": 31486
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4941600972214655e-06,
+      "loss": 0.583,
+      "step": 31487
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4937117099626286e-06,
+      "loss": 0.8165,
+      "step": 31488
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4932633845624656e-06,
+      "loss": 0.6715,
+      "step": 31489
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.492815121024236e-06,
+      "loss": 1.0264,
+      "step": 31490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4923669193511969e-06,
+      "loss": 0.7782,
+      "step": 31491
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4919187795466127e-06,
+      "loss": 0.8429,
+      "step": 31492
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.491470701613742e-06,
+      "loss": 0.8103,
+      "step": 31493
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4910226855558385e-06,
+      "loss": 0.7972,
+      "step": 31494
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4905747313761643e-06,
+      "loss": 0.9591,
+      "step": 31495
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4901268390779721e-06,
+      "loss": 0.5767,
+      "step": 31496
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4896790086645253e-06,
+      "loss": 0.7475,
+      "step": 31497
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4892312401390774e-06,
+      "loss": 0.8555,
+      "step": 31498
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4887835335048862e-06,
+      "loss": 0.8848,
+      "step": 31499
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4883358887652044e-06,
+      "loss": 0.8013,
+      "step": 31500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4878883059232907e-06,
+      "loss": 0.684,
+      "step": 31501
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.487440784982398e-06,
+      "loss": 0.8165,
+      "step": 31502
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4869933259457815e-06,
+      "loss": 0.7664,
+      "step": 31503
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.486545928816695e-06,
+      "loss": 0.798,
+      "step": 31504
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4860985935983908e-06,
+      "loss": 1.0216,
+      "step": 31505
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4856513202941236e-06,
+      "loss": 0.8291,
+      "step": 31506
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4852041089071457e-06,
+      "loss": 0.7814,
+      "step": 31507
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4847569594407095e-06,
+      "loss": 0.8039,
+      "step": 31508
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4843098718980631e-06,
+      "loss": 0.8871,
+      "step": 31509
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4838628462824645e-06,
+      "loss": 0.7669,
+      "step": 31510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.48341588259716e-06,
+      "loss": 0.6845,
+      "step": 31511
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4829689808454007e-06,
+      "loss": 0.7701,
+      "step": 31512
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4825221410304402e-06,
+      "loss": 0.8565,
+      "step": 31513
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4820753631555185e-06,
+      "loss": 0.8985,
+      "step": 31514
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4816286472238939e-06,
+      "loss": 0.7103,
+      "step": 31515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4811819932388115e-06,
+      "loss": 1.0486,
+      "step": 31516
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4807354012035203e-06,
+      "loss": 0.9529,
+      "step": 31517
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.480288871121266e-06,
+      "loss": 0.5593,
+      "step": 31518
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4798424029952985e-06,
+      "loss": 0.8949,
+      "step": 31519
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4793959968288618e-06,
+      "loss": 0.7858,
+      "step": 31520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4789496526252023e-06,
+      "loss": 0.9518,
+      "step": 31521
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4785033703875685e-06,
+      "loss": 0.7143,
+      "step": 31522
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4780571501192077e-06,
+      "loss": 0.8519,
+      "step": 31523
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4776109918233594e-06,
+      "loss": 0.7323,
+      "step": 31524
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4771648955032702e-06,
+      "loss": 0.9265,
+      "step": 31525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.476718861162184e-06,
+      "loss": 0.7038,
+      "step": 31526
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4762728888033428e-06,
+      "loss": 0.7173,
+      "step": 31527
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4758269784299928e-06,
+      "loss": 0.7037,
+      "step": 31528
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4753811300453769e-06,
+      "loss": 0.7469,
+      "step": 31529
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.474935343652736e-06,
+      "loss": 0.7553,
+      "step": 31530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4744896192553116e-06,
+      "loss": 0.7182,
+      "step": 31531
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4740439568563458e-06,
+      "loss": 0.6148,
+      "step": 31532
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4735983564590784e-06,
+      "loss": 1.0234,
+      "step": 31533
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.473152818066751e-06,
+      "loss": 0.8542,
+      "step": 31534
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4727073416826033e-06,
+      "loss": 1.0501,
+      "step": 31535
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4722619273098749e-06,
+      "loss": 0.4902,
+      "step": 31536
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4718165749518044e-06,
+      "loss": 0.951,
+      "step": 31537
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4713712846116313e-06,
+      "loss": 0.8138,
+      "step": 31538
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4709260562925898e-06,
+      "loss": 0.7939,
+      "step": 31539
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.470480889997924e-06,
+      "loss": 0.9028,
+      "step": 31540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.470035785730869e-06,
+      "loss": 0.6402,
+      "step": 31541
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4695907434946633e-06,
+      "loss": 0.8145,
+      "step": 31542
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4691457632925376e-06,
+      "loss": 0.678,
+      "step": 31543
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4687008451277317e-06,
+      "loss": 0.6076,
+      "step": 31544
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4682559890034787e-06,
+      "loss": 0.8784,
+      "step": 31545
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4678111949230178e-06,
+      "loss": 0.8131,
+      "step": 31546
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4673664628895811e-06,
+      "loss": 0.7715,
+      "step": 31547
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.466921792906404e-06,
+      "loss": 0.8965,
+      "step": 31548
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4664771849767178e-06,
+      "loss": 0.906,
+      "step": 31549
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4660326391037583e-06,
+      "loss": 0.6464,
+      "step": 31550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4655881552907569e-06,
+      "loss": 0.9466,
+      "step": 31551
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4651437335409458e-06,
+      "loss": 0.991,
+      "step": 31552
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4646993738575566e-06,
+      "loss": 0.8842,
+      "step": 31553
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4642550762438224e-06,
+      "loss": 0.8293,
+      "step": 31554
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4638108407029728e-06,
+      "loss": 0.9024,
+      "step": 31555
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4633666672382385e-06,
+      "loss": 0.7385,
+      "step": 31556
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4629225558528482e-06,
+      "loss": 0.9829,
+      "step": 31557
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4624785065500348e-06,
+      "loss": 0.8128,
+      "step": 31558
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4620345193330255e-06,
+      "loss": 0.8705,
+      "step": 31559
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.461590594205049e-06,
+      "loss": 0.5938,
+      "step": 31560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.461146731169335e-06,
+      "loss": 0.6343,
+      "step": 31561
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4607029302291121e-06,
+      "loss": 0.671,
+      "step": 31562
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4602591913876007e-06,
+      "loss": 0.6678,
+      "step": 31563
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4598155146480352e-06,
+      "loss": 0.9659,
+      "step": 31564
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.459371900013641e-06,
+      "loss": 0.739,
+      "step": 31565
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4589283474876404e-06,
+      "loss": 0.8731,
+      "step": 31566
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4584848570732634e-06,
+      "loss": 1.022,
+      "step": 31567
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4580414287737322e-06,
+      "loss": 0.7619,
+      "step": 31568
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4575980625922715e-06,
+      "loss": 0.8389,
+      "step": 31569
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4571547585321055e-06,
+      "loss": 0.818,
+      "step": 31570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4567115165964597e-06,
+      "loss": 0.5469,
+      "step": 31571
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.45626833678856e-06,
+      "loss": 0.73,
+      "step": 31572
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4558252191116229e-06,
+      "loss": 0.7829,
+      "step": 31573
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4553821635688725e-06,
+      "loss": 1.0708,
+      "step": 31574
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4549391701635308e-06,
+      "loss": 0.9288,
+      "step": 31575
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4544962388988226e-06,
+      "loss": 0.9396,
+      "step": 31576
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4540533697779669e-06,
+      "loss": 0.7705,
+      "step": 31577
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.453610562804183e-06,
+      "loss": 1.0508,
+      "step": 31578
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.453167817980693e-06,
+      "loss": 0.7025,
+      "step": 31579
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4527251353107163e-06,
+      "loss": 0.947,
+      "step": 31580
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4522825147974707e-06,
+      "loss": 0.6757,
+      "step": 31581
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4518399564441755e-06,
+      "loss": 0.9706,
+      "step": 31582
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4513974602540505e-06,
+      "loss": 0.5967,
+      "step": 31583
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4509550262303117e-06,
+      "loss": 0.8298,
+      "step": 31584
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4505126543761772e-06,
+      "loss": 0.8713,
+      "step": 31585
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.450070344694865e-06,
+      "loss": 0.6649,
+      "step": 31586
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4496280971895893e-06,
+      "loss": 0.5949,
+      "step": 31587
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4491859118635665e-06,
+      "loss": 0.9947,
+      "step": 31588
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448743788720015e-06,
+      "loss": 0.9026,
+      "step": 31589
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4483017277621482e-06,
+      "loss": 0.9236,
+      "step": 31590
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4478597289931806e-06,
+      "loss": 1.0137,
+      "step": 31591
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44741779241633e-06,
+      "loss": 0.4576,
+      "step": 31592
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446975918034802e-06,
+      "loss": 0.8092,
+      "step": 31593
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4465341058518168e-06,
+      "loss": 1.0122,
+      "step": 31594
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4460923558705863e-06,
+      "loss": 0.7058,
+      "step": 31595
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4456506680943217e-06,
+      "loss": 0.6912,
+      "step": 31596
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452090425262344e-06,
+      "loss": 0.9485,
+      "step": 31597
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4447674791695377e-06,
+      "loss": 0.6771,
+      "step": 31598
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444325978027442e-06,
+      "loss": 0.6088,
+      "step": 31599
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4438845391031553e-06,
+      "loss": 0.7512,
+      "step": 31600
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4434431623998935e-06,
+      "loss": 0.8765,
+      "step": 31601
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.443001847920864e-06,
+      "loss": 0.9431,
+      "step": 31602
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4425605956692745e-06,
+      "loss": 0.4873,
+      "step": 31603
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4421194056483333e-06,
+      "loss": 0.599,
+      "step": 31604
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4416782778612514e-06,
+      "loss": 0.8437,
+      "step": 31605
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4412372123112329e-06,
+      "loss": 0.8729,
+      "step": 31606
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4407962090014892e-06,
+      "loss": 0.8381,
+      "step": 31607
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4403552679352263e-06,
+      "loss": 0.5845,
+      "step": 31608
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4399143891156509e-06,
+      "loss": 0.8068,
+      "step": 31609
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4394735725459686e-06,
+      "loss": 0.7836,
+      "step": 31610
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4390328182293845e-06,
+      "loss": 0.6959,
+      "step": 31611
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.438592126169105e-06,
+      "loss": 0.6663,
+      "step": 31612
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.438151496368334e-06,
+      "loss": 0.7111,
+      "step": 31613
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.437710928830276e-06,
+      "loss": 0.879,
+      "step": 31614
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4372704235581358e-06,
+      "loss": 0.8512,
+      "step": 31615
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4368299805551145e-06,
+      "loss": 0.7365,
+      "step": 31616
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4363895998244172e-06,
+      "loss": 0.9228,
+      "step": 31617
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4359492813692433e-06,
+      "loss": 0.617,
+      "step": 31618
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4355090251928005e-06,
+      "loss": 0.7179,
+      "step": 31619
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4350688312982864e-06,
+      "loss": 1.0014,
+      "step": 31620
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4346286996889025e-06,
+      "loss": 0.822,
+      "step": 31621
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.434188630367852e-06,
+      "loss": 0.7663,
+      "step": 31622
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4337486233383312e-06,
+      "loss": 0.7305,
+      "step": 31623
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.43330867860354e-06,
+      "loss": 0.9568,
+      "step": 31624
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.432868796166681e-06,
+      "loss": 1.0277,
+      "step": 31625
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4324289760309507e-06,
+      "loss": 0.8829,
+      "step": 31626
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4319892181995498e-06,
+      "loss": 0.9474,
+      "step": 31627
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4315495226756738e-06,
+      "loss": 0.4792,
+      "step": 31628
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4311098894625208e-06,
+      "loss": 1.0554,
+      "step": 31629
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4306703185632899e-06,
+      "loss": 0.9637,
+      "step": 31630
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4302308099811724e-06,
+      "loss": 0.7844,
+      "step": 31631
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4297913637193728e-06,
+      "loss": 0.8955,
+      "step": 31632
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4293519797810796e-06,
+      "loss": 0.878,
+      "step": 31633
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4289126581694912e-06,
+      "loss": 0.9061,
+      "step": 31634
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.428473398887802e-06,
+      "loss": 0.8239,
+      "step": 31635
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4280342019392024e-06,
+      "loss": 0.8169,
+      "step": 31636
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4275950673268934e-06,
+      "loss": 0.9701,
+      "step": 31637
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4271559950540648e-06,
+      "loss": 0.8374,
+      "step": 31638
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4267169851239105e-06,
+      "loss": 1.0522,
+      "step": 31639
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4262780375396212e-06,
+      "loss": 0.8402,
+      "step": 31640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.42583915230439e-06,
+      "loss": 0.5374,
+      "step": 31641
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4254003294214092e-06,
+      "loss": 0.7352,
+      "step": 31642
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.424961568893869e-06,
+      "loss": 1.0473,
+      "step": 31643
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4245228707249613e-06,
+      "loss": 0.7233,
+      "step": 31644
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4240842349178752e-06,
+      "loss": 0.6947,
+      "step": 31645
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.423645661475801e-06,
+      "loss": 0.6091,
+      "step": 31646
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4232071504019274e-06,
+      "loss": 0.7846,
+      "step": 31647
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.422768701699444e-06,
+      "loss": 0.9575,
+      "step": 31648
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4223303153715373e-06,
+      "loss": 0.9155,
+      "step": 31649
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.421891991421399e-06,
+      "loss": 0.7519,
+      "step": 31650
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4214537298522147e-06,
+      "loss": 0.903,
+      "step": 31651
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4210155306671736e-06,
+      "loss": 0.9616,
+      "step": 31652
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4205773938694578e-06,
+      "loss": 0.7449,
+      "step": 31653
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4201393194622536e-06,
+      "loss": 0.597,
+      "step": 31654
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.419701307448752e-06,
+      "loss": 0.9338,
+      "step": 31655
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4192633578321346e-06,
+      "loss": 0.5328,
+      "step": 31656
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.418825470615588e-06,
+      "loss": 1.0457,
+      "step": 31657
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.418387645802295e-06,
+      "loss": 0.9316,
+      "step": 31658
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4179498833954397e-06,
+      "loss": 0.8167,
+      "step": 31659
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4175121833982052e-06,
+      "loss": 0.9278,
+      "step": 31660
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4170745458137747e-06,
+      "loss": 0.653,
+      "step": 31661
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4166369706453343e-06,
+      "loss": 0.9888,
+      "step": 31662
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4161994578960615e-06,
+      "loss": 0.934,
+      "step": 31663
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4157620075691392e-06,
+      "loss": 1.0005,
+      "step": 31664
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4153246196677483e-06,
+      "loss": 0.8484,
+      "step": 31665
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.414887294195071e-06,
+      "loss": 0.998,
+      "step": 31666
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.414450031154283e-06,
+      "loss": 0.9058,
+      "step": 31667
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4140128305485712e-06,
+      "loss": 0.9597,
+      "step": 31668
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4135756923811117e-06,
+      "loss": 0.7492,
+      "step": 31669
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4131386166550832e-06,
+      "loss": 0.832,
+      "step": 31670
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412701603373663e-06,
+      "loss": 0.7072,
+      "step": 31671
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.412264652540032e-06,
+      "loss": 1.141,
+      "step": 31672
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4118277641573653e-06,
+      "loss": 0.8117,
+      "step": 31673
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4113909382288405e-06,
+      "loss": 0.8874,
+      "step": 31674
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4109541747576349e-06,
+      "loss": 0.9007,
+      "step": 31675
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.410517473746924e-06,
+      "loss": 0.5672,
+      "step": 31676
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.410080835199884e-06,
+      "loss": 0.6167,
+      "step": 31677
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4096442591196902e-06,
+      "loss": 1.0322,
+      "step": 31678
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4092077455095155e-06,
+      "loss": 0.6749,
+      "step": 31679
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4087712943725384e-06,
+      "loss": 0.9716,
+      "step": 31680
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.40833490571193e-06,
+      "loss": 0.7874,
+      "step": 31681
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4078985795308664e-06,
+      "loss": 0.7084,
+      "step": 31682
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4074623158325163e-06,
+      "loss": 0.8294,
+      "step": 31683
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4070261146200536e-06,
+      "loss": 0.7724,
+      "step": 31684
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4065899758966496e-06,
+      "loss": 0.6632,
+      "step": 31685
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.40615389966548e-06,
+      "loss": 0.8258,
+      "step": 31686
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4057178859297139e-06,
+      "loss": 0.6821,
+      "step": 31687
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4052819346925217e-06,
+      "loss": 0.6984,
+      "step": 31688
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4048460459570735e-06,
+      "loss": 0.7396,
+      "step": 31689
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4044102197265387e-06,
+      "loss": 0.902,
+      "step": 31690
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.403974456004088e-06,
+      "loss": 0.8209,
+      "step": 31691
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.403538754792889e-06,
+      "loss": 0.8566,
+      "step": 31692
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4031031160961105e-06,
+      "loss": 0.934,
+      "step": 31693
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.402667539916922e-06,
+      "loss": 0.3897,
+      "step": 31694
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.40223202625849e-06,
+      "loss": 0.7115,
+      "step": 31695
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4017965751239805e-06,
+      "loss": 0.7444,
+      "step": 31696
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4013611865165589e-06,
+      "loss": 0.816,
+      "step": 31697
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.400925860439396e-06,
+      "loss": 0.9202,
+      "step": 31698
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4004905968956551e-06,
+      "loss": 0.9929,
+      "step": 31699
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4000553958885022e-06,
+      "loss": 0.6809,
+      "step": 31700
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3996202574211003e-06,
+      "loss": 0.6693,
+      "step": 31701
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3991851814966184e-06,
+      "loss": 0.6765,
+      "step": 31702
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3987501681182114e-06,
+      "loss": 0.8679,
+      "step": 31703
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3983152172890513e-06,
+      "loss": 0.7345,
+      "step": 31704
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3978803290122978e-06,
+      "loss": 0.5402,
+      "step": 31705
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3974455032911137e-06,
+      "loss": 0.84,
+      "step": 31706
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3970107401286615e-06,
+      "loss": 0.6164,
+      "step": 31707
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3965760395281025e-06,
+      "loss": 0.7386,
+      "step": 31708
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3961414014925977e-06,
+      "loss": 0.7749,
+      "step": 31709
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.395706826025306e-06,
+      "loss": 0.8192,
+      "step": 31710
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3952723131293922e-06,
+      "loss": 0.7708,
+      "step": 31711
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3948378628080162e-06,
+      "loss": 0.4919,
+      "step": 31712
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.394403475064332e-06,
+      "loss": 0.837,
+      "step": 31713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3939691499015007e-06,
+      "loss": 0.7439,
+      "step": 31714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3935348873226807e-06,
+      "loss": 0.7802,
+      "step": 31715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3931006873310316e-06,
+      "loss": 0.6476,
+      "step": 31716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.392666549929711e-06,
+      "loss": 0.6183,
+      "step": 31717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.392232475121874e-06,
+      "loss": 0.7863,
+      "step": 31718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3917984629106795e-06,
+      "loss": 0.7009,
+      "step": 31719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3913645132992826e-06,
+      "loss": 0.9897,
+      "step": 31720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3909306262908383e-06,
+      "loss": 0.5244,
+      "step": 31721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3904968018885024e-06,
+      "loss": 0.6598,
+      "step": 31722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3900630400954295e-06,
+      "loss": 0.7771,
+      "step": 31723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3896293409147754e-06,
+      "loss": 0.9959,
+      "step": 31724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3891957043496917e-06,
+      "loss": 0.847,
+      "step": 31725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3887621304033338e-06,
+      "loss": 1.0644,
+      "step": 31726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3883286190788535e-06,
+      "loss": 0.9681,
+      "step": 31727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3878951703794019e-06,
+      "loss": 0.7083,
+      "step": 31728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.387461784308135e-06,
+      "loss": 1.0898,
+      "step": 31729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3870284608682028e-06,
+      "loss": 0.5993,
+      "step": 31730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3865952000627558e-06,
+      "loss": 0.6795,
+      "step": 31731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3861620018949473e-06,
+      "loss": 1.1532,
+      "step": 31732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3857288663679202e-06,
+      "loss": 0.9558,
+      "step": 31733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3852957934848321e-06,
+      "loss": 0.7209,
+      "step": 31734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3848627832488304e-06,
+      "loss": 1.0386,
+      "step": 31735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3844298356630626e-06,
+      "loss": 0.9822,
+      "step": 31736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3839969507306773e-06,
+      "loss": 0.5466,
+      "step": 31737
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3835641284548229e-06,
+      "loss": 0.7997,
+      "step": 31738
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3831313688386471e-06,
+      "loss": 0.7168,
+      "step": 31739
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3826986718852952e-06,
+      "loss": 0.8259,
+      "step": 31740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.382266037597918e-06,
+      "loss": 0.8285,
+      "step": 31741
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3818334659796605e-06,
+      "loss": 1.0496,
+      "step": 31742
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3814009570336651e-06,
+      "loss": 0.5382,
+      "step": 31743
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3809685107630788e-06,
+      "loss": 0.8947,
+      "step": 31744
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3805361271710471e-06,
+      "loss": 0.6908,
+      "step": 31745
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3801038062607118e-06,
+      "loss": 0.9711,
+      "step": 31746
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3796715480352207e-06,
+      "loss": 0.9796,
+      "step": 31747
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3792393524977154e-06,
+      "loss": 0.6265,
+      "step": 31748
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3788072196513392e-06,
+      "loss": 0.7807,
+      "step": 31749
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3783751494992338e-06,
+      "loss": 0.9052,
+      "step": 31750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.377943142044541e-06,
+      "loss": 1.004,
+      "step": 31751
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3775111972904042e-06,
+      "loss": 0.5859,
+      "step": 31752
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3770793152399631e-06,
+      "loss": 0.9335,
+      "step": 31753
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3766474958963594e-06,
+      "loss": 0.7272,
+      "step": 31754
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3762157392627317e-06,
+      "loss": 0.8248,
+      "step": 31755
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.375784045342221e-06,
+      "loss": 0.889,
+      "step": 31756
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.375352414137967e-06,
+      "loss": 0.8461,
+      "step": 31757
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3749208456531048e-06,
+      "loss": 0.9509,
+      "step": 31758
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.37448933989078e-06,
+      "loss": 0.879,
+      "step": 31759
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3740578968541252e-06,
+      "loss": 1.0065,
+      "step": 31760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3736265165462792e-06,
+      "loss": 1.0118,
+      "step": 31761
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3731951989703817e-06,
+      "loss": 0.9509,
+      "step": 31762
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3727639441295648e-06,
+      "loss": 0.7127,
+      "step": 31763
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3723327520269636e-06,
+      "loss": 1.0434,
+      "step": 31764
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3719016226657189e-06,
+      "loss": 0.6883,
+      "step": 31765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3714705560489638e-06,
+      "loss": 1.066,
+      "step": 31766
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3710395521798336e-06,
+      "loss": 1.2739,
+      "step": 31767
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3706086110614614e-06,
+      "loss": 0.9989,
+      "step": 31768
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3701777326969823e-06,
+      "loss": 0.9044,
+      "step": 31769
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3697469170895282e-06,
+      "loss": 0.6675,
+      "step": 31770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3693161642422315e-06,
+      "loss": 0.7847,
+      "step": 31771
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3688854741582302e-06,
+      "loss": 0.6115,
+      "step": 31772
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.36845484684065e-06,
+      "loss": 0.8387,
+      "step": 31773
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3680242822926247e-06,
+      "loss": 1.0959,
+      "step": 31774
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3675937805172844e-06,
+      "loss": 0.7904,
+      "step": 31775
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3671633415177598e-06,
+      "loss": 0.6508,
+      "step": 31776
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3667329652971839e-06,
+      "loss": 0.777,
+      "step": 31777
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3663026518586842e-06,
+      "loss": 0.8866,
+      "step": 31778
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3658724012053914e-06,
+      "loss": 0.8201,
+      "step": 31779
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.365442213340432e-06,
+      "loss": 1.0393,
+      "step": 31780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.365012088266936e-06,
+      "loss": 0.7025,
+      "step": 31781
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3645820259880315e-06,
+      "loss": 0.7608,
+      "step": 31782
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3641520265068453e-06,
+      "loss": 0.9499,
+      "step": 31783
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3637220898265046e-06,
+      "loss": 1.0574,
+      "step": 31784
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.363292215950135e-06,
+      "loss": 0.752,
+      "step": 31785
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3628624048808648e-06,
+      "loss": 0.9217,
+      "step": 31786
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3624326566218172e-06,
+      "loss": 0.7949,
+      "step": 31787
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3620029711761197e-06,
+      "loss": 0.559,
+      "step": 31788
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3615733485468929e-06,
+      "loss": 0.9186,
+      "step": 31789
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3611437887372657e-06,
+      "loss": 1.0167,
+      "step": 31790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3607142917503613e-06,
+      "loss": 0.8299,
+      "step": 31791
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3602848575893034e-06,
+      "loss": 1.0875,
+      "step": 31792
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.359855486257211e-06,
+      "loss": 0.8974,
+      "step": 31793
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.359426177757207e-06,
+      "loss": 0.8309,
+      "step": 31794
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.358996932092418e-06,
+      "loss": 0.8272,
+      "step": 31795
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3585677492659622e-06,
+      "loss": 0.7866,
+      "step": 31796
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3581386292809617e-06,
+      "loss": 0.7043,
+      "step": 31797
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3577095721405375e-06,
+      "loss": 0.5021,
+      "step": 31798
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3572805778478083e-06,
+      "loss": 0.5749,
+      "step": 31799
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3568516464058946e-06,
+      "loss": 0.8288,
+      "step": 31800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3564227778179128e-06,
+      "loss": 0.9693,
+      "step": 31801
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3559939720869886e-06,
+      "loss": 0.7563,
+      "step": 31802
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3555652292162348e-06,
+      "loss": 0.8631,
+      "step": 31803
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3551365492087698e-06,
+      "loss": 0.8321,
+      "step": 31804
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3547079320677126e-06,
+      "loss": 0.743,
+      "step": 31805
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3542793777961794e-06,
+      "loss": 0.7661,
+      "step": 31806
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3538508863972843e-06,
+      "loss": 0.8725,
+      "step": 31807
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3534224578741472e-06,
+      "loss": 0.5817,
+      "step": 31808
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3529940922298824e-06,
+      "loss": 0.8392,
+      "step": 31809
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3525657894676059e-06,
+      "loss": 0.6597,
+      "step": 31810
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.35213754959043e-06,
+      "loss": 0.9285,
+      "step": 31811
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3517093726014707e-06,
+      "loss": 0.7318,
+      "step": 31812
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3512812585038405e-06,
+      "loss": 0.6909,
+      "step": 31813
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3508532073006542e-06,
+      "loss": 0.8424,
+      "step": 31814
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.350425218995024e-06,
+      "loss": 0.9,
+      "step": 31815
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3499972935900618e-06,
+      "loss": 0.7334,
+      "step": 31816
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3495694310888795e-06,
+      "loss": 1.029,
+      "step": 31817
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3491416314945894e-06,
+      "loss": 0.4226,
+      "step": 31818
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3487138948102996e-06,
+      "loss": 0.7442,
+      "step": 31819
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3482862210391246e-06,
+      "loss": 0.555,
+      "step": 31820
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.347858610184174e-06,
+      "loss": 0.8669,
+      "step": 31821
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3474310622485587e-06,
+      "loss": 0.7745,
+      "step": 31822
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.347003577235383e-06,
+      "loss": 0.4697,
+      "step": 31823
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3465761551477574e-06,
+      "loss": 1.1032,
+      "step": 31824
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3461487959887898e-06,
+      "loss": 0.8624,
+      "step": 31825
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3457214997615909e-06,
+      "loss": 0.7842,
+      "step": 31826
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3452942664692658e-06,
+      "loss": 0.8895,
+      "step": 31827
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3448670961149224e-06,
+      "loss": 0.8894,
+      "step": 31828
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3444399887016668e-06,
+      "loss": 0.9171,
+      "step": 31829
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3440129442326045e-06,
+      "loss": 1.0011,
+      "step": 31830
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3435859627108415e-06,
+      "loss": 0.7514,
+      "step": 31831
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3431590441394826e-06,
+      "loss": 0.5328,
+      "step": 31832
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3427321885216316e-06,
+      "loss": 0.9505,
+      "step": 31833
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.342305395860395e-06,
+      "loss": 0.7551,
+      "step": 31834
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3418786661588745e-06,
+      "loss": 0.6793,
+      "step": 31835
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3414519994201724e-06,
+      "loss": 0.802,
+      "step": 31836
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3410253956473928e-06,
+      "loss": 0.653,
+      "step": 31837
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3405988548436389e-06,
+      "loss": 0.7643,
+      "step": 31838
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3401723770120124e-06,
+      "loss": 0.798,
+      "step": 31839
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.339745962155613e-06,
+      "loss": 0.6508,
+      "step": 31840
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3393196102775441e-06,
+      "loss": 0.8122,
+      "step": 31841
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3388933213809063e-06,
+      "loss": 1.0636,
+      "step": 31842
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3384670954687929e-06,
+      "loss": 0.9759,
+      "step": 31843
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3380409325443122e-06,
+      "loss": 0.5715,
+      "step": 31844
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3376148326105586e-06,
+      "loss": 0.8465,
+      "step": 31845
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.337188795670632e-06,
+      "loss": 0.7356,
+      "step": 31846
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3367628217276307e-06,
+      "loss": 0.8962,
+      "step": 31847
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3363369107846524e-06,
+      "loss": 0.8482,
+      "step": 31848
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3359110628447948e-06,
+      "loss": 0.6963,
+      "step": 31849
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3354852779111504e-06,
+      "loss": 1.1019,
+      "step": 31850
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3350595559868229e-06,
+      "loss": 0.7455,
+      "step": 31851
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3346338970749052e-06,
+      "loss": 0.6903,
+      "step": 31852
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3342083011784912e-06,
+      "loss": 0.8311,
+      "step": 31853
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3337827683006755e-06,
+      "loss": 0.6955,
+      "step": 31854
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3333572984445552e-06,
+      "loss": 1.0534,
+      "step": 31855
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3329318916132194e-06,
+      "loss": 0.9824,
+      "step": 31856
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3325065478097687e-06,
+      "loss": 0.6899,
+      "step": 31857
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.332081267037293e-06,
+      "loss": 0.7516,
+      "step": 31858
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3316560492988838e-06,
+      "loss": 0.7702,
+      "step": 31859
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3312308945976348e-06,
+      "loss": 0.8293,
+      "step": 31860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3308058029366377e-06,
+      "loss": 1.0621,
+      "step": 31861
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3303807743189834e-06,
+      "loss": 0.9469,
+      "step": 31862
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3299558087477626e-06,
+      "loss": 0.5909,
+      "step": 31863
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3295309062260663e-06,
+      "loss": 0.9053,
+      "step": 31864
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3291060667569843e-06,
+      "loss": 0.6456,
+      "step": 31865
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3286812903436052e-06,
+      "loss": 0.9279,
+      "step": 31866
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3282565769890188e-06,
+      "loss": 0.6409,
+      "step": 31867
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3278319266963125e-06,
+      "loss": 0.8391,
+      "step": 31868
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3274073394685761e-06,
+      "loss": 0.8762,
+      "step": 31869
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.326982815308897e-06,
+      "loss": 0.8875,
+      "step": 31870
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.326558354220362e-06,
+      "loss": 1.0197,
+      "step": 31871
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3261339562060605e-06,
+      "loss": 0.9388,
+      "step": 31872
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3257096212690735e-06,
+      "loss": 0.741,
+      "step": 31873
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3252853494124874e-06,
+      "loss": 0.9358,
+      "step": 31874
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3248611406393918e-06,
+      "loss": 0.8284,
+      "step": 31875
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.324436994952869e-06,
+      "loss": 0.8677,
+      "step": 31876
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3240129123560041e-06,
+      "loss": 0.7785,
+      "step": 31877
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3235888928518804e-06,
+      "loss": 0.7542,
+      "step": 31878
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3231649364435816e-06,
+      "loss": 0.9194,
+      "step": 31879
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3227410431341914e-06,
+      "loss": 1.0383,
+      "step": 31880
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3223172129267891e-06,
+      "loss": 0.8548,
+      "step": 31881
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3218934458244649e-06,
+      "loss": 0.4954,
+      "step": 31882
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3214697418302924e-06,
+      "loss": 0.7584,
+      "step": 31883
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3210461009473562e-06,
+      "loss": 1.0353,
+      "step": 31884
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3206225231787362e-06,
+      "loss": 0.8677,
+      "step": 31885
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3201990085275106e-06,
+      "loss": 0.8451,
+      "step": 31886
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3197755569967641e-06,
+      "loss": 0.8364,
+      "step": 31887
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3193521685895738e-06,
+      "loss": 0.7784,
+      "step": 31888
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3189288433090187e-06,
+      "loss": 0.655,
+      "step": 31889
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.318505581158177e-06,
+      "loss": 0.9246,
+      "step": 31890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3180823821401267e-06,
+      "loss": 0.3159,
+      "step": 31891
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3176592462579451e-06,
+      "loss": 0.8935,
+      "step": 31892
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3172361735147098e-06,
+      "loss": 1.0738,
+      "step": 31893
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3168131639134973e-06,
+      "loss": 0.7825,
+      "step": 31894
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.316390217457384e-06,
+      "loss": 0.8496,
+      "step": 31895
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3159673341494438e-06,
+      "loss": 1.0236,
+      "step": 31896
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3155445139927558e-06,
+      "loss": 1.0449,
+      "step": 31897
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3151217569903908e-06,
+      "loss": 0.8381,
+      "step": 31898
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3146990631454237e-06,
+      "loss": 0.8136,
+      "step": 31899
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3142764324609303e-06,
+      "loss": 0.6584,
+      "step": 31900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3138538649399846e-06,
+      "loss": 0.5593,
+      "step": 31901
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3134313605856596e-06,
+      "loss": 0.8519,
+      "step": 31902
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3130089194010242e-06,
+      "loss": 0.8039,
+      "step": 31903
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3125865413891504e-06,
+      "loss": 0.9452,
+      "step": 31904
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3121642265531154e-06,
+      "loss": 0.765,
+      "step": 31905
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3117419748959858e-06,
+      "loss": 1.1168,
+      "step": 31906
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3113197864208327e-06,
+      "loss": 0.8738,
+      "step": 31907
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3108976611307278e-06,
+      "loss": 0.9111,
+      "step": 31908
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.31047559902874e-06,
+      "loss": 0.6763,
+      "step": 31909
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3100536001179387e-06,
+      "loss": 0.9986,
+      "step": 31910
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3096316644013907e-06,
+      "loss": 0.9926,
+      "step": 31911
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.30920979188217e-06,
+      "loss": 0.8886,
+      "step": 31912
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3087879825633386e-06,
+      "loss": 0.6986,
+      "step": 31913
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3083662364479665e-06,
+      "loss": 0.8754,
+      "step": 31914
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.30794455353912e-06,
+      "loss": 0.861,
+      "step": 31915
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3075229338398665e-06,
+      "loss": 0.6843,
+      "step": 31916
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3071013773532682e-06,
+      "loss": 0.7121,
+      "step": 31917
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3066798840823968e-06,
+      "loss": 1.1992,
+      "step": 31918
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3062584540303146e-06,
+      "loss": 0.7448,
+      "step": 31919
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.305837087200087e-06,
+      "loss": 0.7857,
+      "step": 31920
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3054157835947768e-06,
+      "loss": 1.1656,
+      "step": 31921
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3049945432174492e-06,
+      "loss": 0.9127,
+      "step": 31922
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3045733660711668e-06,
+      "loss": 0.5651,
+      "step": 31923
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.304152252158991e-06,
+      "loss": 0.822,
+      "step": 31924
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3037312014839875e-06,
+      "loss": 0.5758,
+      "step": 31925
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.303310214049216e-06,
+      "loss": 0.6538,
+      "step": 31926
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3028892898577372e-06,
+      "loss": 0.6449,
+      "step": 31927
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3024684289126143e-06,
+      "loss": 0.7637,
+      "step": 31928
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.302047631216905e-06,
+      "loss": 0.8675,
+      "step": 31929
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3016268967736722e-06,
+      "loss": 0.8046,
+      "step": 31930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3012062255859759e-06,
+      "loss": 1.2066,
+      "step": 31931
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3007856176568757e-06,
+      "loss": 0.7139,
+      "step": 31932
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3003650729894258e-06,
+      "loss": 0.7301,
+      "step": 31933
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2999445915866871e-06,
+      "loss": 0.5278,
+      "step": 31934
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.299524173451715e-06,
+      "loss": 0.8595,
+      "step": 31935
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2991038185875726e-06,
+      "loss": 0.7775,
+      "step": 31936
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.298683526997313e-06,
+      "loss": 0.5623,
+      "step": 31937
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2982632986839926e-06,
+      "loss": 0.806,
+      "step": 31938
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.297843133650668e-06,
+      "loss": 0.4824,
+      "step": 31939
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2974230319003944e-06,
+      "loss": 0.7477,
+      "step": 31940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2970029934362272e-06,
+      "loss": 0.5513,
+      "step": 31941
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2965830182612205e-06,
+      "loss": 0.939,
+      "step": 31942
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2961631063784287e-06,
+      "loss": 0.7929,
+      "step": 31943
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2957432577909056e-06,
+      "loss": 0.9755,
+      "step": 31944
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2953234725017039e-06,
+      "loss": 0.831,
+      "step": 31945
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2949037505138762e-06,
+      "loss": 0.8032,
+      "step": 31946
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2944840918304735e-06,
+      "loss": 0.7008,
+      "step": 31947
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.29406449645455e-06,
+      "loss": 0.8434,
+      "step": 31948
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.293644964389158e-06,
+      "loss": 0.5993,
+      "step": 31949
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2932254956373457e-06,
+      "loss": 0.9204,
+      "step": 31950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2928060902021643e-06,
+      "loss": 0.871,
+      "step": 31951
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.292386748086667e-06,
+      "loss": 0.6181,
+      "step": 31952
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2919674692938956e-06,
+      "loss": 1.0401,
+      "step": 31953
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2915482538269054e-06,
+      "loss": 0.811,
+      "step": 31954
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2911291016887439e-06,
+      "loss": 1.0383,
+      "step": 31955
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2907100128824579e-06,
+      "loss": 0.6546,
+      "step": 31956
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2902909874110968e-06,
+      "loss": 0.8523,
+      "step": 31957
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2898720252777063e-06,
+      "loss": 0.8561,
+      "step": 31958
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.289453126485334e-06,
+      "loss": 0.6801,
+      "step": 31959
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2890342910370234e-06,
+      "loss": 0.9573,
+      "step": 31960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.288615518935825e-06,
+      "loss": 0.7638,
+      "step": 31961
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2881968101847841e-06,
+      "loss": 0.8185,
+      "step": 31962
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2877781647869414e-06,
+      "loss": 0.7199,
+      "step": 31963
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2873595827453433e-06,
+      "loss": 0.8269,
+      "step": 31964
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.286941064063031e-06,
+      "loss": 0.7045,
+      "step": 31965
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2865226087430539e-06,
+      "loss": 0.8411,
+      "step": 31966
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2861042167884518e-06,
+      "loss": 0.7104,
+      "step": 31967
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2856858882022671e-06,
+      "loss": 0.7847,
+      "step": 31968
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2852676229875427e-06,
+      "loss": 0.9068,
+      "step": 31969
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2848494211473196e-06,
+      "loss": 0.853,
+      "step": 31970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2844312826846395e-06,
+      "loss": 0.88,
+      "step": 31971
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2840132076025425e-06,
+      "loss": 1.0085,
+      "step": 31972
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2835951959040693e-06,
+      "loss": 0.7679,
+      "step": 31973
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.283177247592261e-06,
+      "loss": 0.9908,
+      "step": 31974
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.282759362670155e-06,
+      "loss": 1.003,
+      "step": 31975
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2823415411407902e-06,
+      "loss": 0.7912,
+      "step": 31976
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2819237830072073e-06,
+      "loss": 0.6291,
+      "step": 31977
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2815060882724395e-06,
+      "loss": 0.9876,
+      "step": 31978
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.28108845693953e-06,
+      "loss": 0.7755,
+      "step": 31979
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2806708890115138e-06,
+      "loss": 0.9225,
+      "step": 31980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2802533844914266e-06,
+      "loss": 0.8679,
+      "step": 31981
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2798359433823071e-06,
+      "loss": 0.6751,
+      "step": 31982
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.279418565687185e-06,
+      "loss": 0.6872,
+      "step": 31983
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2790012514091022e-06,
+      "loss": 0.8668,
+      "step": 31984
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2785840005510897e-06,
+      "loss": 0.6715,
+      "step": 31985
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.278166813116184e-06,
+      "loss": 0.7764,
+      "step": 31986
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2777496891074171e-06,
+      "loss": 0.9087,
+      "step": 31987
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.277332628527823e-06,
+      "loss": 1.1865,
+      "step": 31988
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2769156313804343e-06,
+      "loss": 0.9768,
+      "step": 31989
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2764986976682825e-06,
+      "loss": 0.9978,
+      "step": 31990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.276081827394403e-06,
+      "loss": 0.4919,
+      "step": 31991
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.275665020561827e-06,
+      "loss": 0.689,
+      "step": 31992
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2752482771735808e-06,
+      "loss": 0.7955,
+      "step": 31993
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2748315972326997e-06,
+      "loss": 0.814,
+      "step": 31994
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2744149807422113e-06,
+      "loss": 0.7409,
+      "step": 31995
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2739984277051432e-06,
+      "loss": 0.694,
+      "step": 31996
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2735819381245307e-06,
+      "loss": 1.0512,
+      "step": 31997
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2731655120033991e-06,
+      "loss": 0.8412,
+      "step": 31998
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2727491493447773e-06,
+      "loss": 0.7373,
+      "step": 31999
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2723328501516918e-06,
+      "loss": 0.7657,
+      "step": 32000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2719166144271722e-06,
+      "loss": 0.5606,
+      "step": 32001
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.271500442174244e-06,
+      "loss": 1.1088,
+      "step": 32002
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2710843333959333e-06,
+      "loss": 0.8569,
+      "step": 32003
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.270668288095267e-06,
+      "loss": 0.8958,
+      "step": 32004
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2702523062752704e-06,
+      "loss": 0.8979,
+      "step": 32005
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2698363879389685e-06,
+      "loss": 0.9026,
+      "step": 32006
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.269420533089385e-06,
+      "loss": 0.5692,
+      "step": 32007
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2690047417295437e-06,
+      "loss": 0.7728,
+      "step": 32008
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2685890138624713e-06,
+      "loss": 0.8775,
+      "step": 32009
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2681733494911897e-06,
+      "loss": 0.8209,
+      "step": 32010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2677577486187199e-06,
+      "loss": 1.0124,
+      "step": 32011
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2673422112480882e-06,
+      "loss": 1.0229,
+      "step": 32012
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2669267373823123e-06,
+      "loss": 0.7141,
+      "step": 32013
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2665113270244123e-06,
+      "loss": 0.8514,
+      "step": 32014
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.266095980177413e-06,
+      "loss": 0.6777,
+      "step": 32015
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2656806968443348e-06,
+      "loss": 0.6967,
+      "step": 32016
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2652654770281958e-06,
+      "loss": 0.7038,
+      "step": 32017
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2648503207320162e-06,
+      "loss": 0.7354,
+      "step": 32018
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2644352279588156e-06,
+      "loss": 0.6993,
+      "step": 32019
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.264020198711612e-06,
+      "loss": 0.8756,
+      "step": 32020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2636052329934212e-06,
+      "loss": 0.7964,
+      "step": 32021
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.263190330807268e-06,
+      "loss": 0.8349,
+      "step": 32022
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.262775492156162e-06,
+      "loss": 0.701,
+      "step": 32023
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.262360717043124e-06,
+      "loss": 0.8658,
+      "step": 32024
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2619460054711685e-06,
+      "loss": 0.7552,
+      "step": 32025
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2615313574433098e-06,
+      "loss": 0.862,
+      "step": 32026
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2611167729625673e-06,
+      "loss": 0.7352,
+      "step": 32027
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2607022520319545e-06,
+      "loss": 0.8571,
+      "step": 32028
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2602877946544856e-06,
+      "loss": 0.8255,
+      "step": 32029
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2598734008331737e-06,
+      "loss": 0.8349,
+      "step": 32030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.259459070571033e-06,
+      "loss": 0.6946,
+      "step": 32031
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2590448038710778e-06,
+      "loss": 0.7544,
+      "step": 32032
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.258630600736318e-06,
+      "loss": 1.0562,
+      "step": 32033
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2582164611697678e-06,
+      "loss": 0.9082,
+      "step": 32034
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.257802385174438e-06,
+      "loss": 0.7197,
+      "step": 32035
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.257388372753341e-06,
+      "loss": 0.7285,
+      "step": 32036
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.256974423909485e-06,
+      "loss": 0.8637,
+      "step": 32037
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2565605386458835e-06,
+      "loss": 0.8561,
+      "step": 32038
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2561467169655417e-06,
+      "loss": 0.7883,
+      "step": 32039
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2557329588714739e-06,
+      "loss": 0.7019,
+      "step": 32040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2553192643666878e-06,
+      "loss": 0.9392,
+      "step": 32041
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.254905633454192e-06,
+      "loss": 0.9719,
+      "step": 32042
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.254492066136993e-06,
+      "loss": 0.696,
+      "step": 32043
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2540785624180952e-06,
+      "loss": 1.0774,
+      "step": 32044
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2536651223005125e-06,
+      "loss": 1.0096,
+      "step": 32045
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2532517457872484e-06,
+      "loss": 0.9509,
+      "step": 32046
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.252838432881308e-06,
+      "loss": 0.8172,
+      "step": 32047
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2524251835856981e-06,
+      "loss": 0.8918,
+      "step": 32048
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.252011997903424e-06,
+      "loss": 0.8221,
+      "step": 32049
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2515988758374897e-06,
+      "loss": 0.7301,
+      "step": 32050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2511858173908986e-06,
+      "loss": 0.8663,
+      "step": 32051
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2507728225666594e-06,
+      "loss": 0.7148,
+      "step": 32052
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2503598913677694e-06,
+      "loss": 0.5903,
+      "step": 32053
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2499470237972345e-06,
+      "loss": 0.7304,
+      "step": 32054
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2495342198580562e-06,
+      "loss": 1.0033,
+      "step": 32055
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.249121479553237e-06,
+      "loss": 0.8515,
+      "step": 32056
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2487088028857751e-06,
+      "loss": 0.7074,
+      "step": 32057
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2482961898586776e-06,
+      "loss": 0.7703,
+      "step": 32058
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2478836404749416e-06,
+      "loss": 0.729,
+      "step": 32059
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2474711547375684e-06,
+      "loss": 0.7597,
+      "step": 32060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2470587326495564e-06,
+      "loss": 0.8123,
+      "step": 32061
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2466463742139056e-06,
+      "loss": 0.699,
+      "step": 32062
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2462340794336136e-06,
+      "loss": 0.8533,
+      "step": 32063
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2458218483116802e-06,
+      "loss": 0.9821,
+      "step": 32064
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2454096808511029e-06,
+      "loss": 0.8749,
+      "step": 32065
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2449975770548773e-06,
+      "loss": 1.1844,
+      "step": 32066
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2445855369260028e-06,
+      "loss": 0.7954,
+      "step": 32067
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2441735604674731e-06,
+      "loss": 0.7172,
+      "step": 32068
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2437616476822846e-06,
+      "loss": 0.8675,
+      "step": 32069
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2433497985734356e-06,
+      "loss": 1.0264,
+      "step": 32070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2429380131439195e-06,
+      "loss": 1.0396,
+      "step": 32071
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2425262913967317e-06,
+      "loss": 0.6595,
+      "step": 32072
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2421146333348643e-06,
+      "loss": 1.0403,
+      "step": 32073
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2417030389613116e-06,
+      "loss": 0.8546,
+      "step": 32074
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2412915082790645e-06,
+      "loss": 1.0197,
+      "step": 32075
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2408800412911203e-06,
+      "loss": 0.9217,
+      "step": 32076
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2404686380004694e-06,
+      "loss": 0.5507,
+      "step": 32077
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2400572984101034e-06,
+      "loss": 0.7392,
+      "step": 32078
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2396460225230134e-06,
+      "loss": 0.8623,
+      "step": 32079
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2392348103421903e-06,
+      "loss": 0.8327,
+      "step": 32080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.238823661870624e-06,
+      "loss": 0.8996,
+      "step": 32081
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2384125771113053e-06,
+      "loss": 0.8361,
+      "step": 32082
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2380015560672233e-06,
+      "loss": 0.8438,
+      "step": 32083
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2375905987413673e-06,
+      "loss": 1.0842,
+      "step": 32084
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.237179705136725e-06,
+      "loss": 1.0715,
+      "step": 32085
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2367688752562857e-06,
+      "loss": 0.7318,
+      "step": 32086
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.236358109103033e-06,
+      "loss": 0.7246,
+      "step": 32087
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2359474066799605e-06,
+      "loss": 0.6406,
+      "step": 32088
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.235536767990051e-06,
+      "loss": 0.6358,
+      "step": 32089
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2351261930362912e-06,
+      "loss": 0.8065,
+      "step": 32090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2347156818216677e-06,
+      "loss": 0.7682,
+      "step": 32091
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2343052343491668e-06,
+      "loss": 0.78,
+      "step": 32092
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2338948506217685e-06,
+      "loss": 0.6735,
+      "step": 32093
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2334845306424615e-06,
+      "loss": 0.9914,
+      "step": 32094
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2330742744142277e-06,
+      "loss": 0.9544,
+      "step": 32095
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2326640819400525e-06,
+      "loss": 0.7124,
+      "step": 32096
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2322539532229172e-06,
+      "loss": 0.9058,
+      "step": 32097
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2318438882658057e-06,
+      "loss": 1.0035,
+      "step": 32098
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.231433887071698e-06,
+      "loss": 0.7211,
+      "step": 32099
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2310239496435749e-06,
+      "loss": 1.0076,
+      "step": 32100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2306140759844222e-06,
+      "loss": 0.8964,
+      "step": 32101
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2302042660972192e-06,
+      "loss": 0.8047,
+      "step": 32102
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2297945199849426e-06,
+      "loss": 0.8284,
+      "step": 32103
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2293848376505747e-06,
+      "loss": 0.7178,
+      "step": 32104
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2289752190970905e-06,
+      "loss": 0.749,
+      "step": 32105
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2285656643274758e-06,
+      "loss": 0.5104,
+      "step": 32106
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2281561733447057e-06,
+      "loss": 0.6814,
+      "step": 32107
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2277467461517579e-06,
+      "loss": 0.6956,
+      "step": 32108
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.22733738275161e-06,
+      "loss": 0.8234,
+      "step": 32109
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2269280831472386e-06,
+      "loss": 0.7804,
+      "step": 32110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2265188473416212e-06,
+      "loss": 0.5363,
+      "step": 32111
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2261096753377322e-06,
+      "loss": 1.0578,
+      "step": 32112
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.225700567138548e-06,
+      "loss": 0.9525,
+      "step": 32113
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2252915227470442e-06,
+      "loss": 0.4934,
+      "step": 32114
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2248825421661937e-06,
+      "loss": 0.7921,
+      "step": 32115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.224473625398973e-06,
+      "loss": 0.8062,
+      "step": 32116
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2240647724483533e-06,
+      "loss": 0.9137,
+      "step": 32117
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2236559833173078e-06,
+      "loss": 0.7544,
+      "step": 32118
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2232472580088129e-06,
+      "loss": 0.9936,
+      "step": 32119
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2228385965258383e-06,
+      "loss": 0.5918,
+      "step": 32120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2224299988713563e-06,
+      "loss": 0.6701,
+      "step": 32121
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2220214650483397e-06,
+      "loss": 0.6307,
+      "step": 32122
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2216129950597545e-06,
+      "loss": 0.5468,
+      "step": 32123
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2212045889085756e-06,
+      "loss": 1.0805,
+      "step": 32124
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.220796246597773e-06,
+      "loss": 0.7983,
+      "step": 32125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2203879681303132e-06,
+      "loss": 1.324,
+      "step": 32126
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2199797535091683e-06,
+      "loss": 0.6694,
+      "step": 32127
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.219571602737305e-06,
+      "loss": 0.6461,
+      "step": 32128
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2191635158176928e-06,
+      "loss": 1.1002,
+      "step": 32129
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2187554927532963e-06,
+      "loss": 0.9924,
+      "step": 32130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2183475335470863e-06,
+      "loss": 0.8573,
+      "step": 32131
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2179396382020314e-06,
+      "loss": 1.0752,
+      "step": 32132
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2175318067210918e-06,
+      "loss": 0.5769,
+      "step": 32133
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.217124039107237e-06,
+      "loss": 0.4971,
+      "step": 32134
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2167163353634303e-06,
+      "loss": 1.0521,
+      "step": 32135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2163086954926363e-06,
+      "loss": 0.77,
+      "step": 32136
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.215901119497823e-06,
+      "loss": 0.6191,
+      "step": 32137
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2154936073819524e-06,
+      "loss": 0.8905,
+      "step": 32138
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.215086159147988e-06,
+      "loss": 0.6638,
+      "step": 32139
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.214678774798892e-06,
+      "loss": 0.6878,
+      "step": 32140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2142714543376277e-06,
+      "loss": 0.5576,
+      "step": 32141
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2138641977671562e-06,
+      "loss": 0.8932,
+      "step": 32142
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2134570050904414e-06,
+      "loss": 0.87,
+      "step": 32143
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2130498763104415e-06,
+      "loss": 1.0367,
+      "step": 32144
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2126428114301204e-06,
+      "loss": 0.8537,
+      "step": 32145
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2122358104524356e-06,
+      "loss": 0.8019,
+      "step": 32146
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2118288733803474e-06,
+      "loss": 0.8573,
+      "step": 32147
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2114220002168143e-06,
+      "loss": 0.6924,
+      "step": 32148
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2110151909647982e-06,
+      "loss": 0.4993,
+      "step": 32149
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.210608445627255e-06,
+      "loss": 0.8911,
+      "step": 32150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.210201764207144e-06,
+      "loss": 0.6873,
+      "step": 32151
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2097951467074244e-06,
+      "loss": 0.7006,
+      "step": 32152
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2093885931310468e-06,
+      "loss": 1.0027,
+      "step": 32153
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2089821034809713e-06,
+      "loss": 0.8712,
+      "step": 32154
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2085756777601555e-06,
+      "loss": 0.9084,
+      "step": 32155
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2081693159715535e-06,
+      "loss": 0.7037,
+      "step": 32156
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.20776301811812e-06,
+      "loss": 0.7164,
+      "step": 32157
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.207356784202811e-06,
+      "loss": 0.8688,
+      "step": 32158
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.206950614228579e-06,
+      "loss": 0.6194,
+      "step": 32159
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2065445081983795e-06,
+      "loss": 0.7074,
+      "step": 32160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2061384661151632e-06,
+      "loss": 0.7863,
+      "step": 32161
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2057324879818877e-06,
+      "loss": 0.8578,
+      "step": 32162
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2053265738015008e-06,
+      "loss": 0.9129,
+      "step": 32163
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2049207235769544e-06,
+      "loss": 0.8998,
+      "step": 32164
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.204514937311202e-06,
+      "loss": 0.9247,
+      "step": 32165
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.204109215007192e-06,
+      "loss": 0.8522,
+      "step": 32166
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2037035566678778e-06,
+      "loss": 0.6533,
+      "step": 32167
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2032979622962083e-06,
+      "loss": 1.0442,
+      "step": 32168
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.202892431895133e-06,
+      "loss": 0.8927,
+      "step": 32169
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.202486965467602e-06,
+      "loss": 0.9971,
+      "step": 32170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2020815630165606e-06,
+      "loss": 0.8087,
+      "step": 32171
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2016762245449609e-06,
+      "loss": 0.7742,
+      "step": 32172
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2012709500557474e-06,
+      "loss": 0.7569,
+      "step": 32173
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2008657395518696e-06,
+      "loss": 0.9623,
+      "step": 32174
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2004605930362724e-06,
+      "loss": 0.7152,
+      "step": 32175
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.2000555105119039e-06,
+      "loss": 0.9619,
+      "step": 32176
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1996504919817087e-06,
+      "loss": 0.9831,
+      "step": 32177
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1992455374486312e-06,
+      "loss": 0.8027,
+      "step": 32178
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1988406469156167e-06,
+      "loss": 0.8266,
+      "step": 32179
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1984358203856117e-06,
+      "loss": 1.0006,
+      "step": 32180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1980310578615595e-06,
+      "loss": 0.8777,
+      "step": 32181
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1976263593464043e-06,
+      "loss": 0.6601,
+      "step": 32182
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1972217248430862e-06,
+      "loss": 0.6335,
+      "step": 32183
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1968171543545471e-06,
+      "loss": 0.6753,
+      "step": 32184
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1964126478837323e-06,
+      "loss": 0.6312,
+      "step": 32185
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1960082054335831e-06,
+      "loss": 0.7854,
+      "step": 32186
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1956038270070402e-06,
+      "loss": 0.8782,
+      "step": 32187
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1951995126070437e-06,
+      "loss": 0.596,
+      "step": 32188
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1947952622365343e-06,
+      "loss": 0.5877,
+      "step": 32189
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.194391075898451e-06,
+      "loss": 1.0355,
+      "step": 32190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1939869535957315e-06,
+      "loss": 0.9313,
+      "step": 32191
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1935828953313212e-06,
+      "loss": 0.6069,
+      "step": 32192
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1931789011081529e-06,
+      "loss": 0.7459,
+      "step": 32193
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1927749709291647e-06,
+      "loss": 0.6643,
+      "step": 32194
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1923711047972941e-06,
+      "loss": 0.746,
+      "step": 32195
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1919673027154798e-06,
+      "loss": 0.4263,
+      "step": 32196
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1915635646866552e-06,
+      "loss": 0.8899,
+      "step": 32197
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.191159890713761e-06,
+      "loss": 0.9606,
+      "step": 32198
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1907562807997297e-06,
+      "loss": 0.9436,
+      "step": 32199
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1903527349474975e-06,
+      "loss": 0.9057,
+      "step": 32200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1899492531599976e-06,
+      "loss": 0.8247,
+      "step": 32201
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1895458354401657e-06,
+      "loss": 0.7991,
+      "step": 32202
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1891424817909348e-06,
+      "loss": 1.0271,
+      "step": 32203
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.188739192215238e-06,
+      "loss": 0.5973,
+      "step": 32204
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1883359667160087e-06,
+      "loss": 0.7559,
+      "step": 32205
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1879328052961792e-06,
+      "loss": 0.6377,
+      "step": 32206
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.18752970795868e-06,
+      "loss": 0.8089,
+      "step": 32207
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1871266747064435e-06,
+      "loss": 0.7575,
+      "step": 32208
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1867237055423985e-06,
+      "loss": 0.702,
+      "step": 32209
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1863208004694793e-06,
+      "loss": 0.9794,
+      "step": 32210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1859179594906146e-06,
+      "loss": 0.7458,
+      "step": 32211
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1855151826087341e-06,
+      "loss": 0.713,
+      "step": 32212
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1851124698267646e-06,
+      "loss": 0.8097,
+      "step": 32213
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.184709821147637e-06,
+      "loss": 0.7428,
+      "step": 32214
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1843072365742759e-06,
+      "loss": 0.6739,
+      "step": 32215
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1839047161096128e-06,
+      "loss": 0.8824,
+      "step": 32216
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1835022597565737e-06,
+      "loss": 0.9622,
+      "step": 32217
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.183099867518086e-06,
+      "loss": 0.7308,
+      "step": 32218
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1826975393970753e-06,
+      "loss": 0.9704,
+      "step": 32219
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1822952753964667e-06,
+      "loss": 0.5803,
+      "step": 32220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.181893075519186e-06,
+      "loss": 0.8478,
+      "step": 32221
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1814909397681584e-06,
+      "loss": 0.8595,
+      "step": 32222
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1810888681463072e-06,
+      "loss": 0.7907,
+      "step": 32223
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.180686860656558e-06,
+      "loss": 0.9837,
+      "step": 32224
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1802849173018328e-06,
+      "loss": 1.058,
+      "step": 32225
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1798830380850546e-06,
+      "loss": 1.0156,
+      "step": 32226
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1794812230091457e-06,
+      "loss": 0.8361,
+      "step": 32227
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1790794720770293e-06,
+      "loss": 0.749,
+      "step": 32228
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1786777852916275e-06,
+      "loss": 0.7654,
+      "step": 32229
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1782761626558603e-06,
+      "loss": 0.8378,
+      "step": 32230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1778746041726486e-06,
+      "loss": 0.9203,
+      "step": 32231
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1774731098449144e-06,
+      "loss": 1.089,
+      "step": 32232
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.177071679675572e-06,
+      "loss": 0.7002,
+      "step": 32233
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.176670313667545e-06,
+      "loss": 0.6974,
+      "step": 32234
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1762690118237518e-06,
+      "loss": 0.9301,
+      "step": 32235
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1758677741471104e-06,
+      "loss": 0.7811,
+      "step": 32236
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.175466600640538e-06,
+      "loss": 0.5935,
+      "step": 32237
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1750654913069525e-06,
+      "loss": 1.0203,
+      "step": 32238
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1746644461492707e-06,
+      "loss": 0.6976,
+      "step": 32239
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1742634651704076e-06,
+      "loss": 1.0727,
+      "step": 32240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1738625483732823e-06,
+      "loss": 0.8568,
+      "step": 32241
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1734616957608102e-06,
+      "loss": 0.698,
+      "step": 32242
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1730609073359034e-06,
+      "loss": 1.0671,
+      "step": 32243
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1726601831014782e-06,
+      "loss": 1.1094,
+      "step": 32244
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1722595230604462e-06,
+      "loss": 0.8836,
+      "step": 32245
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1718589272157244e-06,
+      "loss": 0.6659,
+      "step": 32246
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1714583955702263e-06,
+      "loss": 0.9157,
+      "step": 32247
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.171057928126862e-06,
+      "loss": 0.9624,
+      "step": 32248
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1706575248885455e-06,
+      "loss": 0.886,
+      "step": 32249
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1702571858581867e-06,
+      "loss": 0.816,
+      "step": 32250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1698569110386993e-06,
+      "loss": 0.5992,
+      "step": 32251
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1694567004329927e-06,
+      "loss": 0.7436,
+      "step": 32252
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.169056554043978e-06,
+      "loss": 0.9375,
+      "step": 32253
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1686564718745642e-06,
+      "loss": 0.6733,
+      "step": 32254
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.168256453927662e-06,
+      "loss": 0.8983,
+      "step": 32255
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1678565002061782e-06,
+      "loss": 0.8152,
+      "step": 32256
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1674566107130236e-06,
+      "loss": 0.8925,
+      "step": 32257
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.167056785451104e-06,
+      "loss": 0.6083,
+      "step": 32258
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1666570244233288e-06,
+      "loss": 0.8867,
+      "step": 32259
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1662573276326062e-06,
+      "loss": 0.8282,
+      "step": 32260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1658576950818402e-06,
+      "loss": 0.5131,
+      "step": 32261
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1654581267739406e-06,
+      "loss": 0.7409,
+      "step": 32262
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1650586227118055e-06,
+      "loss": 0.932,
+      "step": 32263
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.164659182898349e-06,
+      "loss": 0.8853,
+      "step": 32264
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1642598073364707e-06,
+      "loss": 0.5999,
+      "step": 32265
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1638604960290768e-06,
+      "loss": 0.7484,
+      "step": 32266
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1634612489790697e-06,
+      "loss": 0.8508,
+      "step": 32267
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1630620661893554e-06,
+      "loss": 0.6544,
+      "step": 32268
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1626629476628336e-06,
+      "loss": 0.9868,
+      "step": 32269
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1622638934024066e-06,
+      "loss": 1.1048,
+      "step": 32270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1618649034109796e-06,
+      "loss": 0.7564,
+      "step": 32271
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.161465977691456e-06,
+      "loss": 1.0469,
+      "step": 32272
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1610671162467302e-06,
+      "loss": 1.0963,
+      "step": 32273
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1606683190797053e-06,
+      "loss": 0.861,
+      "step": 32274
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1602695861932833e-06,
+      "loss": 0.7051,
+      "step": 32275
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1598709175903588e-06,
+      "loss": 0.8393,
+      "step": 32276
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1594723132738383e-06,
+      "loss": 0.7625,
+      "step": 32277
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1590737732466163e-06,
+      "loss": 0.9577,
+      "step": 32278
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.158675297511591e-06,
+      "loss": 0.754,
+      "step": 32279
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.158276886071662e-06,
+      "loss": 0.8631,
+      "step": 32280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1578785389297241e-06,
+      "loss": 0.8028,
+      "step": 32281
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1574802560886755e-06,
+      "loss": 1.0696,
+      "step": 32282
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1570820375514136e-06,
+      "loss": 1.0866,
+      "step": 32283
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1566838833208317e-06,
+      "loss": 0.8876,
+      "step": 32284
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1562857933998274e-06,
+      "loss": 0.8181,
+      "step": 32285
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.155887767791295e-06,
+      "loss": 0.7253,
+      "step": 32286
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1554898064981279e-06,
+      "loss": 0.8948,
+      "step": 32287
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1550919095232204e-06,
+      "loss": 0.8342,
+      "step": 32288
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1546940768694681e-06,
+      "loss": 1.0252,
+      "step": 32289
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1542963085397619e-06,
+      "loss": 0.8001,
+      "step": 32290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.153898604536996e-06,
+      "loss": 0.9036,
+      "step": 32291
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1535009648640638e-06,
+      "loss": 0.7262,
+      "step": 32292
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.153103389523853e-06,
+      "loss": 1.0332,
+      "step": 32293
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1527058785192546e-06,
+      "loss": 0.9147,
+      "step": 32294
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1523084318531641e-06,
+      "loss": 0.7399,
+      "step": 32295
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.151911049528469e-06,
+      "loss": 0.5875,
+      "step": 32296
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1515137315480585e-06,
+      "loss": 0.9938,
+      "step": 32297
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1511164779148231e-06,
+      "loss": 0.9644,
+      "step": 32298
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.150719288631652e-06,
+      "loss": 0.6459,
+      "step": 32299
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1503221637014327e-06,
+      "loss": 0.8731,
+      "step": 32300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1499251031270508e-06,
+      "loss": 0.6286,
+      "step": 32301
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1495281069114005e-06,
+      "loss": 0.6805,
+      "step": 32302
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.149131175057363e-06,
+      "loss": 0.8256,
+      "step": 32303
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1487343075678258e-06,
+      "loss": 0.8896,
+      "step": 32304
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1483375044456767e-06,
+      "loss": 0.8447,
+      "step": 32305
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1479407656937969e-06,
+      "loss": 0.7268,
+      "step": 32306
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1475440913150782e-06,
+      "loss": 1.0295,
+      "step": 32307
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1471474813124005e-06,
+      "loss": 1.0131,
+      "step": 32308
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1467509356886509e-06,
+      "loss": 0.4519,
+      "step": 32309
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1463544544467109e-06,
+      "loss": 0.5184,
+      "step": 32310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.145958037589464e-06,
+      "loss": 0.6588,
+      "step": 32311
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1455616851197937e-06,
+      "loss": 0.7129,
+      "step": 32312
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1451653970405828e-06,
+      "loss": 0.7013,
+      "step": 32313
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1447691733547117e-06,
+      "loss": 0.8265,
+      "step": 32314
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1443730140650633e-06,
+      "loss": 0.6917,
+      "step": 32315
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1439769191745165e-06,
+      "loss": 0.7059,
+      "step": 32316
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1435808886859545e-06,
+      "loss": 0.7767,
+      "step": 32317
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.143184922602254e-06,
+      "loss": 0.692,
+      "step": 32318
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.142789020926296e-06,
+      "loss": 0.7234,
+      "step": 32319
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1423931836609604e-06,
+      "loss": 0.9931,
+      "step": 32320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1419974108091259e-06,
+      "loss": 0.9549,
+      "step": 32321
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1416017023736724e-06,
+      "loss": 0.8011,
+      "step": 32322
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1412060583574724e-06,
+      "loss": 0.8051,
+      "step": 32323
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1408104787634033e-06,
+      "loss": 0.9287,
+      "step": 32324
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1404149635943462e-06,
+      "loss": 0.7611,
+      "step": 32325
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1400195128531766e-06,
+      "loss": 1.206,
+      "step": 32326
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1396241265427676e-06,
+      "loss": 0.8683,
+      "step": 32327
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.139228804665996e-06,
+      "loss": 0.8001,
+      "step": 32328
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1388335472257373e-06,
+      "loss": 0.7992,
+      "step": 32329
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1384383542248646e-06,
+      "loss": 0.6387,
+      "step": 32330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.13804322566625e-06,
+      "loss": 0.8963,
+      "step": 32331
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1376481615527735e-06,
+      "loss": 0.9385,
+      "step": 32332
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1372531618873017e-06,
+      "loss": 0.877,
+      "step": 32333
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.136858226672709e-06,
+      "loss": 0.5548,
+      "step": 32334
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1364633559118676e-06,
+      "loss": 0.8149,
+      "step": 32335
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1360685496076485e-06,
+      "loss": 0.7417,
+      "step": 32336
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1356738077629215e-06,
+      "loss": 0.7649,
+      "step": 32337
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.135279130380561e-06,
+      "loss": 0.7621,
+      "step": 32338
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.134884517463436e-06,
+      "loss": 0.7076,
+      "step": 32339
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.134489969014414e-06,
+      "loss": 0.607,
+      "step": 32340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.134095485036365e-06,
+      "loss": 0.4925,
+      "step": 32341
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.133701065532159e-06,
+      "loss": 0.9754,
+      "step": 32342
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1333067105046635e-06,
+      "loss": 0.6734,
+      "step": 32343
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1329124199567464e-06,
+      "loss": 0.7764,
+      "step": 32344
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1325181938912743e-06,
+      "loss": 0.7342,
+      "step": 32345
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1321240323111148e-06,
+      "loss": 1.107,
+      "step": 32346
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1317299352191346e-06,
+      "loss": 0.8472,
+      "step": 32347
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.131335902618198e-06,
+      "loss": 0.5235,
+      "step": 32348
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1309419345111705e-06,
+      "loss": 0.9731,
+      "step": 32349
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1305480309009199e-06,
+      "loss": 0.8216,
+      "step": 32350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1301541917903092e-06,
+      "loss": 0.7939,
+      "step": 32351
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.129760417182204e-06,
+      "loss": 0.8633,
+      "step": 32352
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1293667070794633e-06,
+      "loss": 0.8605,
+      "step": 32353
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1289730614849536e-06,
+      "loss": 0.8346,
+      "step": 32354
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1285794804015349e-06,
+      "loss": 0.8816,
+      "step": 32355
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1281859638320736e-06,
+      "loss": 0.6854,
+      "step": 32356
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1277925117794286e-06,
+      "loss": 0.6327,
+      "step": 32357
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1273991242464622e-06,
+      "loss": 0.7792,
+      "step": 32358
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1270058012360352e-06,
+      "loss": 0.847,
+      "step": 32359
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1266125427510066e-06,
+      "loss": 0.7923,
+      "step": 32360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1262193487942363e-06,
+      "loss": 0.6614,
+      "step": 32361
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1258262193685853e-06,
+      "loss": 0.9605,
+      "step": 32362
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1254331544769116e-06,
+      "loss": 0.9999,
+      "step": 32363
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1250401541220723e-06,
+      "loss": 0.6222,
+      "step": 32364
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1246472183069279e-06,
+      "loss": 0.8259,
+      "step": 32365
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1242543470343337e-06,
+      "loss": 0.5621,
+      "step": 32366
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1238615403071473e-06,
+      "loss": 0.6631,
+      "step": 32367
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1234687981282245e-06,
+      "loss": 0.8944,
+      "step": 32368
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1230761205004238e-06,
+      "loss": 0.4595,
+      "step": 32369
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1226835074265985e-06,
+      "loss": 0.705,
+      "step": 32370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1222909589096054e-06,
+      "loss": 0.8457,
+      "step": 32371
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1218984749522999e-06,
+      "loss": 0.6798,
+      "step": 32372
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1215060555575309e-06,
+      "loss": 1.0221,
+      "step": 32373
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1211137007281581e-06,
+      "loss": 0.9443,
+      "step": 32374
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1207214104670316e-06,
+      "loss": 0.5828,
+      "step": 32375
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1203291847770059e-06,
+      "loss": 0.7862,
+      "step": 32376
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.119937023660932e-06,
+      "loss": 0.6917,
+      "step": 32377
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.119544927121662e-06,
+      "loss": 0.9402,
+      "step": 32378
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.119152895162048e-06,
+      "loss": 0.836,
+      "step": 32379
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1187609277849376e-06,
+      "loss": 0.8402,
+      "step": 32380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1183690249931866e-06,
+      "loss": 0.7285,
+      "step": 32381
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1179771867896439e-06,
+      "loss": 0.7712,
+      "step": 32382
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1175854131771557e-06,
+      "loss": 0.7044,
+      "step": 32383
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1171937041585723e-06,
+      "loss": 0.5464,
+      "step": 32384
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1168020597367435e-06,
+      "loss": 0.9964,
+      "step": 32385
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1164104799145138e-06,
+      "loss": 1.0539,
+      "step": 32386
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1160189646947362e-06,
+      "loss": 0.6732,
+      "step": 32387
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.115627514080254e-06,
+      "loss": 1.0016,
+      "step": 32388
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1152361280739167e-06,
+      "loss": 1.0549,
+      "step": 32389
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1148448066785678e-06,
+      "loss": 0.6872,
+      "step": 32390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1144535498970543e-06,
+      "loss": 0.781,
+      "step": 32391
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1140623577322219e-06,
+      "loss": 0.7928,
+      "step": 32392
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1136712301869146e-06,
+      "loss": 0.9388,
+      "step": 32393
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1132801672639771e-06,
+      "loss": 0.9812,
+      "step": 32394
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1128891689662525e-06,
+      "loss": 0.6929,
+      "step": 32395
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1124982352965852e-06,
+      "loss": 0.8286,
+      "step": 32396
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1121073662578175e-06,
+      "loss": 0.6552,
+      "step": 32397
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1117165618527904e-06,
+      "loss": 0.7818,
+      "step": 32398
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.111325822084348e-06,
+      "loss": 0.8496,
+      "step": 32399
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1109351469553331e-06,
+      "loss": 0.5454,
+      "step": 32400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1105445364685841e-06,
+      "loss": 1.1,
+      "step": 32401
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1101539906269443e-06,
+      "loss": 0.5252,
+      "step": 32402
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1097635094332504e-06,
+      "loss": 0.8652,
+      "step": 32403
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1093730928903412e-06,
+      "loss": 0.7468,
+      "step": 32404
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1089827410010601e-06,
+      "loss": 0.6104,
+      "step": 32405
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1085924537682435e-06,
+      "loss": 0.7801,
+      "step": 32406
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1082022311947305e-06,
+      "loss": 0.4463,
+      "step": 32407
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1078120732833576e-06,
+      "loss": 0.8905,
+      "step": 32408
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1074219800369633e-06,
+      "loss": 0.537,
+      "step": 32409
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1070319514583839e-06,
+      "loss": 0.5584,
+      "step": 32410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1066419875504531e-06,
+      "loss": 1.1231,
+      "step": 32411
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1062520883160144e-06,
+      "loss": 0.6935,
+      "step": 32412
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1058622537578955e-06,
+      "loss": 0.8251,
+      "step": 32413
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1054724838789344e-06,
+      "loss": 0.8688,
+      "step": 32414
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.105082778681964e-06,
+      "loss": 0.7548,
+      "step": 32415
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.104693138169818e-06,
+      "loss": 0.7517,
+      "step": 32416
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1043035623453336e-06,
+      "loss": 1.0847,
+      "step": 32417
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.103914051211341e-06,
+      "loss": 0.667,
+      "step": 32418
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1035246047706738e-06,
+      "loss": 0.755,
+      "step": 32419
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1031352230261638e-06,
+      "loss": 0.8389,
+      "step": 32420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1027459059806422e-06,
+      "loss": 0.7143,
+      "step": 32421
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1023566536369402e-06,
+      "loss": 0.877,
+      "step": 32422
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1019674659978885e-06,
+      "loss": 0.8085,
+      "step": 32423
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1015783430663175e-06,
+      "loss": 0.7678,
+      "step": 32424
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.101189284845058e-06,
+      "loss": 0.7472,
+      "step": 32425
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1008002913369364e-06,
+      "loss": 0.8678,
+      "step": 32426
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1004113625447843e-06,
+      "loss": 1.0338,
+      "step": 32427
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1000224984714292e-06,
+      "loss": 0.9438,
+      "step": 32428
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0996336991196965e-06,
+      "loss": 0.6102,
+      "step": 32429
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0992449644924186e-06,
+      "loss": 0.939,
+      "step": 32430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0988562945924187e-06,
+      "loss": 0.9608,
+      "step": 32431
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0984676894225277e-06,
+      "loss": 0.724,
+      "step": 32432
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0980791489855646e-06,
+      "loss": 0.954,
+      "step": 32433
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0976906732843573e-06,
+      "loss": 1.0396,
+      "step": 32434
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0973022623217334e-06,
+      "loss": 0.7802,
+      "step": 32435
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.096913916100517e-06,
+      "loss": 0.614,
+      "step": 32436
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0965256346235308e-06,
+      "loss": 0.8718,
+      "step": 32437
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.096137417893598e-06,
+      "loss": 1.1787,
+      "step": 32438
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0957492659135437e-06,
+      "loss": 0.9338,
+      "step": 32439
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0953611786861896e-06,
+      "loss": 0.8557,
+      "step": 32440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0949731562143552e-06,
+      "loss": 0.925,
+      "step": 32441
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0945851985008694e-06,
+      "loss": 1.0911,
+      "step": 32442
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0941973055485467e-06,
+      "loss": 0.8676,
+      "step": 32443
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0938094773602103e-06,
+      "loss": 0.9204,
+      "step": 32444
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0934217139386805e-06,
+      "loss": 0.8413,
+      "step": 32445
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.093034015286777e-06,
+      "loss": 0.7748,
+      "step": 32446
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0926463814073173e-06,
+      "loss": 0.8086,
+      "step": 32447
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.092258812303124e-06,
+      "loss": 0.817,
+      "step": 32448
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0918713079770138e-06,
+      "loss": 0.8845,
+      "step": 32449
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.091483868431804e-06,
+      "loss": 0.6717,
+      "step": 32450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0910964936703138e-06,
+      "loss": 0.8686,
+      "step": 32451
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0907091836953587e-06,
+      "loss": 1.0833,
+      "step": 32452
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0903219385097564e-06,
+      "loss": 0.9637,
+      "step": 32453
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0899347581163222e-06,
+      "loss": 1.2215,
+      "step": 32454
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0895476425178708e-06,
+      "loss": 0.9249,
+      "step": 32455
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0891605917172198e-06,
+      "loss": 0.6393,
+      "step": 32456
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0887736057171817e-06,
+      "loss": 0.7583,
+      "step": 32457
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0883866845205714e-06,
+      "loss": 0.6452,
+      "step": 32458
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0879998281302017e-06,
+      "loss": 0.943,
+      "step": 32459
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.087613036548888e-06,
+      "loss": 0.5496,
+      "step": 32460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0872263097794444e-06,
+      "loss": 0.9636,
+      "step": 32461
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.086839647824679e-06,
+      "loss": 0.685,
+      "step": 32462
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.086453050687405e-06,
+      "loss": 0.9023,
+      "step": 32463
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0860665183704344e-06,
+      "loss": 0.8347,
+      "step": 32464
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0856800508765763e-06,
+      "loss": 0.9903,
+      "step": 32465
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.085293648208644e-06,
+      "loss": 0.7142,
+      "step": 32466
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0849073103694474e-06,
+      "loss": 0.8326,
+      "step": 32467
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0845210373617942e-06,
+      "loss": 0.81,
+      "step": 32468
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0841348291884934e-06,
+      "loss": 0.9594,
+      "step": 32469
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0837486858523548e-06,
+      "loss": 1.0404,
+      "step": 32470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0833626073561865e-06,
+      "loss": 0.4559,
+      "step": 32471
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.0829765937027948e-06,
+      "loss": 0.7301,
+      "step": 32472
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0825906448949875e-06,
+      "loss": 1.0364,
+      "step": 32473
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0822047609355713e-06,
+      "loss": 0.4789,
+      "step": 32474
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0818189418273527e-06,
+      "loss": 0.7394,
+      "step": 32475
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0814331875731365e-06,
+      "loss": 0.6086,
+      "step": 32476
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0810474981757268e-06,
+      "loss": 0.7896,
+      "step": 32477
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0806618736379327e-06,
+      "loss": 1.0108,
+      "step": 32478
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.080276313962555e-06,
+      "loss": 0.7772,
+      "step": 32479
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0798908191523993e-06,
+      "loss": 0.7256,
+      "step": 32480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0795053892102691e-06,
+      "loss": 0.8111,
+      "step": 32481
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0791200241389655e-06,
+      "loss": 0.9348,
+      "step": 32482
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0787347239412893e-06,
+      "loss": 0.59,
+      "step": 32483
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0783494886200464e-06,
+      "loss": 0.5998,
+      "step": 32484
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0779643181780364e-06,
+      "loss": 0.7563,
+      "step": 32485
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0775792126180618e-06,
+      "loss": 0.7574,
+      "step": 32486
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0771941719429213e-06,
+      "loss": 0.8594,
+      "step": 32487
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.076809196155416e-06,
+      "loss": 0.6915,
+      "step": 32488
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0764242852583451e-06,
+      "loss": 0.9555,
+      "step": 32489
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0760394392545058e-06,
+      "loss": 0.8077,
+      "step": 32490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.075654658146702e-06,
+      "loss": 0.9,
+      "step": 32491
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0752699419377267e-06,
+      "loss": 0.7701,
+      "step": 32492
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0748852906303798e-06,
+      "loss": 0.6867,
+      "step": 32493
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.074500704227458e-06,
+      "loss": 0.6438,
+      "step": 32494
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.074116182731757e-06,
+      "loss": 0.6298,
+      "step": 32495
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0737317261460756e-06,
+      "loss": 0.6642,
+      "step": 32496
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0733473344732082e-06,
+      "loss": 0.62,
+      "step": 32497
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0729630077159503e-06,
+      "loss": 0.7859,
+      "step": 32498
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0725787458770975e-06,
+      "loss": 0.6816,
+      "step": 32499
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.072194548959442e-06,
+      "loss": 0.7542,
+      "step": 32500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0718104169657805e-06,
+      "loss": 1.1645,
+      "step": 32501
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0714263498989042e-06,
+      "loss": 0.8237,
+      "step": 32502
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0710423477616084e-06,
+      "loss": 0.7595,
+      "step": 32503
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.070658410556683e-06,
+      "loss": 0.6999,
+      "step": 32504
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0702745382869207e-06,
+      "loss": 0.7065,
+      "step": 32505
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0698907309551143e-06,
+      "loss": 0.553,
+      "step": 32506
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.069506988564054e-06,
+      "loss": 1.0188,
+      "step": 32507
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0691233111165289e-06,
+      "loss": 0.7239,
+      "step": 32508
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.068739698615333e-06,
+      "loss": 0.9585,
+      "step": 32509
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0683561510632533e-06,
+      "loss": 0.7626,
+      "step": 32510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.067972668463082e-06,
+      "loss": 0.8541,
+      "step": 32511
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0675892508176023e-06,
+      "loss": 0.7539,
+      "step": 32512
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.067205898129604e-06,
+      "loss": 0.6751,
+      "step": 32513
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0668226104018786e-06,
+      "loss": 1.0858,
+      "step": 32514
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0664393876372114e-06,
+      "loss": 0.7528,
+      "step": 32515
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0660562298383892e-06,
+      "loss": 0.8438,
+      "step": 32516
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0656731370081975e-06,
+      "loss": 0.9626,
+      "step": 32517
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0652901091494228e-06,
+      "loss": 0.733,
+      "step": 32518
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0649071462648507e-06,
+      "loss": 0.6789,
+      "step": 32519
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.064524248357265e-06,
+      "loss": 0.7627,
+      "step": 32520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0641414154294539e-06,
+      "loss": 0.8687,
+      "step": 32521
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0637586474841976e-06,
+      "loss": 0.8179,
+      "step": 32522
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0633759445242808e-06,
+      "loss": 0.7978,
+      "step": 32523
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0629933065524856e-06,
+      "loss": 0.7133,
+      "step": 32524
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0626107335715963e-06,
+      "loss": 0.7646,
+      "step": 32525
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.062228225584392e-06,
+      "loss": 0.8744,
+      "step": 32526
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.061845782593659e-06,
+      "loss": 0.9121,
+      "step": 32527
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0614634046021754e-06,
+      "loss": 0.8179,
+      "step": 32528
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0610810916127223e-06,
+      "loss": 1.1009,
+      "step": 32529
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0606988436280806e-06,
+      "loss": 0.5936,
+      "step": 32530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0603166606510296e-06,
+      "loss": 0.847,
+      "step": 32531
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0599345426843488e-06,
+      "loss": 0.7281,
+      "step": 32532
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0595524897308163e-06,
+      "loss": 0.9466,
+      "step": 32533
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0591705017932118e-06,
+      "loss": 0.7648,
+      "step": 32534
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0587885788743112e-06,
+      "loss": 0.9057,
+      "step": 32535
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0584067209768934e-06,
+      "loss": 0.6624,
+      "step": 32536
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0580249281037347e-06,
+      "loss": 0.7009,
+      "step": 32537
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0576432002576099e-06,
+      "loss": 0.8895,
+      "step": 32538
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0572615374412986e-06,
+      "loss": 0.9108,
+      "step": 32539
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0568799396575747e-06,
+      "loss": 0.6581,
+      "step": 32540
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0564984069092154e-06,
+      "loss": 1.0332,
+      "step": 32541
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0561169391989901e-06,
+      "loss": 0.5586,
+      "step": 32542
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0557355365296752e-06,
+      "loss": 0.9245,
+      "step": 32543
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.055354198904044e-06,
+      "loss": 0.7978,
+      "step": 32544
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.054972926324871e-06,
+      "loss": 0.729,
+      "step": 32545
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0545917187949295e-06,
+      "loss": 1.0132,
+      "step": 32546
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0542105763169896e-06,
+      "loss": 0.7101,
+      "step": 32547
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0538294988938235e-06,
+      "loss": 0.7064,
+      "step": 32548
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0534484865282035e-06,
+      "loss": 0.9967,
+      "step": 32549
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0530675392228995e-06,
+      "loss": 1.112,
+      "step": 32550
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0526866569806816e-06,
+      "loss": 0.811,
+      "step": 32551
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0523058398043196e-06,
+      "loss": 0.9,
+      "step": 32552
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0519250876965837e-06,
+      "loss": 0.7072,
+      "step": 32553
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0515444006602426e-06,
+      "loss": 0.8233,
+      "step": 32554
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0511637786980632e-06,
+      "loss": 0.7658,
+      "step": 32555
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0507832218128134e-06,
+      "loss": 0.8387,
+      "step": 32556
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0504027300072638e-06,
+      "loss": 0.6614,
+      "step": 32557
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0500223032841783e-06,
+      "loss": 0.5589,
+      "step": 32558
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0496419416463254e-06,
+      "loss": 0.8686,
+      "step": 32559
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0492616450964698e-06,
+      "loss": 0.664,
+      "step": 32560
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0488814136373805e-06,
+      "loss": 0.7981,
+      "step": 32561
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0485012472718147e-06,
+      "loss": 0.6458,
+      "step": 32562
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0481211460025443e-06,
+      "loss": 1.4651,
+      "step": 32563
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0477411098323298e-06,
+      "loss": 0.8451,
+      "step": 32564
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.047361138763937e-06,
+      "loss": 0.9996,
+      "step": 32565
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0469812328001283e-06,
+      "loss": 0.7649,
+      "step": 32566
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0466013919436668e-06,
+      "loss": 0.6861,
+      "step": 32567
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0462216161973137e-06,
+      "loss": 0.6312,
+      "step": 32568
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.045841905563829e-06,
+      "loss": 0.639,
+      "step": 32569
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0454622600459785e-06,
+      "loss": 0.4349,
+      "step": 32570
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.045082679646522e-06,
+      "loss": 1.0217,
+      "step": 32571
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0447031643682181e-06,
+      "loss": 0.721,
+      "step": 32572
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0443237142138262e-06,
+      "loss": 1.3377,
+      "step": 32573
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.043944329186105e-06,
+      "loss": 0.6671,
+      "step": 32574
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.043565009287818e-06,
+      "loss": 0.722,
+      "step": 32575
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0431857545217194e-06,
+      "loss": 0.7708,
+      "step": 32576
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0428065648905695e-06,
+      "loss": 0.8895,
+      "step": 32577
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0424274403971257e-06,
+      "loss": 0.8798,
+      "step": 32578
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.042048381044143e-06,
+      "loss": 0.6788,
+      "step": 32579
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0416693868343796e-06,
+      "loss": 1.1153,
+      "step": 32580
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0412904577705907e-06,
+      "loss": 0.6568,
+      "step": 32581
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0409115938555336e-06,
+      "loss": 0.7899,
+      "step": 32582
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0405327950919618e-06,
+      "loss": 0.8948,
+      "step": 32583
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.04015406148263e-06,
+      "loss": 1.1286,
+      "step": 32584
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0397753930302935e-06,
+      "loss": 0.9675,
+      "step": 32585
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0393967897377043e-06,
+      "loss": 0.9832,
+      "step": 32586
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0390182516076152e-06,
+      "loss": 1.0033,
+      "step": 32587
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0386397786427838e-06,
+      "loss": 0.8261,
+      "step": 32588
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0382613708459578e-06,
+      "loss": 0.5916,
+      "step": 32589
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0378830282198905e-06,
+      "loss": 0.6902,
+      "step": 32590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0375047507673342e-06,
+      "loss": 0.7724,
+      "step": 32591
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0371265384910357e-06,
+      "loss": 1.1056,
+      "step": 32592
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0367483913937493e-06,
+      "loss": 0.7539,
+      "step": 32593
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0363703094782252e-06,
+      "loss": 0.8462,
+      "step": 32594
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.035992292747211e-06,
+      "loss": 0.9027,
+      "step": 32595
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0356143412034557e-06,
+      "loss": 0.6708,
+      "step": 32596
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0352364548497095e-06,
+      "loss": 0.5515,
+      "step": 32597
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.034858633688719e-06,
+      "loss": 0.6375,
+      "step": 32598
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0344808777232295e-06,
+      "loss": 0.8624,
+      "step": 32599
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0341031869559937e-06,
+      "loss": 0.6402,
+      "step": 32600
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0337255613897557e-06,
+      "loss": 1.0014,
+      "step": 32601
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0333480010272601e-06,
+      "loss": 0.9372,
+      "step": 32602
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0329705058712536e-06,
+      "loss": 0.7236,
+      "step": 32603
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0325930759244806e-06,
+      "loss": 0.7433,
+      "step": 32604
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0322157111896858e-06,
+      "loss": 0.779,
+      "step": 32605
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0318384116696146e-06,
+      "loss": 0.7251,
+      "step": 32606
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0314611773670114e-06,
+      "loss": 0.9178,
+      "step": 32607
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0310840082846185e-06,
+      "loss": 0.8778,
+      "step": 32608
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0307069044251783e-06,
+      "loss": 0.9467,
+      "step": 32609
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.030329865791434e-06,
+      "loss": 0.6217,
+      "step": 32610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0299528923861258e-06,
+      "loss": 0.9296,
+      "step": 32611
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.029575984211998e-06,
+      "loss": 0.9276,
+      "step": 32612
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0291991412717883e-06,
+      "loss": 0.8264,
+      "step": 32613
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0288223635682394e-06,
+      "loss": 0.7103,
+      "step": 32614
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0284456511040896e-06,
+      "loss": 0.7311,
+      "step": 32615
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0280690038820807e-06,
+      "loss": 0.6808,
+      "step": 32616
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0276924219049478e-06,
+      "loss": 0.7881,
+      "step": 32617
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0273159051754333e-06,
+      "loss": 0.7152,
+      "step": 32618
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.026939453696274e-06,
+      "loss": 0.9538,
+      "step": 32619
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0265630674702077e-06,
+      "loss": 1.0026,
+      "step": 32620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0261867464999741e-06,
+      "loss": 0.6447,
+      "step": 32621
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0258104907883038e-06,
+      "loss": 0.522,
+      "step": 32622
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0254343003379353e-06,
+      "loss": 0.6448,
+      "step": 32623
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0250581751516053e-06,
+      "loss": 0.9111,
+      "step": 32624
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0246821152320507e-06,
+      "loss": 1.0072,
+      "step": 32625
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0243061205820038e-06,
+      "loss": 0.8069,
+      "step": 32626
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0239301912042e-06,
+      "loss": 0.7973,
+      "step": 32627
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0235543271013715e-06,
+      "loss": 0.7931,
+      "step": 32628
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0231785282762541e-06,
+      "loss": 0.4269,
+      "step": 32629
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0228027947315766e-06,
+      "loss": 0.6836,
+      "step": 32630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.022427126470078e-06,
+      "loss": 0.8538,
+      "step": 32631
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0220515234944838e-06,
+      "loss": 0.7085,
+      "step": 32632
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0216759858075287e-06,
+      "loss": 1.042,
+      "step": 32633
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0213005134119425e-06,
+      "loss": 0.6118,
+      "step": 32634
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0209251063104542e-06,
+      "loss": 0.5387,
+      "step": 32635
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0205497645057972e-06,
+      "loss": 1.0039,
+      "step": 32636
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0201744880007003e-06,
+      "loss": 0.8742,
+      "step": 32637
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0197992767978916e-06,
+      "loss": 0.9482,
+      "step": 32638
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0194241309000985e-06,
+      "loss": 0.8373,
+      "step": 32639
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0190490503100515e-06,
+      "loss": 0.6864,
+      "step": 32640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.018674035030477e-06,
+      "loss": 0.715,
+      "step": 32641
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0182990850641018e-06,
+      "loss": 0.6455,
+      "step": 32642
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0179242004136536e-06,
+      "loss": 0.9571,
+      "step": 32643
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0175493810818582e-06,
+      "loss": 0.9589,
+      "step": 32644
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.017174627071441e-06,
+      "loss": 0.8222,
+      "step": 32645
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0167999383851268e-06,
+      "loss": 0.8237,
+      "step": 32646
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.016425315025642e-06,
+      "loss": 0.7608,
+      "step": 32647
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0160507569957067e-06,
+      "loss": 0.801,
+      "step": 32648
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0156762642980511e-06,
+      "loss": 0.6733,
+      "step": 32649
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0153018369353951e-06,
+      "loss": 0.8714,
+      "step": 32650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0149274749104632e-06,
+      "loss": 0.9163,
+      "step": 32651
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0145531782259754e-06,
+      "loss": 1.0348,
+      "step": 32652
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0141789468846518e-06,
+      "loss": 0.658,
+      "step": 32653
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0138047808892193e-06,
+      "loss": 0.8217,
+      "step": 32654
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0134306802423965e-06,
+      "loss": 0.9526,
+      "step": 32655
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0130566449469038e-06,
+      "loss": 0.8529,
+      "step": 32656
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0126826750054608e-06,
+      "loss": 0.8238,
+      "step": 32657
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0123087704207868e-06,
+      "loss": 0.8404,
+      "step": 32658
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0119349311956028e-06,
+      "loss": 0.5547,
+      "step": 32659
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0115611573326235e-06,
+      "loss": 0.6883,
+      "step": 32660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.011187448834573e-06,
+      "loss": 0.7428,
+      "step": 32661
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0108138057041638e-06,
+      "loss": 0.6712,
+      "step": 32662
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.010440227944115e-06,
+      "loss": 0.8164,
+      "step": 32663
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0100667155571441e-06,
+      "loss": 0.8652,
+      "step": 32664
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0096932685459648e-06,
+      "loss": 0.9991,
+      "step": 32665
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0093198869132937e-06,
+      "loss": 1.0481,
+      "step": 32666
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0089465706618484e-06,
+      "loss": 0.5962,
+      "step": 32667
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0085733197943414e-06,
+      "loss": 0.7762,
+      "step": 32668
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0082001343134884e-06,
+      "loss": 0.8461,
+      "step": 32669
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0078270142220015e-06,
+      "loss": 0.8035,
+      "step": 32670
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0074539595225963e-06,
+      "loss": 0.831,
+      "step": 32671
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.007080970217984e-06,
+      "loss": 0.71,
+      "step": 32672
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.006708046310878e-06,
+      "loss": 0.796,
+      "step": 32673
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0063351878039885e-06,
+      "loss": 1.0521,
+      "step": 32674
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0059623947000297e-06,
+      "loss": 0.8962,
+      "step": 32675
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0055896670017106e-06,
+      "loss": 0.6996,
+      "step": 32676
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0052170047117417e-06,
+      "loss": 0.6758,
+      "step": 32677
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0048444078328324e-06,
+      "loss": 0.6313,
+      "step": 32678
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0044718763676952e-06,
+      "loss": 0.611,
+      "step": 32679
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.004099410319037e-06,
+      "loss": 0.8094,
+      "step": 32680
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.003727009689569e-06,
+      "loss": 0.7966,
+      "step": 32681
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0033546744819955e-06,
+      "loss": 0.7412,
+      "step": 32682
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0029824046990267e-06,
+      "loss": 0.6732,
+      "step": 32683
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0026102003433659e-06,
+      "loss": 0.8653,
+      "step": 32684
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0022380614177251e-06,
+      "loss": 0.7434,
+      "step": 32685
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0018659879248082e-06,
+      "loss": 0.6426,
+      "step": 32686
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0014939798673206e-06,
+      "loss": 0.8433,
+      "step": 32687
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0011220372479679e-06,
+      "loss": 0.8473,
+      "step": 32688
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0007501600694557e-06,
+      "loss": 0.5948,
+      "step": 32689
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0003783483344887e-06,
+      "loss": 0.7925,
+      "step": 32690
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.0000066020457677e-06,
+      "loss": 0.7363,
+      "step": 32691
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.996349212059998e-07,
+      "loss": 0.9934,
+      "step": 32692
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.992633058178847e-07,
+      "loss": 0.6232,
+      "step": 32693
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.988917558841272e-07,
+      "loss": 0.838,
+      "step": 32694
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.985202714074281e-07,
+      "loss": 0.6975,
+      "step": 32695
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.981488523904881e-07,
+      "loss": 0.6018,
+      "step": 32696
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.9777749883601e-07,
+      "loss": 0.6338,
+      "step": 32697
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.974062107466941e-07,
+      "loss": 0.6771,
+      "step": 32698
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.970349881252406e-07,
+      "loss": 0.7616,
+      "step": 32699
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.966638309743481e-07,
+      "loss": 0.9555,
+      "step": 32700
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.96292739296717e-07,
+      "loss": 0.73,
+      "step": 32701
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.959217130950428e-07,
+      "loss": 1.0518,
+      "step": 32702
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.955507523720265e-07,
+      "loss": 0.8255,
+      "step": 32703
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.95179857130365e-07,
+      "loss": 0.3841,
+      "step": 32704
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.948090273727561e-07,
+      "loss": 0.9195,
+      "step": 32705
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.944382631018957e-07,
+      "loss": 0.8283,
+      "step": 32706
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.9406756432048e-07,
+      "loss": 0.4877,
+      "step": 32707
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.93696931031205e-07,
+      "loss": 0.5396,
+      "step": 32708
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.93326363236764e-07,
+      "loss": 0.6854,
+      "step": 32709
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.92955860939857e-07,
+      "loss": 0.5406,
+      "step": 32710
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.925854241431764e-07,
+      "loss": 0.9398,
+      "step": 32711
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.922150528494122e-07,
+      "loss": 0.8748,
+      "step": 32712
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.918447470612614e-07,
+      "loss": 0.5058,
+      "step": 32713
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.914745067814135e-07,
+      "loss": 0.8971,
+      "step": 32714
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.911043320125657e-07,
+      "loss": 0.9081,
+      "step": 32715
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.907342227574068e-07,
+      "loss": 0.811,
+      "step": 32716
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.903641790186303e-07,
+      "loss": 0.9943,
+      "step": 32717
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.899942007989248e-07,
+      "loss": 0.884,
+      "step": 32718
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.896242881009833e-07,
+      "loss": 1.0288,
+      "step": 32719
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.89254440927493e-07,
+      "loss": 0.9267,
+      "step": 32720
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.888846592811462e-07,
+      "loss": 0.8998,
+      "step": 32721
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.885149431646302e-07,
+      "loss": 0.8007,
+      "step": 32722
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.881452925806346e-07,
+      "loss": 0.987,
+      "step": 32723
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.877757075318473e-07,
+      "loss": 0.6995,
+      "step": 32724
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.87406188020955e-07,
+      "loss": 0.7075,
+      "step": 32725
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.87036734050647e-07,
+      "loss": 0.9878,
+      "step": 32726
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.866673456236054e-07,
+      "loss": 0.8786,
+      "step": 32727
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.862980227425222e-07,
+      "loss": 0.9029,
+      "step": 32728
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.859287654100813e-07,
+      "loss": 0.903,
+      "step": 32729
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.85559573628967e-07,
+      "loss": 0.87,
+      "step": 32730
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.85190447401867e-07,
+      "loss": 0.9617,
+      "step": 32731
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.848213867314582e-07,
+      "loss": 0.7942,
+      "step": 32732
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.844523916204318e-07,
+      "loss": 0.8906,
+      "step": 32733
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.84083462071469e-07,
+      "loss": 0.822,
+      "step": 32734
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.83714598087252e-07,
+      "loss": 0.5526,
+      "step": 32735
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.833457996704643e-07,
+      "loss": 0.7069,
+      "step": 32736
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.82977066823787e-07,
+      "loss": 0.677,
+      "step": 32737
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.826083995499014e-07,
+      "loss": 1.0078,
+      "step": 32738
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.822397978514874e-07,
+      "loss": 0.623,
+      "step": 32739
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.818712617312286e-07,
+      "loss": 0.9985,
+      "step": 32740
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.815027911918063e-07,
+      "loss": 0.8815,
+      "step": 32741
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.811343862358936e-07,
+      "loss": 0.9077,
+      "step": 32742
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.807660468661738e-07,
+      "loss": 0.7437,
+      "step": 32743
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.803977730853253e-07,
+      "loss": 0.7894,
+      "step": 32744
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.800295648960245e-07,
+      "loss": 0.8973,
+      "step": 32745
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.796614223009516e-07,
+      "loss": 0.743,
+      "step": 32746
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.792933453027819e-07,
+      "loss": 0.5861,
+      "step": 32747
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.789253339041926e-07,
+      "loss": 0.9761,
+      "step": 32748
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.785573881078614e-07,
+      "loss": 0.8562,
+      "step": 32749
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.781895079164617e-07,
+      "loss": 0.944,
+      "step": 32750
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.7782169333267e-07,
+      "loss": 0.9762,
+      "step": 32751
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.774539443591601e-07,
+      "loss": 1.1025,
+      "step": 32752
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.770862609986087e-07,
+      "loss": 0.9285,
+      "step": 32753
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.767186432536869e-07,
+      "loss": 0.6695,
+      "step": 32754
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.763510911270691e-07,
+      "loss": 0.9526,
+      "step": 32755
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.759836046214277e-07,
+      "loss": 0.8829,
+      "step": 32756
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.756161837394352e-07,
+      "loss": 0.8293,
+      "step": 32757
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.752488284837647e-07,
+      "loss": 0.9297,
+      "step": 32758
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.748815388570876e-07,
+      "loss": 0.6837,
+      "step": 32759
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.74514314862074e-07,
+      "loss": 0.7308,
+      "step": 32760
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.74147156501396e-07,
+      "loss": 0.8844,
+      "step": 32761
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.737800637777206e-07,
+      "loss": 0.8943,
+      "step": 32762
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.734130366937167e-07,
+      "loss": 0.8933,
+      "step": 32763
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.730460752520577e-07,
+      "loss": 0.7335,
+      "step": 32764
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.726791794554102e-07,
+      "loss": 0.7842,
+      "step": 32765
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.723123493064423e-07,
+      "loss": 0.9608,
+      "step": 32766
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.719455848078208e-07,
+      "loss": 0.6871,
+      "step": 32767
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.715788859622143e-07,
+      "loss": 0.8754,
+      "step": 32768
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.712122527722879e-07,
+      "loss": 0.7589,
+      "step": 32769
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.708456852407078e-07,
+      "loss": 0.7491,
+      "step": 32770
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.704791833701433e-07,
+      "loss": 1.0874,
+      "step": 32771
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.701127471632554e-07,
+      "loss": 0.9163,
+      "step": 32772
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.697463766227112e-07,
+      "loss": 0.8334,
+      "step": 32773
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.693800717511725e-07,
+      "loss": 0.6945,
+      "step": 32774
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.690138325513043e-07,
+      "loss": 0.6981,
+      "step": 32775
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.686476590257709e-07,
+      "loss": 0.8517,
+      "step": 32776
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.68281551177236e-07,
+      "loss": 0.6886,
+      "step": 32777
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.679155090083603e-07,
+      "loss": 0.8346,
+      "step": 32778
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.675495325218055e-07,
+      "loss": 0.9376,
+      "step": 32779
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.671836217202334e-07,
+      "loss": 0.789,
+      "step": 32780
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.668177766063059e-07,
+      "loss": 0.493,
+      "step": 32781
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.664519971826813e-07,
+      "loss": 0.7946,
+      "step": 32782
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.660862834520224e-07,
+      "loss": 0.9036,
+      "step": 32783
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.657206354169868e-07,
+      "loss": 0.6922,
+      "step": 32784
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.653550530802335e-07,
+      "loss": 0.7096,
+      "step": 32785
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.649895364444217e-07,
+      "loss": 0.8733,
+      "step": 32786
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.646240855122091e-07,
+      "loss": 0.8527,
+      "step": 32787
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.642587002862524e-07,
+      "loss": 0.9799,
+      "step": 32788
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.638933807692109e-07,
+      "loss": 0.8037,
+      "step": 32789
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.635281269637409e-07,
+      "loss": 0.7889,
+      "step": 32790
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.631629388724994e-07,
+      "loss": 0.5996,
+      "step": 32791
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.627978164981388e-07,
+      "loss": 1.153,
+      "step": 32792
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.624327598433137e-07,
+      "loss": 0.5468,
+      "step": 32793
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.620677689106829e-07,
+      "loss": 0.5774,
+      "step": 32794
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.617028437028996e-07,
+      "loss": 0.9645,
+      "step": 32795
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.613379842226167e-07,
+      "loss": 0.792,
+      "step": 32796
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.609731904724872e-07,
+      "loss": 0.9995,
+      "step": 32797
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.606084624551648e-07,
+      "loss": 0.732,
+      "step": 32798
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.602438001733017e-07,
+      "loss": 0.6784,
+      "step": 32799
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.598792036295479e-07,
+      "loss": 0.8895,
+      "step": 32800
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.595146728265593e-07,
+      "loss": 0.7699,
+      "step": 32801
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.591502077669824e-07,
+      "loss": 0.8645,
+      "step": 32802
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.587858084534695e-07,
+      "loss": 0.7979,
+      "step": 32803
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.584214748886699e-07,
+      "loss": 0.7207,
+      "step": 32804
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.580572070752335e-07,
+      "loss": 0.8041,
+      "step": 32805
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.576930050158072e-07,
+      "loss": 0.5369,
+      "step": 32806
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.573288687130434e-07,
+      "loss": 0.6626,
+      "step": 32807
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.569647981695884e-07,
+      "loss": 0.8715,
+      "step": 32808
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.566007933880894e-07,
+      "loss": 0.8437,
+      "step": 32809
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.56236854371193e-07,
+      "loss": 0.6941,
+      "step": 32810
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.558729811215473e-07,
+      "loss": 0.8283,
+      "step": 32811
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.555091736417976e-07,
+      "loss": 0.8576,
+      "step": 32812
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.551454319345898e-07,
+      "loss": 1.0118,
+      "step": 32813
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.547817560025685e-07,
+      "loss": 0.7465,
+      "step": 32814
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.54418145848378e-07,
+      "loss": 0.7146,
+      "step": 32815
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.540546014746644e-07,
+      "loss": 0.8685,
+      "step": 32816
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.536911228840695e-07,
+      "loss": 0.8903,
+      "step": 32817
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.53327710079236e-07,
+      "loss": 0.8278,
+      "step": 32818
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.529643630628093e-07,
+      "loss": 0.7201,
+      "step": 32819
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.52601081837431e-07,
+      "loss": 0.7563,
+      "step": 32820
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.52237866405743e-07,
+      "loss": 0.8774,
+      "step": 32821
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.518747167703857e-07,
+      "loss": 0.6229,
+      "step": 32822
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.515116329339991e-07,
+      "loss": 0.6845,
+      "step": 32823
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.511486148992232e-07,
+      "loss": 0.9744,
+      "step": 32824
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.507856626687017e-07,
+      "loss": 0.5854,
+      "step": 32825
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.504227762450713e-07,
+      "loss": 0.8806,
+      "step": 32826
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.500599556309709e-07,
+      "loss": 0.6865,
+      "step": 32827
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.496972008290406e-07,
+      "loss": 0.9239,
+      "step": 32828
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.493345118419161e-07,
+      "loss": 0.9861,
+      "step": 32829
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.489718886722377e-07,
+      "loss": 0.7419,
+      "step": 32830
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.486093313226397e-07,
+      "loss": 0.8635,
+      "step": 32831
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.482468397957601e-07,
+      "loss": 0.7624,
+      "step": 32832
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.478844140942345e-07,
+      "loss": 0.7244,
+      "step": 32833
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.475220542206975e-07,
+      "loss": 0.7936,
+      "step": 32834
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.471597601777871e-07,
+      "loss": 0.6419,
+      "step": 32835
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.467975319681333e-07,
+      "loss": 1.0222,
+      "step": 32836
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.464353695943762e-07,
+      "loss": 0.9126,
+      "step": 32837
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.460732730591459e-07,
+      "loss": 0.9653,
+      "step": 32838
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.457112423650761e-07,
+      "loss": 1.0299,
+      "step": 32839
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.453492775147988e-07,
+      "loss": 0.5698,
+      "step": 32840
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.449873785109498e-07,
+      "loss": 0.7366,
+      "step": 32841
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.446255453561548e-07,
+      "loss": 0.675,
+      "step": 32842
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.442637780530495e-07,
+      "loss": 0.739,
+      "step": 32843
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.439020766042639e-07,
+      "loss": 0.755,
+      "step": 32844
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.435404410124271e-07,
+      "loss": 0.5949,
+      "step": 32845
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.431788712801715e-07,
+      "loss": 0.8192,
+      "step": 32846
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.428173674101237e-07,
+      "loss": 0.6277,
+      "step": 32847
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.424559294049152e-07,
+      "loss": 0.7906,
+      "step": 32848
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.420945572671691e-07,
+      "loss": 0.9132,
+      "step": 32849
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.417332509995203e-07,
+      "loss": 0.9306,
+      "step": 32850
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.413720106045954e-07,
+      "loss": 0.7073,
+      "step": 32851
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.410108360850168e-07,
+      "loss": 0.7223,
+      "step": 32852
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.406497274434134e-07,
+      "loss": 0.8217,
+      "step": 32853
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.402886846824099e-07,
+      "loss": 0.772,
+      "step": 32854
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.399277078046343e-07,
+      "loss": 0.8522,
+      "step": 32855
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.395667968127109e-07,
+      "loss": 0.9291,
+      "step": 32856
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.392059517092633e-07,
+      "loss": 0.8863,
+      "step": 32857
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.388451724969161e-07,
+      "loss": 0.5489,
+      "step": 32858
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.384844591782938e-07,
+      "loss": 0.7871,
+      "step": 32859
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.381238117560188e-07,
+      "loss": 0.8734,
+      "step": 32860
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.377632302327133e-07,
+      "loss": 0.7895,
+      "step": 32861
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.374027146109998e-07,
+      "loss": 0.9118,
+      "step": 32862
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.370422648935007e-07,
+      "loss": 0.5393,
+      "step": 32863
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.36681881082836e-07,
+      "loss": 0.774,
+      "step": 32864
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.36321563181628e-07,
+      "loss": 0.6342,
+      "step": 32865
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.359613111924948e-07,
+      "loss": 0.9852,
+      "step": 32866
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.356011251180563e-07,
+      "loss": 0.606,
+      "step": 32867
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.352410049609351e-07,
+      "loss": 0.8374,
+      "step": 32868
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.348809507237477e-07,
+      "loss": 1.0546,
+      "step": 32869
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.345209624091133e-07,
+      "loss": 0.6137,
+      "step": 32870
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.341610400196499e-07,
+      "loss": 0.8548,
+      "step": 32871
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.33801183557973e-07,
+      "loss": 0.9739,
+      "step": 32872
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.334413930266995e-07,
+      "loss": 1.0319,
+      "step": 32873
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.330816684284472e-07,
+      "loss": 0.8886,
+      "step": 32874
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.32722009765833e-07,
+      "loss": 0.8246,
+      "step": 32875
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.323624170414714e-07,
+      "loss": 0.8587,
+      "step": 32876
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.320028902579759e-07,
+      "loss": 0.8186,
+      "step": 32877
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.316434294179632e-07,
+      "loss": 0.8775,
+      "step": 32878
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.31284034524047e-07,
+      "loss": 0.9793,
+      "step": 32879
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.309247055788384e-07,
+      "loss": 0.8271,
+      "step": 32880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.305654425849553e-07,
+      "loss": 0.7554,
+      "step": 32881
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.302062455450067e-07,
+      "loss": 0.8161,
+      "step": 32882
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.298471144616039e-07,
+      "loss": 0.9099,
+      "step": 32883
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.294880493373615e-07,
+      "loss": 0.8322,
+      "step": 32884
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.291290501748873e-07,
+      "loss": 0.8447,
+      "step": 32885
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.287701169767948e-07,
+      "loss": 0.8417,
+      "step": 32886
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.284112497456943e-07,
+      "loss": 0.8541,
+      "step": 32887
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.280524484841947e-07,
+      "loss": 0.6635,
+      "step": 32888
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.276937131949038e-07,
+      "loss": 0.8269,
+      "step": 32889
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.273350438804329e-07,
+      "loss": 0.8435,
+      "step": 32890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.26976440543389e-07,
+      "loss": 0.6178,
+      "step": 32891
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.266179031863797e-07,
+      "loss": 0.7977,
+      "step": 32892
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.262594318120122e-07,
+      "loss": 0.8351,
+      "step": 32893
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.25901026422894e-07,
+      "loss": 0.8043,
+      "step": 32894
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.255426870216311e-07,
+      "loss": 0.9503,
+      "step": 32895
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.251844136108301e-07,
+      "loss": 1.0133,
+      "step": 32896
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.248262061930946e-07,
+      "loss": 0.8852,
+      "step": 32897
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.244680647710303e-07,
+      "loss": 0.8218,
+      "step": 32898
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.241099893472427e-07,
+      "loss": 0.7657,
+      "step": 32899
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.237519799243355e-07,
+      "loss": 0.8416,
+      "step": 32900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.233940365049143e-07,
+      "loss": 0.7756,
+      "step": 32901
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.23036159091577e-07,
+      "loss": 0.9002,
+      "step": 32902
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.226783476869272e-07,
+      "loss": 0.8484,
+      "step": 32903
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.223206022935704e-07,
+      "loss": 0.6985,
+      "step": 32904
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.219629229141059e-07,
+      "loss": 0.6096,
+      "step": 32905
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.216053095511346e-07,
+      "loss": 0.5292,
+      "step": 32906
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.21247762207258e-07,
+      "loss": 0.9526,
+      "step": 32907
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.208902808850761e-07,
+      "loss": 0.8255,
+      "step": 32908
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.205328655871892e-07,
+      "loss": 0.9778,
+      "step": 32909
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.201755163161918e-07,
+      "loss": 0.7265,
+      "step": 32910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.198182330746918e-07,
+      "loss": 0.7987,
+      "step": 32911
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.194610158652795e-07,
+      "loss": 0.827,
+      "step": 32912
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.191038646905548e-07,
+      "loss": 1.0153,
+      "step": 32913
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.187467795531147e-07,
+      "loss": 0.7879,
+      "step": 32914
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.18389760455557e-07,
+      "loss": 0.8013,
+      "step": 32915
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.180328074004763e-07,
+      "loss": 0.8837,
+      "step": 32916
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.176759203904706e-07,
+      "loss": 0.9764,
+      "step": 32917
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.173190994281334e-07,
+      "loss": 0.5998,
+      "step": 32918
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.169623445160613e-07,
+      "loss": 0.6609,
+      "step": 32919
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.166056556568481e-07,
+      "loss": 0.7286,
+      "step": 32920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.162490328530871e-07,
+      "loss": 0.7168,
+      "step": 32921
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.158924761073707e-07,
+      "loss": 0.8077,
+      "step": 32922
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.155359854222945e-07,
+      "loss": 0.6842,
+      "step": 32923
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.151795608004488e-07,
+      "loss": 0.7098,
+      "step": 32924
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.148232022444259e-07,
+      "loss": 0.7935,
+      "step": 32925
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.144669097568181e-07,
+      "loss": 0.8576,
+      "step": 32926
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.141106833402147e-07,
+      "loss": 0.8522,
+      "step": 32927
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.137545229972067e-07,
+      "loss": 0.9477,
+      "step": 32928
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.133984287303865e-07,
+      "loss": 1.0461,
+      "step": 32929
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.130424005423411e-07,
+      "loss": 0.6178,
+      "step": 32930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.126864384356638e-07,
+      "loss": 1.0545,
+      "step": 32931
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.123305424129369e-07,
+      "loss": 0.8092,
+      "step": 32932
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.119747124767519e-07,
+      "loss": 0.7971,
+      "step": 32933
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.116189486296934e-07,
+      "loss": 0.9091,
+      "step": 32934
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.112632508743535e-07,
+      "loss": 0.4943,
+      "step": 32935
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.10907619213317e-07,
+      "loss": 0.9964,
+      "step": 32936
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.105520536491697e-07,
+      "loss": 0.4021,
+      "step": 32937
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.101965541844959e-07,
+      "loss": 0.7525,
+      "step": 32938
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.098411208218827e-07,
+      "loss": 0.6523,
+      "step": 32939
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.094857535639157e-07,
+      "loss": 0.7011,
+      "step": 32940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.091304524131761e-07,
+      "loss": 0.8325,
+      "step": 32941
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.087752173722497e-07,
+      "loss": 0.6771,
+      "step": 32942
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.084200484437201e-07,
+      "loss": 0.7344,
+      "step": 32943
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.080649456301694e-07,
+      "loss": 0.7266,
+      "step": 32944
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.07709908934179e-07,
+      "loss": 0.655,
+      "step": 32945
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.073549383583314e-07,
+      "loss": 1.0459,
+      "step": 32946
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.070000339052087e-07,
+      "loss": 0.8921,
+      "step": 32947
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.066451955773925e-07,
+      "loss": 0.9885,
+      "step": 32948
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.062904233774628e-07,
+      "loss": 0.5983,
+      "step": 32949
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.059357173079986e-07,
+      "loss": 0.9742,
+      "step": 32950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.055810773715812e-07,
+      "loss": 0.9483,
+      "step": 32951
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.052265035707841e-07,
+      "loss": 0.6484,
+      "step": 32952
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.04871995908192e-07,
+      "loss": 0.9103,
+      "step": 32953
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.045175543863805e-07,
+      "loss": 0.8265,
+      "step": 32954
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.041631790079275e-07,
+      "loss": 0.7859,
+      "step": 32955
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.03808869775411e-07,
+      "loss": 0.7385,
+      "step": 32956
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.034546266914046e-07,
+      "loss": 0.9137,
+      "step": 32957
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.031004497584883e-07,
+      "loss": 0.6045,
+      "step": 32958
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.027463389792324e-07,
+      "loss": 0.7162,
+      "step": 32959
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.02392294356218e-07,
+      "loss": 1.0713,
+      "step": 32960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.020383158920187e-07,
+      "loss": 0.9079,
+      "step": 32961
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.016844035892047e-07,
+      "loss": 0.6376,
+      "step": 32962
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.013305574503529e-07,
+      "loss": 0.8597,
+      "step": 32963
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.009767774780332e-07,
+      "loss": 1.1445,
+      "step": 32964
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.006230636748226e-07,
+      "loss": 0.7208,
+      "step": 32965
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.002694160432923e-07,
+      "loss": 0.7727,
+      "step": 32966
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.999158345860127e-07,
+      "loss": 0.8999,
+      "step": 32967
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.995623193055558e-07,
+      "loss": 0.9422,
+      "step": 32968
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.992088702044921e-07,
+      "loss": 0.8272,
+      "step": 32969
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.988554872853927e-07,
+      "loss": 0.9877,
+      "step": 32970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.985021705508256e-07,
+      "loss": 0.8217,
+      "step": 32971
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.981489200033633e-07,
+      "loss": 0.812,
+      "step": 32972
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.977957356455713e-07,
+      "loss": 0.4405,
+      "step": 32973
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.974426174800199e-07,
+      "loss": 0.8331,
+      "step": 32974
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.970895655092771e-07,
+      "loss": 0.9072,
+      "step": 32975
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.967365797359096e-07,
+      "loss": 0.5462,
+      "step": 32976
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.96383660162482e-07,
+      "loss": 1.0212,
+      "step": 32977
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.960308067915658e-07,
+      "loss": 0.787,
+      "step": 32978
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.956780196257243e-07,
+      "loss": 1.2123,
+      "step": 32979
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.953252986675232e-07,
+      "loss": 0.9492,
+      "step": 32980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.949726439195306e-07,
+      "loss": 0.8077,
+      "step": 32981
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.946200553843032e-07,
+      "loss": 0.8278,
+      "step": 32982
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.942675330644113e-07,
+      "loss": 0.6231,
+      "step": 32983
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.939150769624183e-07,
+      "loss": 0.8562,
+      "step": 32984
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.935626870808856e-07,
+      "loss": 1.0798,
+      "step": 32985
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.932103634223754e-07,
+      "loss": 0.8326,
+      "step": 32986
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.928581059894514e-07,
+      "loss": 0.8368,
+      "step": 32987
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.925059147846749e-07,
+      "loss": 0.867,
+      "step": 32988
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.921537898106047e-07,
+      "loss": 0.8448,
+      "step": 32989
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.918017310698057e-07,
+      "loss": 0.8816,
+      "step": 32990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.914497385648369e-07,
+      "loss": 0.8103,
+      "step": 32991
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.910978122982561e-07,
+      "loss": 1.0376,
+      "step": 32992
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.907459522726236e-07,
+      "loss": 0.8121,
+      "step": 32993
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.903941584904974e-07,
+      "loss": 0.7567,
+      "step": 32994
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.900424309544354e-07,
+      "loss": 0.9578,
+      "step": 32995
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.896907696669976e-07,
+      "loss": 0.7683,
+      "step": 32996
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.8933917463074e-07,
+      "loss": 0.5334,
+      "step": 32997
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.889876458482205e-07,
+      "loss": 0.8169,
+      "step": 32998
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.886361833219937e-07,
+      "loss": 0.9541,
+      "step": 32999
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.882847870546174e-07,
+      "loss": 0.8272,
+      "step": 33000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.879334570486453e-07,
+      "loss": 0.7996,
+      "step": 33001
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.875821933066331e-07,
+      "loss": 0.3461,
+      "step": 33002
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.872309958311354e-07,
+      "loss": 0.6142,
+      "step": 33003
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.868798646247057e-07,
+      "loss": 0.7731,
+      "step": 33004
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.865287996898975e-07,
+      "loss": 0.7771,
+      "step": 33005
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.861778010292632e-07,
+      "loss": 0.8707,
+      "step": 33006
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.858268686453553e-07,
+      "loss": 0.8109,
+      "step": 33007
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.854760025407272e-07,
+      "loss": 0.7726,
+      "step": 33008
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.851252027179302e-07,
+      "loss": 0.8396,
+      "step": 33009
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.847744691795157e-07,
+      "loss": 0.7053,
+      "step": 33010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.844238019280349e-07,
+      "loss": 0.6092,
+      "step": 33011
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.840732009660346e-07,
+      "loss": 0.8683,
+      "step": 33012
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.83722666296064e-07,
+      "loss": 0.9326,
+      "step": 33013
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.833721979206767e-07,
+      "loss": 0.8428,
+      "step": 33014
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.830217958424192e-07,
+      "loss": 0.7917,
+      "step": 33015
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.826714600638398e-07,
+      "loss": 0.7997,
+      "step": 33016
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.823211905874851e-07,
+      "loss": 0.8205,
+      "step": 33017
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.819709874159044e-07,
+      "loss": 0.8355,
+      "step": 33018
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.816208505516422e-07,
+      "loss": 0.7443,
+      "step": 33019
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.812707799972442e-07,
+      "loss": 0.9356,
+      "step": 33020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.809207757552618e-07,
+      "loss": 0.7351,
+      "step": 33021
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.805708378282341e-07,
+      "loss": 0.8245,
+      "step": 33022
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.802209662187078e-07,
+      "loss": 0.7021,
+      "step": 33023
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.798711609292277e-07,
+      "loss": 0.727,
+      "step": 33024
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.795214219623349e-07,
+      "loss": 0.7288,
+      "step": 33025
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.791717493205776e-07,
+      "loss": 1.0655,
+      "step": 33026
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.788221430064958e-07,
+      "loss": 0.7956,
+      "step": 33027
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.784726030226321e-07,
+      "loss": 0.9329,
+      "step": 33028
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.781231293715286e-07,
+      "loss": 0.8773,
+      "step": 33029
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.77773722055727e-07,
+      "loss": 0.8958,
+      "step": 33030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.774243810777683e-07,
+      "loss": 1.1718,
+      "step": 33031
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.770751064401916e-07,
+      "loss": 0.9984,
+      "step": 33032
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.767258981455384e-07,
+      "loss": 0.6479,
+      "step": 33033
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.763767561963477e-07,
+      "loss": 0.7269,
+      "step": 33034
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.760276805951573e-07,
+      "loss": 0.8004,
+      "step": 33035
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.756786713445075e-07,
+      "loss": 0.7962,
+      "step": 33036
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.753297284469365e-07,
+      "loss": 0.754,
+      "step": 33037
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.749808519049774e-07,
+      "loss": 0.9615,
+      "step": 33038
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.746320417211739e-07,
+      "loss": 0.8742,
+      "step": 33039
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.742832978980598e-07,
+      "loss": 0.6191,
+      "step": 33040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.739346204381715e-07,
+      "loss": 0.6094,
+      "step": 33041
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.735860093440429e-07,
+      "loss": 0.7321,
+      "step": 33042
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.732374646182085e-07,
+      "loss": 1.0177,
+      "step": 33043
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.728889862632072e-07,
+      "loss": 0.8648,
+      "step": 33044
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.725405742815695e-07,
+      "loss": 0.8577,
+      "step": 33045
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.721922286758311e-07,
+      "loss": 0.5595,
+      "step": 33046
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.718439494485254e-07,
+      "loss": 0.657,
+      "step": 33047
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.714957366021825e-07,
+      "loss": 0.8123,
+      "step": 33048
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.711475901393374e-07,
+      "loss": 0.4,
+      "step": 33049
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.707995100625177e-07,
+      "loss": 0.8938,
+      "step": 33050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.704514963742628e-07,
+      "loss": 0.7743,
+      "step": 33051
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.70103549077096e-07,
+      "loss": 0.7449,
+      "step": 33052
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.697556681735497e-07,
+      "loss": 0.5775,
+      "step": 33053
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.694078536661544e-07,
+      "loss": 0.7802,
+      "step": 33054
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.690601055574388e-07,
+      "loss": 0.9598,
+      "step": 33055
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.6871242384993e-07,
+      "loss": 1.0313,
+      "step": 33056
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.683648085461616e-07,
+      "loss": 0.7177,
+      "step": 33057
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.680172596486569e-07,
+      "loss": 1.136,
+      "step": 33058
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.676697771599452e-07,
+      "loss": 0.7068,
+      "step": 33059
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.673223610825532e-07,
+      "loss": 0.8797,
+      "step": 33060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.669750114190078e-07,
+      "loss": 1.0731,
+      "step": 33061
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.666277281718339e-07,
+      "loss": 0.7404,
+      "step": 33062
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.662805113435569e-07,
+      "loss": 0.6467,
+      "step": 33063
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.659333609367026e-07,
+      "loss": 0.7244,
+      "step": 33064
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.655862769537948e-07,
+      "loss": 0.6555,
+      "step": 33065
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.65239259397358e-07,
+      "loss": 0.7226,
+      "step": 33066
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.648923082699168e-07,
+      "loss": 0.8503,
+      "step": 33067
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.645454235739903e-07,
+      "loss": 0.8111,
+      "step": 33068
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.641986053121065e-07,
+      "loss": 0.6193,
+      "step": 33069
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.638518534867846e-07,
+      "loss": 0.5015,
+      "step": 33070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.63505168100548e-07,
+      "loss": 0.7435,
+      "step": 33071
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.631585491559147e-07,
+      "loss": 0.676,
+      "step": 33072
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.628119966554071e-07,
+      "loss": 0.714,
+      "step": 33073
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.624655106015434e-07,
+      "loss": 0.9889,
+      "step": 33074
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.62119090996848e-07,
+      "loss": 0.896,
+      "step": 33075
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.617727378438357e-07,
+      "loss": 0.8813,
+      "step": 33076
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.614264511450277e-07,
+      "loss": 0.8056,
+      "step": 33077
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.610802309029421e-07,
+      "loss": 0.6243,
+      "step": 33078
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.607340771200945e-07,
+      "loss": 0.6071,
+      "step": 33079
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.603879897990042e-07,
+      "loss": 1.005,
+      "step": 33080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.600419689421879e-07,
+      "loss": 0.773,
+      "step": 33081
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.596960145521605e-07,
+      "loss": 0.6741,
+      "step": 33082
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.593501266314397e-07,
+      "loss": 0.8232,
+      "step": 33083
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.59004305182538e-07,
+      "loss": 0.8384,
+      "step": 33084
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.586585502079736e-07,
+      "loss": 0.8138,
+      "step": 33085
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.583128617102576e-07,
+      "loss": 0.741,
+      "step": 33086
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.57967239691907e-07,
+      "loss": 0.8137,
+      "step": 33087
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.576216841554341e-07,
+      "loss": 0.8202,
+      "step": 33088
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.572761951033514e-07,
+      "loss": 0.7735,
+      "step": 33089
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.569307725381715e-07,
+      "loss": 0.8585,
+      "step": 33090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.565854164624077e-07,
+      "loss": 0.7238,
+      "step": 33091
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.56240126878567e-07,
+      "loss": 0.5976,
+      "step": 33092
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.558949037891651e-07,
+      "loss": 0.8266,
+      "step": 33093
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.555497471967112e-07,
+      "loss": 0.7425,
+      "step": 33094
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.552046571037142e-07,
+      "loss": 0.8307,
+      "step": 33095
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.548596335126846e-07,
+      "loss": 0.8961,
+      "step": 33096
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.545146764261325e-07,
+      "loss": 1.0863,
+      "step": 33097
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.541697858465636e-07,
+      "loss": 0.7701,
+      "step": 33098
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.538249617764871e-07,
+      "loss": 0.8477,
+      "step": 33099
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.53480204218412e-07,
+      "loss": 1.0587,
+      "step": 33100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.531355131748475e-07,
+      "loss": 0.6888,
+      "step": 33101
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.52790888648295e-07,
+      "loss": 0.8441,
+      "step": 33102
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.524463306412623e-07,
+      "loss": 0.9462,
+      "step": 33103
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.521018391562541e-07,
+      "loss": 0.8905,
+      "step": 33104
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.517574141957796e-07,
+      "loss": 0.9958,
+      "step": 33105
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.514130557623401e-07,
+      "loss": 0.8805,
+      "step": 33106
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.510687638584414e-07,
+      "loss": 0.8219,
+      "step": 33107
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.50724538486587e-07,
+      "loss": 0.8573,
+      "step": 33108
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.503803796492794e-07,
+      "loss": 1.0216,
+      "step": 33109
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.50036287349022e-07,
+      "loss": 1.1644,
+      "step": 33110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.496922615883163e-07,
+      "loss": 0.8935,
+      "step": 33111
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.493483023696647e-07,
+      "loss": 1.0539,
+      "step": 33112
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.490044096955685e-07,
+      "loss": 0.8096,
+      "step": 33113
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.48660583568528e-07,
+      "loss": 0.7398,
+      "step": 33114
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.483168239910445e-07,
+      "loss": 0.9751,
+      "step": 33115
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.479731309656169e-07,
+      "loss": 0.473,
+      "step": 33116
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.476295044947436e-07,
+      "loss": 0.9186,
+      "step": 33117
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.472859445809267e-07,
+      "loss": 0.774,
+      "step": 33118
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.469424512266621e-07,
+      "loss": 0.7494,
+      "step": 33119
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.46599024434449e-07,
+      "loss": 0.9811,
+      "step": 33120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.462556642067865e-07,
+      "loss": 0.5196,
+      "step": 33121
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.459123705461658e-07,
+      "loss": 1.1032,
+      "step": 33122
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.455691434550873e-07,
+      "loss": 0.7403,
+      "step": 33123
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.452259829360476e-07,
+      "loss": 1.0048,
+      "step": 33124
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.448828889915406e-07,
+      "loss": 0.7174,
+      "step": 33125
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.44539861624063e-07,
+      "loss": 0.8085,
+      "step": 33126
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.441969008361073e-07,
+      "loss": 0.8259,
+      "step": 33127
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.438540066301682e-07,
+      "loss": 0.9209,
+      "step": 33128
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.43511179008738e-07,
+      "loss": 0.6039,
+      "step": 33129
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.431684179743139e-07,
+      "loss": 0.7011,
+      "step": 33130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.428257235293868e-07,
+      "loss": 0.8176,
+      "step": 33131
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.424830956764463e-07,
+      "loss": 0.8718,
+      "step": 33132
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.421405344179856e-07,
+      "loss": 0.5656,
+      "step": 33133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.417980397564963e-07,
+      "loss": 0.6513,
+      "step": 33134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.414556116944672e-07,
+      "loss": 0.7551,
+      "step": 33135
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.41113250234391e-07,
+      "loss": 0.8728,
+      "step": 33136
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.407709553787557e-07,
+      "loss": 0.776,
+      "step": 33137
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.404287271300526e-07,
+      "loss": 0.9657,
+      "step": 33138
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.400865654907675e-07,
+      "loss": 0.8034,
+      "step": 33139
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.397444704633906e-07,
+      "loss": 0.892,
+      "step": 33140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.394024420504099e-07,
+      "loss": 0.6034,
+      "step": 33141
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.390604802543112e-07,
+      "loss": 0.7758,
+      "step": 33142
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.387185850775814e-07,
+      "loss": 0.6664,
+      "step": 33143
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.383767565227085e-07,
+      "loss": 0.8726,
+      "step": 33144
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.380349945921773e-07,
+      "loss": 0.9864,
+      "step": 33145
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.376932992884723e-07,
+      "loss": 0.9614,
+      "step": 33146
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.373516706140772e-07,
+      "loss": 0.7486,
+      "step": 33147
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.370101085714799e-07,
+      "loss": 0.9211,
+      "step": 33148
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.366686131631629e-07,
+      "loss": 0.6953,
+      "step": 33149
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.363271843916099e-07,
+      "loss": 0.9268,
+      "step": 33150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.359858222593043e-07,
+      "loss": 0.8149,
+      "step": 33151
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.356445267687263e-07,
+      "loss": 0.9263,
+      "step": 33152
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.353032979223563e-07,
+      "loss": 1.0788,
+      "step": 33153
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.349621357226811e-07,
+      "loss": 0.6719,
+      "step": 33154
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.346210401721788e-07,
+      "loss": 0.7677,
+      "step": 33155
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.342800112733296e-07,
+      "loss": 0.6037,
+      "step": 33156
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.339390490286147e-07,
+      "loss": 0.889,
+      "step": 33157
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.335981534405124e-07,
+      "loss": 0.8518,
+      "step": 33158
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.332573245115016e-07,
+      "loss": 0.9299,
+      "step": 33159
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.329165622440605e-07,
+      "loss": 0.7095,
+      "step": 33160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.325758666406724e-07,
+      "loss": 0.6847,
+      "step": 33161
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.322352377038079e-07,
+      "loss": 0.8114,
+      "step": 33162
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.318946754359469e-07,
+      "loss": 0.7294,
+      "step": 33163
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.315541798395654e-07,
+      "loss": 0.6765,
+      "step": 33164
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.312137509171392e-07,
+      "loss": 0.9189,
+      "step": 33165
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.308733886711462e-07,
+      "loss": 0.596,
+      "step": 33166
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.3053309310406e-07,
+      "loss": 0.5246,
+      "step": 33167
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.301928642183554e-07,
+      "loss": 0.8882,
+      "step": 33168
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.29852702016507e-07,
+      "loss": 0.9406,
+      "step": 33169
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.295126065009873e-07,
+      "loss": 0.7985,
+      "step": 33170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.291725776742709e-07,
+      "loss": 0.8745,
+      "step": 33171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.288326155388304e-07,
+      "loss": 0.8886,
+      "step": 33172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.28492720097136e-07,
+      "loss": 0.7726,
+      "step": 33173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.281528913516623e-07,
+      "loss": 0.8465,
+      "step": 33174
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.278131293048785e-07,
+      "loss": 0.4319,
+      "step": 33175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.274734339592572e-07,
+      "loss": 0.9041,
+      "step": 33176
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.271338053172673e-07,
+      "loss": 0.8715,
+      "step": 33177
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.26794243381377e-07,
+      "loss": 0.5624,
+      "step": 33178
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.264547481540586e-07,
+      "loss": 0.7608,
+      "step": 33179
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.261153196377814e-07,
+      "loss": 0.8927,
+      "step": 33180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.257759578350133e-07,
+      "loss": 0.9108,
+      "step": 33181
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.254366627482191e-07,
+      "loss": 0.8757,
+      "step": 33182
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.250974343798668e-07,
+      "loss": 0.8885,
+      "step": 33183
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.247582727324277e-07,
+      "loss": 1.0075,
+      "step": 33184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.244191778083643e-07,
+      "loss": 0.9111,
+      "step": 33185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.240801496101436e-07,
+      "loss": 0.6894,
+      "step": 33186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.237411881402313e-07,
+      "loss": 0.8132,
+      "step": 33187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.23402293401091e-07,
+      "loss": 0.9261,
+      "step": 33188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.230634653951897e-07,
+      "loss": 0.873,
+      "step": 33189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.227247041249875e-07,
+      "loss": 0.7152,
+      "step": 33190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.223860095929537e-07,
+      "loss": 0.7891,
+      "step": 33191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.220473818015462e-07,
+      "loss": 0.7613,
+      "step": 33192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.217088207532298e-07,
+      "loss": 0.6649,
+      "step": 33193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.213703264504658e-07,
+      "loss": 0.782,
+      "step": 33194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.210318988957166e-07,
+      "loss": 1.0127,
+      "step": 33195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.206935380914405e-07,
+      "loss": 0.6727,
+      "step": 33196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.203552440401019e-07,
+      "loss": 0.9227,
+      "step": 33197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.200170167441601e-07,
+      "loss": 0.9228,
+      "step": 33198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.196788562060731e-07,
+      "loss": 1.0952,
+      "step": 33199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.193407624283011e-07,
+      "loss": 0.7583,
+      "step": 33200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.190027354133035e-07,
+      "loss": 0.9719,
+      "step": 33201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.186647751635357e-07,
+      "loss": 0.7406,
+      "step": 33202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.183268816814583e-07,
+      "loss": 0.6309,
+      "step": 33203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.179890549695279e-07,
+      "loss": 1.0402,
+      "step": 33204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.176512950302007e-07,
+      "loss": 0.7869,
+      "step": 33205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.173136018659323e-07,
+      "loss": 0.5772,
+      "step": 33206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.169759754791784e-07,
+      "loss": 0.9089,
+      "step": 33207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.16638415872395e-07,
+      "loss": 0.7947,
+      "step": 33208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.163009230480368e-07,
+      "loss": 0.8017,
+      "step": 33209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.159634970085595e-07,
+      "loss": 1.006,
+      "step": 33210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.156261377564167e-07,
+      "loss": 0.5993,
+      "step": 33211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.152888452940589e-07,
+      "loss": 0.9262,
+      "step": 33212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.149516196239404e-07,
+      "loss": 0.7246,
+      "step": 33213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.146144607485118e-07,
+      "loss": 0.6349,
+      "step": 33214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.142773686702287e-07,
+      "loss": 0.6546,
+      "step": 33215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.139403433915405e-07,
+      "loss": 0.92,
+      "step": 33216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.136033849148995e-07,
+      "loss": 0.8821,
+      "step": 33217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.132664932427536e-07,
+      "loss": 0.5256,
+      "step": 33218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.129296683775534e-07,
+      "loss": 0.9017,
+      "step": 33219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.1259291032175e-07,
+      "loss": 1.0369,
+      "step": 33220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.122562190777905e-07,
+      "loss": 0.7268,
+      "step": 33221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.119195946481239e-07,
+      "loss": 0.9591,
+      "step": 33222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.115830370351996e-07,
+      "loss": 0.5681,
+      "step": 33223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.11246546241462e-07,
+      "loss": 0.6207,
+      "step": 33224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.109101222693616e-07,
+      "loss": 0.877,
+      "step": 33225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.105737651213408e-07,
+      "loss": 0.9753,
+      "step": 33226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.102374747998487e-07,
+      "loss": 0.7391,
+      "step": 33227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.099012513073312e-07,
+      "loss": 0.9738,
+      "step": 33228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.09565094646233e-07,
+      "loss": 0.6157,
+      "step": 33229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.092290048189966e-07,
+      "loss": 0.6711,
+      "step": 33230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.088929818280699e-07,
+      "loss": 0.6075,
+      "step": 33231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 8.085570256758913e-07,
+      "loss": 0.7431,
+      "step": 33232
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.082211363649073e-07,
+      "loss": 0.9127,
+      "step": 33233
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.078853138975607e-07,
+      "loss": 1.1534,
+      "step": 33234
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.075495582762927e-07,
+      "loss": 0.827,
+      "step": 33235
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.072138695035448e-07,
+      "loss": 0.876,
+      "step": 33236
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.068782475817582e-07,
+      "loss": 0.7758,
+      "step": 33237
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.065426925133746e-07,
+      "loss": 1.2219,
+      "step": 33238
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.062072043008318e-07,
+      "loss": 0.6988,
+      "step": 33239
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.058717829465723e-07,
+      "loss": 0.7978,
+      "step": 33240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.055364284530353e-07,
+      "loss": 0.82,
+      "step": 33241
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.052011408226579e-07,
+      "loss": 0.7965,
+      "step": 33242
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.04865920057879e-07,
+      "loss": 0.9603,
+      "step": 33243
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.045307661611335e-07,
+      "loss": 0.8548,
+      "step": 33244
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.041956791348637e-07,
+      "loss": 0.8026,
+      "step": 33245
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.038606589815034e-07,
+      "loss": 0.801,
+      "step": 33246
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.035257057034906e-07,
+      "loss": 0.9296,
+      "step": 33247
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.031908193032601e-07,
+      "loss": 1.0012,
+      "step": 33248
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.028559997832463e-07,
+      "loss": 0.8804,
+      "step": 33249
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.025212471458854e-07,
+      "loss": 0.6851,
+      "step": 33250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.02186561393612e-07,
+      "loss": 0.6931,
+      "step": 33251
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.018519425288584e-07,
+      "loss": 0.6862,
+      "step": 33252
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.015173905540597e-07,
+      "loss": 0.8492,
+      "step": 33253
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.011829054716469e-07,
+      "loss": 0.8636,
+      "step": 33254
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.008484872840538e-07,
+      "loss": 1.0684,
+      "step": 33255
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.005141359937129e-07,
+      "loss": 0.6722,
+      "step": 33256
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 8.001798516030524e-07,
+      "loss": 0.8539,
+      "step": 33257
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.998456341145077e-07,
+      "loss": 0.9523,
+      "step": 33258
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.995114835305073e-07,
+      "loss": 0.8069,
+      "step": 33259
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.991773998534802e-07,
+      "loss": 0.7989,
+      "step": 33260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.988433830858599e-07,
+      "loss": 0.7539,
+      "step": 33261
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.98509433230068e-07,
+      "loss": 0.5921,
+      "step": 33262
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.981755502885391e-07,
+      "loss": 0.85,
+      "step": 33263
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.97841734263699e-07,
+      "loss": 1.1715,
+      "step": 33264
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.975079851579759e-07,
+      "loss": 0.7774,
+      "step": 33265
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.971743029737977e-07,
+      "loss": 0.6985,
+      "step": 33266
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.968406877135881e-07,
+      "loss": 0.8389,
+      "step": 33267
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.965071393797763e-07,
+      "loss": 0.6566,
+      "step": 33268
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.961736579747847e-07,
+      "loss": 0.8418,
+      "step": 33269
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.958402435010415e-07,
+      "loss": 0.6852,
+      "step": 33270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.955068959609724e-07,
+      "loss": 0.9994,
+      "step": 33271
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.951736153569977e-07,
+      "loss": 1.1095,
+      "step": 33272
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.948404016915423e-07,
+      "loss": 0.86,
+      "step": 33273
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.945072549670296e-07,
+      "loss": 0.8616,
+      "step": 33274
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.941741751858811e-07,
+      "loss": 0.8577,
+      "step": 33275
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.938411623505216e-07,
+      "loss": 0.8717,
+      "step": 33276
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.935082164633723e-07,
+      "loss": 0.865,
+      "step": 33277
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.931753375268525e-07,
+      "loss": 0.5642,
+      "step": 33278
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.928425255433847e-07,
+      "loss": 0.6653,
+      "step": 33279
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.92509780515388e-07,
+      "loss": 0.904,
+      "step": 33280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.921771024452829e-07,
+      "loss": 0.8465,
+      "step": 33281
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.918444913354884e-07,
+      "loss": 0.8866,
+      "step": 33282
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.915119471884236e-07,
+      "loss": 0.7601,
+      "step": 33283
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.911794700065056e-07,
+      "loss": 0.6861,
+      "step": 33284
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.908470597921547e-07,
+      "loss": 0.4326,
+      "step": 33285
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.905147165477855e-07,
+      "loss": 0.8953,
+      "step": 33286
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.901824402758141e-07,
+      "loss": 0.549,
+      "step": 33287
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.898502309786604e-07,
+      "loss": 0.7887,
+      "step": 33288
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.895180886587394e-07,
+      "loss": 0.6893,
+      "step": 33289
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.891860133184648e-07,
+      "loss": 0.7329,
+      "step": 33290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.888540049602545e-07,
+      "loss": 1.0239,
+      "step": 33291
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.885220635865199e-07,
+      "loss": 0.8224,
+      "step": 33292
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.881901891996724e-07,
+      "loss": 0.7288,
+      "step": 33293
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.878583818021312e-07,
+      "loss": 0.7579,
+      "step": 33294
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.875266413963068e-07,
+      "loss": 0.7236,
+      "step": 33295
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.871949679846125e-07,
+      "loss": 0.8771,
+      "step": 33296
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.868633615694588e-07,
+      "loss": 0.7689,
+      "step": 33297
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.86531822153257e-07,
+      "loss": 0.6297,
+      "step": 33298
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.862003497384196e-07,
+      "loss": 0.5665,
+      "step": 33299
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.858689443273548e-07,
+      "loss": 0.9851,
+      "step": 33300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.855376059224784e-07,
+      "loss": 0.6804,
+      "step": 33301
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.852063345261929e-07,
+      "loss": 0.8993,
+      "step": 33302
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.848751301409108e-07,
+      "loss": 0.6879,
+      "step": 33303
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.845439927690402e-07,
+      "loss": 0.4702,
+      "step": 33304
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.842129224129868e-07,
+      "loss": 0.6558,
+      "step": 33305
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.838819190751623e-07,
+      "loss": 0.9672,
+      "step": 33306
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.835509827579712e-07,
+      "loss": 0.5772,
+      "step": 33307
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.832201134638218e-07,
+      "loss": 0.6469,
+      "step": 33308
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.828893111951197e-07,
+      "loss": 0.8591,
+      "step": 33309
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.825585759542687e-07,
+      "loss": 1.0285,
+      "step": 33310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.822279077436767e-07,
+      "loss": 0.7935,
+      "step": 33311
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.818973065657453e-07,
+      "loss": 0.6357,
+      "step": 33312
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.815667724228814e-07,
+      "loss": 0.5329,
+      "step": 33313
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.812363053174877e-07,
+      "loss": 0.9946,
+      "step": 33314
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.809059052519674e-07,
+      "loss": 0.962,
+      "step": 33315
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.805755722287223e-07,
+      "loss": 0.9091,
+      "step": 33316
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.80245306250157e-07,
+      "loss": 0.6698,
+      "step": 33317
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.799151073186684e-07,
+      "loss": 0.5586,
+      "step": 33318
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.795849754366635e-07,
+      "loss": 0.5892,
+      "step": 33319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.792549106065395e-07,
+      "loss": 0.6195,
+      "step": 33320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.789249128307008e-07,
+      "loss": 0.5464,
+      "step": 33321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.785949821115413e-07,
+      "loss": 0.8865,
+      "step": 33322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.782651184514622e-07,
+      "loss": 0.702,
+      "step": 33323
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.779353218528651e-07,
+      "loss": 0.6274,
+      "step": 33324
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.776055923181458e-07,
+      "loss": 0.7771,
+      "step": 33325
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.772759298497024e-07,
+      "loss": 0.9389,
+      "step": 33326
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.769463344499339e-07,
+      "loss": 0.8112,
+      "step": 33327
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.766168061212342e-07,
+      "loss": 0.7393,
+      "step": 33328
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.762873448660025e-07,
+      "loss": 0.7829,
+      "step": 33329
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.759579506866311e-07,
+      "loss": 0.9292,
+      "step": 33330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.756286235855215e-07,
+      "loss": 0.5123,
+      "step": 33331
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.752993635650641e-07,
+      "loss": 0.6028,
+      "step": 33332
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.749701706276525e-07,
+      "loss": 1.1255,
+      "step": 33333
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.746410447756835e-07,
+      "loss": 0.7098,
+      "step": 33334
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.743119860115489e-07,
+      "loss": 1.2249,
+      "step": 33335
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.739829943376398e-07,
+      "loss": 0.8815,
+      "step": 33336
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.736540697563532e-07,
+      "loss": 0.7745,
+      "step": 33337
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.733252122700785e-07,
+      "loss": 0.8279,
+      "step": 33338
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.729964218812081e-07,
+      "loss": 0.6186,
+      "step": 33339
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.72667698592131e-07,
+      "loss": 1.0487,
+      "step": 33340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.723390424052401e-07,
+      "loss": 0.8377,
+      "step": 33341
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.720104533229244e-07,
+      "loss": 0.7067,
+      "step": 33342
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.716819313475731e-07,
+      "loss": 1.0444,
+      "step": 33343
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.713534764815755e-07,
+      "loss": 0.7542,
+      "step": 33344
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.710250887273196e-07,
+      "loss": 0.7408,
+      "step": 33345
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.706967680871957e-07,
+      "loss": 1.0378,
+      "step": 33346
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.703685145635886e-07,
+      "loss": 1.0456,
+      "step": 33347
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.700403281588853e-07,
+      "loss": 0.6569,
+      "step": 33348
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.69712208875475e-07,
+      "loss": 0.8954,
+      "step": 33349
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.693841567157434e-07,
+      "loss": 0.7488,
+      "step": 33350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.690561716820766e-07,
+      "loss": 0.7034,
+      "step": 33351
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.687282537768559e-07,
+      "loss": 0.7969,
+      "step": 33352
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.684004030024694e-07,
+      "loss": 0.7609,
+      "step": 33353
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.680726193612986e-07,
+      "loss": 0.7665,
+      "step": 33354
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.677449028557293e-07,
+      "loss": 0.8051,
+      "step": 33355
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.674172534881463e-07,
+      "loss": 0.6867,
+      "step": 33356
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.670896712609289e-07,
+      "loss": 0.9429,
+      "step": 33357
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.667621561764604e-07,
+      "loss": 0.5418,
+      "step": 33358
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.664347082371237e-07,
+      "loss": 0.9198,
+      "step": 33359
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.66107327445299e-07,
+      "loss": 0.7393,
+      "step": 33360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.657800138033677e-07,
+      "loss": 1.0136,
+      "step": 33361
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.654527673137091e-07,
+      "loss": 1.0114,
+      "step": 33362
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.651255879787034e-07,
+      "loss": 0.6968,
+      "step": 33363
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.647984758007298e-07,
+      "loss": 0.4769,
+      "step": 33364
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.644714307821666e-07,
+      "loss": 0.9599,
+      "step": 33365
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.641444529253917e-07,
+      "loss": 0.7123,
+      "step": 33366
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.638175422327865e-07,
+      "loss": 0.8963,
+      "step": 33367
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.634906987067248e-07,
+      "loss": 0.6817,
+      "step": 33368
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.631639223495846e-07,
+      "loss": 0.8017,
+      "step": 33369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.628372131637419e-07,
+      "loss": 0.7742,
+      "step": 33370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.625105711515746e-07,
+      "loss": 0.898,
+      "step": 33371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.621839963154531e-07,
+      "loss": 0.7316,
+      "step": 33372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.618574886577568e-07,
+      "loss": 0.6684,
+      "step": 33373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.615310481808591e-07,
+      "loss": 0.8062,
+      "step": 33374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.612046748871327e-07,
+      "loss": 0.7767,
+      "step": 33375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.608783687789522e-07,
+      "loss": 0.7693,
+      "step": 33376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.605521298586904e-07,
+      "loss": 0.6093,
+      "step": 33377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.602259581287197e-07,
+      "loss": 0.7445,
+      "step": 33378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.598998535914104e-07,
+      "loss": 0.7263,
+      "step": 33379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.595738162491384e-07,
+      "loss": 0.7253,
+      "step": 33380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.592478461042696e-07,
+      "loss": 0.9341,
+      "step": 33381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.589219431591776e-07,
+      "loss": 0.7171,
+      "step": 33382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.585961074162307e-07,
+      "loss": 0.7728,
+      "step": 33383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.58270338877799e-07,
+      "loss": 0.7196,
+      "step": 33384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.579446375462507e-07,
+      "loss": 0.7276,
+      "step": 33385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.576190034239562e-07,
+      "loss": 0.9499,
+      "step": 33386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.572934365132834e-07,
+      "loss": 0.6897,
+      "step": 33387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.569679368165994e-07,
+      "loss": 0.8523,
+      "step": 33388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.566425043362702e-07,
+      "loss": 0.8291,
+      "step": 33389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.563171390746627e-07,
+      "loss": 0.6892,
+      "step": 33390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.559918410341449e-07,
+      "loss": 0.7898,
+      "step": 33391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.556666102170795e-07,
+      "loss": 0.4829,
+      "step": 33392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.553414466258336e-07,
+      "loss": 0.8055,
+      "step": 33393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.550163502627716e-07,
+      "loss": 0.7448,
+      "step": 33394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.546913211302576e-07,
+      "loss": 0.7092,
+      "step": 33395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.543663592306538e-07,
+      "loss": 0.8756,
+      "step": 33396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.540414645663241e-07,
+      "loss": 1.014,
+      "step": 33397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.537166371396321e-07,
+      "loss": 0.8851,
+      "step": 33398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.533918769529403e-07,
+      "loss": 0.8473,
+      "step": 33399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.530671840086113e-07,
+      "loss": 0.6882,
+      "step": 33400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.527425583090031e-07,
+      "loss": 0.8108,
+      "step": 33401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.524179998564773e-07,
+      "loss": 0.6793,
+      "step": 33402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.520935086533932e-07,
+      "loss": 0.5593,
+      "step": 33403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.517690847021142e-07,
+      "loss": 0.583,
+      "step": 33404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.514447280049964e-07,
+      "loss": 0.7489,
+      "step": 33405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.511204385644e-07,
+      "loss": 0.8997,
+      "step": 33406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.507962163826831e-07,
+      "loss": 0.5717,
+      "step": 33407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.504720614622029e-07,
+      "loss": 0.7822,
+      "step": 33408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.501479738053174e-07,
+      "loss": 0.8322,
+      "step": 33409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.498239534143836e-07,
+      "loss": 0.9851,
+      "step": 33410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.495000002917562e-07,
+      "loss": 0.7715,
+      "step": 33411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.491761144397925e-07,
+      "loss": 0.9883,
+      "step": 33412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.488522958608469e-07,
+      "loss": 0.9133,
+      "step": 33413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.485285445572743e-07,
+      "loss": 0.6263,
+      "step": 33414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.482048605314285e-07,
+      "loss": 0.7988,
+      "step": 33415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.478812437856664e-07,
+      "loss": 0.6709,
+      "step": 33416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.475576943223395e-07,
+      "loss": 0.751,
+      "step": 33417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.472342121437992e-07,
+      "loss": 0.8375,
+      "step": 33418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.469107972524003e-07,
+      "loss": 0.9104,
+      "step": 33419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.465874496504944e-07,
+      "loss": 0.5665,
+      "step": 33420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.462641693404293e-07,
+      "loss": 0.9801,
+      "step": 33421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.459409563245612e-07,
+      "loss": 0.7988,
+      "step": 33422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.456178106052369e-07,
+      "loss": 0.6914,
+      "step": 33423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.452947321848081e-07,
+      "loss": 0.8286,
+      "step": 33424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.449717210656237e-07,
+      "loss": 0.6906,
+      "step": 33425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.446487772500321e-07,
+      "loss": 0.7709,
+      "step": 33426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.443259007403836e-07,
+      "loss": 0.6966,
+      "step": 33427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.440030915390228e-07,
+      "loss": 0.8392,
+      "step": 33428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.436803496483002e-07,
+      "loss": 0.6426,
+      "step": 33429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.43357675070564e-07,
+      "loss": 0.9266,
+      "step": 33430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.430350678081577e-07,
+      "loss": 0.9604,
+      "step": 33431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.427125278634283e-07,
+      "loss": 0.8639,
+      "step": 33432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.423900552387197e-07,
+      "loss": 0.7244,
+      "step": 33433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.420676499363799e-07,
+      "loss": 0.7236,
+      "step": 33434
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.417453119587525e-07,
+      "loss": 0.5196,
+      "step": 33435
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.414230413081802e-07,
+      "loss": 0.7358,
+      "step": 33436
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.411008379870088e-07,
+      "loss": 0.814,
+      "step": 33437
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.40778701997581e-07,
+      "loss": 0.9068,
+      "step": 33438
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.40456633342237e-07,
+      "loss": 0.7726,
+      "step": 33439
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.401346320233215e-07,
+      "loss": 0.6979,
+      "step": 33440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.398126980431753e-07,
+      "loss": 0.7782,
+      "step": 33441
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.394908314041394e-07,
+      "loss": 0.8315,
+      "step": 33442
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.391690321085543e-07,
+      "loss": 0.8495,
+      "step": 33443
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.388473001587604e-07,
+      "loss": 0.8184,
+      "step": 33444
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.38525635557098e-07,
+      "loss": 0.9398,
+      "step": 33445
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.38204038305903e-07,
+      "loss": 0.815,
+      "step": 33446
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.378825084075192e-07,
+      "loss": 0.7941,
+      "step": 33447
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.375610458642823e-07,
+      "loss": 0.7089,
+      "step": 33448
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.372396506785296e-07,
+      "loss": 0.9663,
+      "step": 33449
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.369183228526e-07,
+      "loss": 0.4739,
+      "step": 33450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.365970623888263e-07,
+      "loss": 0.7371,
+      "step": 33451
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.362758692895489e-07,
+      "loss": 0.9481,
+      "step": 33452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.359547435571024e-07,
+      "loss": 0.8953,
+      "step": 33453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.356336851938217e-07,
+      "loss": 0.8943,
+      "step": 33454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.353126942020417e-07,
+      "loss": 0.6815,
+      "step": 33455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.349917705840969e-07,
+      "loss": 0.6226,
+      "step": 33456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.346709143423203e-07,
+      "loss": 0.6762,
+      "step": 33457
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.343501254790442e-07,
+      "loss": 1.0646,
+      "step": 33458
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.340294039966045e-07,
+      "loss": 0.7511,
+      "step": 33459
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.337087498973328e-07,
+      "loss": 0.8926,
+      "step": 33460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.333881631835593e-07,
+      "loss": 0.5527,
+      "step": 33461
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.330676438576156e-07,
+      "loss": 0.8769,
+      "step": 33462
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.327471919218343e-07,
+      "loss": 0.8977,
+      "step": 33463
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.324268073785413e-07,
+      "loss": 0.6134,
+      "step": 33464
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.321064902300723e-07,
+      "loss": 0.7294,
+      "step": 33465
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.317862404787535e-07,
+      "loss": 0.8332,
+      "step": 33466
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.314660581269151e-07,
+      "loss": 0.5419,
+      "step": 33467
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.311459431768842e-07,
+      "loss": 0.9951,
+      "step": 33468
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.3082589563099e-07,
+      "loss": 0.6878,
+      "step": 33469
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.305059154915583e-07,
+      "loss": 0.8594,
+      "step": 33470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.301860027609176e-07,
+      "loss": 0.9595,
+      "step": 33471
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.298661574413935e-07,
+      "loss": 0.6752,
+      "step": 33472
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.29546379535312e-07,
+      "loss": 1.0023,
+      "step": 33473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.292266690449978e-07,
+      "loss": 0.9176,
+      "step": 33474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.289070259727782e-07,
+      "loss": 0.7607,
+      "step": 33475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.285874503209733e-07,
+      "loss": 0.7643,
+      "step": 33476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.282679420919125e-07,
+      "loss": 0.9781,
+      "step": 33477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.279485012879173e-07,
+      "loss": 0.7751,
+      "step": 33478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.27629127911309e-07,
+      "loss": 0.8522,
+      "step": 33479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.273098219644137e-07,
+      "loss": 0.9128,
+      "step": 33480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.269905834495494e-07,
+      "loss": 0.8798,
+      "step": 33481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.266714123690377e-07,
+      "loss": 0.9056,
+      "step": 33482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.263523087252033e-07,
+      "loss": 0.6702,
+      "step": 33483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.260332725203645e-07,
+      "loss": 1.1345,
+      "step": 33484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.257143037568415e-07,
+      "loss": 0.7933,
+      "step": 33485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.253954024369536e-07,
+      "loss": 0.6582,
+      "step": 33486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.250765685630201e-07,
+      "loss": 0.4152,
+      "step": 33487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.247578021373602e-07,
+      "loss": 0.6489,
+      "step": 33488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.244391031622899e-07,
+      "loss": 0.6117,
+      "step": 33489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.241204716401318e-07,
+      "loss": 0.8957,
+      "step": 33490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.238019075731983e-07,
+      "loss": 0.8319,
+      "step": 33491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.234834109638078e-07,
+      "loss": 0.8125,
+      "step": 33492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.231649818142749e-07,
+      "loss": 0.8299,
+      "step": 33493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.228466201269157e-07,
+      "loss": 0.7896,
+      "step": 33494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.225283259040472e-07,
+      "loss": 0.992,
+      "step": 33495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.222100991479831e-07,
+      "loss": 0.889,
+      "step": 33496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.218919398610368e-07,
+      "loss": 0.9043,
+      "step": 33497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.215738480455237e-07,
+      "loss": 0.8438,
+      "step": 33498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.212558237037548e-07,
+      "loss": 1.0336,
+      "step": 33499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.209378668380452e-07,
+      "loss": 0.9278,
+      "step": 33500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.20619977450705e-07,
+      "loss": 0.663,
+      "step": 33501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.20302155544047e-07,
+      "loss": 0.7502,
+      "step": 33502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.199844011203816e-07,
+      "loss": 0.6724,
+      "step": 33503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.196667141820213e-07,
+      "loss": 0.8832,
+      "step": 33504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.193490947312742e-07,
+      "loss": 0.791,
+      "step": 33505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.190315427704519e-07,
+      "loss": 0.9868,
+      "step": 33506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.187140583018604e-07,
+      "loss": 0.5857,
+      "step": 33507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.183966413278132e-07,
+      "loss": 0.7529,
+      "step": 33508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.180792918506163e-07,
+      "loss": 0.9998,
+      "step": 33509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.17762009872579e-07,
+      "loss": 0.8333,
+      "step": 33510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.174447953960051e-07,
+      "loss": 0.8433,
+      "step": 33511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.171276484232025e-07,
+      "loss": 0.6786,
+      "step": 33512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.168105689564797e-07,
+      "loss": 0.691,
+      "step": 33513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.164935569981413e-07,
+      "loss": 1.1257,
+      "step": 33514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.161766125504932e-07,
+      "loss": 0.8426,
+      "step": 33515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.158597356158392e-07,
+      "loss": 0.508,
+      "step": 33516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.155429261964842e-07,
+      "loss": 1.0116,
+      "step": 33517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.152261842947328e-07,
+      "loss": 0.5859,
+      "step": 33518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.149095099128845e-07,
+      "loss": 0.6928,
+      "step": 33519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.145929030532495e-07,
+      "loss": 0.7152,
+      "step": 33520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.142763637181249e-07,
+      "loss": 1.1158,
+      "step": 33521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.139598919098134e-07,
+      "loss": 0.8752,
+      "step": 33522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.136434876306164e-07,
+      "loss": 0.6773,
+      "step": 33523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.133271508828355e-07,
+      "loss": 0.9321,
+      "step": 33524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.130108816687687e-07,
+      "loss": 0.5638,
+      "step": 33525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.126946799907197e-07,
+      "loss": 0.9316,
+      "step": 33526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.123785458509868e-07,
+      "loss": 0.6905,
+      "step": 33527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.120624792518682e-07,
+      "loss": 0.637,
+      "step": 33528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.117464801956631e-07,
+      "loss": 0.9031,
+      "step": 33529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.114305486846684e-07,
+      "loss": 0.7475,
+      "step": 33530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.111146847211825e-07,
+      "loss": 0.6819,
+      "step": 33531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.107988883075023e-07,
+      "loss": 0.9616,
+      "step": 33532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.104831594459238e-07,
+      "loss": 0.5024,
+      "step": 33533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.10167498138744e-07,
+      "loss": 1.0536,
+      "step": 33534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.098519043882568e-07,
+      "loss": 0.8313,
+      "step": 33535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.095363781967591e-07,
+      "loss": 0.7908,
+      "step": 33536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.092209195665422e-07,
+      "loss": 0.651,
+      "step": 33537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.089055284999047e-07,
+      "loss": 0.6414,
+      "step": 33538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.085902049991377e-07,
+      "loss": 0.5932,
+      "step": 33539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.082749490665353e-07,
+      "loss": 0.7579,
+      "step": 33540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.079597607043887e-07,
+      "loss": 0.8703,
+      "step": 33541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.076446399149906e-07,
+      "loss": 0.6531,
+      "step": 33542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.073295867006302e-07,
+      "loss": 0.8971,
+      "step": 33543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.070146010636037e-07,
+      "loss": 0.7505,
+      "step": 33544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.066996830061978e-07,
+      "loss": 0.4507,
+      "step": 33545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.063848325307043e-07,
+      "loss": 0.8251,
+      "step": 33546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.060700496394123e-07,
+      "loss": 0.7822,
+      "step": 33547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.057553343346113e-07,
+      "loss": 1.0088,
+      "step": 33548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.054406866185893e-07,
+      "loss": 0.7224,
+      "step": 33549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.051261064936354e-07,
+      "loss": 0.9302,
+      "step": 33550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.048115939620359e-07,
+      "loss": 0.7604,
+      "step": 33551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.044971490260788e-07,
+      "loss": 0.7288,
+      "step": 33552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.041827716880501e-07,
+      "loss": 0.8795,
+      "step": 33553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.038684619502379e-07,
+      "loss": 0.9078,
+      "step": 33554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.035542198149237e-07,
+      "loss": 0.8974,
+      "step": 33555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.03240045284398e-07,
+      "loss": 0.6737,
+      "step": 33556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.029259383609422e-07,
+      "loss": 0.7949,
+      "step": 33557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.026118990468422e-07,
+      "loss": 0.7459,
+      "step": 33558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.022979273443809e-07,
+      "loss": 0.7398,
+      "step": 33559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.019840232558428e-07,
+      "loss": 0.607,
+      "step": 33560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.016701867835063e-07,
+      "loss": 0.701,
+      "step": 33561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.013564179296583e-07,
+      "loss": 0.793,
+      "step": 33562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.010427166965794e-07,
+      "loss": 0.6219,
+      "step": 33563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.007290830865499e-07,
+      "loss": 0.9525,
+      "step": 33564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.004155171018512e-07,
+      "loss": 0.8267,
+      "step": 33565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 7.001020187447638e-07,
+      "loss": 0.6839,
+      "step": 33566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.997885880175681e-07,
+      "loss": 0.683,
+      "step": 33567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.994752249225401e-07,
+      "loss": 0.9386,
+      "step": 33568
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.991619294619622e-07,
+      "loss": 0.771,
+      "step": 33569
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.98848701638114e-07,
+      "loss": 0.6591,
+      "step": 33570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.985355414532691e-07,
+      "loss": 1.0182,
+      "step": 33571
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.982224489097067e-07,
+      "loss": 0.792,
+      "step": 33572
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.979094240097017e-07,
+      "loss": 0.8014,
+      "step": 33573
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.975964667555335e-07,
+      "loss": 1.073,
+      "step": 33574
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.972835771494779e-07,
+      "loss": 0.8813,
+      "step": 33575
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.969707551938087e-07,
+      "loss": 0.8275,
+      "step": 33576
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.966580008908008e-07,
+      "loss": 0.8504,
+      "step": 33577
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.96345314242729e-07,
+      "loss": 0.6244,
+      "step": 33578
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.960326952518681e-07,
+      "loss": 0.9636,
+      "step": 33579
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.957201439204897e-07,
+      "loss": 0.5267,
+      "step": 33580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.954076602508686e-07,
+      "loss": 0.5816,
+      "step": 33581
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.950952442452751e-07,
+      "loss": 1.0239,
+      "step": 33582
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.947828959059833e-07,
+      "loss": 0.7198,
+      "step": 33583
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.944706152352632e-07,
+      "loss": 0.8197,
+      "step": 33584
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.941584022353865e-07,
+      "loss": 0.9578,
+      "step": 33585
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.938462569086212e-07,
+      "loss": 0.5246,
+      "step": 33586
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.935341792572403e-07,
+      "loss": 0.8182,
+      "step": 33587
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.932221692835139e-07,
+      "loss": 0.6546,
+      "step": 33588
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.929102269897081e-07,
+      "loss": 0.8815,
+      "step": 33589
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.925983523780943e-07,
+      "loss": 0.5703,
+      "step": 33590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.922865454509364e-07,
+      "loss": 0.9142,
+      "step": 33591
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.919748062105059e-07,
+      "loss": 0.9411,
+      "step": 33592
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.916631346590674e-07,
+      "loss": 0.7842,
+      "step": 33593
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.913515307988894e-07,
+      "loss": 0.887,
+      "step": 33594
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.910399946322355e-07,
+      "loss": 0.7665,
+      "step": 33595
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.907285261613739e-07,
+      "loss": 0.981,
+      "step": 33596
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.904171253885672e-07,
+      "loss": 0.7718,
+      "step": 33597
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.901057923160793e-07,
+      "loss": 0.5121,
+      "step": 33598
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.897945269461781e-07,
+      "loss": 0.9916,
+      "step": 33599
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.894833292811265e-07,
+      "loss": 0.6164,
+      "step": 33600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.891721993231837e-07,
+      "loss": 0.8042,
+      "step": 33601
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.888611370746157e-07,
+      "loss": 1.018,
+      "step": 33602
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.885501425376828e-07,
+      "loss": 0.5586,
+      "step": 33603
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.882392157146456e-07,
+      "loss": 0.8493,
+      "step": 33604
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.879283566077688e-07,
+      "loss": 0.688,
+      "step": 33605
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.876175652193107e-07,
+      "loss": 0.8393,
+      "step": 33606
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.873068415515327e-07,
+      "loss": 0.8528,
+      "step": 33607
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.869961856066931e-07,
+      "loss": 0.8456,
+      "step": 33608
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.866855973870512e-07,
+      "loss": 0.9357,
+      "step": 33609
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.863750768948651e-07,
+      "loss": 0.8058,
+      "step": 33610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.860646241323943e-07,
+      "loss": 0.7233,
+      "step": 33611
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.857542391018968e-07,
+      "loss": 0.5732,
+      "step": 33612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.854439218056274e-07,
+      "loss": 0.7996,
+      "step": 33613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.851336722458446e-07,
+      "loss": 0.9695,
+      "step": 33614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.848234904248041e-07,
+      "loss": 0.7329,
+      "step": 33615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.845133763447609e-07,
+      "loss": 0.7745,
+      "step": 33616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.84203330007972e-07,
+      "loss": 0.9058,
+      "step": 33617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.838933514166901e-07,
+      "loss": 1.0656,
+      "step": 33618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.835834405731712e-07,
+      "loss": 0.6734,
+      "step": 33619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.83273597479669e-07,
+      "loss": 0.9482,
+      "step": 33620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.82963822138435e-07,
+      "loss": 0.7073,
+      "step": 33621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.826541145517207e-07,
+      "loss": 0.7543,
+      "step": 33622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.823444747217823e-07,
+      "loss": 0.4997,
+      "step": 33623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.820349026508688e-07,
+      "loss": 0.8189,
+      "step": 33624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.81725398341233e-07,
+      "loss": 0.7985,
+      "step": 33625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.814159617951244e-07,
+      "loss": 0.8091,
+      "step": 33626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.811065930147931e-07,
+      "loss": 0.5931,
+      "step": 33627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.807972920024907e-07,
+      "loss": 1.0464,
+      "step": 33628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.804880587604634e-07,
+      "loss": 0.7924,
+      "step": 33629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.801788932909648e-07,
+      "loss": 0.6596,
+      "step": 33630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.798697955962385e-07,
+      "loss": 0.8574,
+      "step": 33631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.795607656785342e-07,
+      "loss": 0.9295,
+      "step": 33632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.792518035400996e-07,
+      "loss": 0.8623,
+      "step": 33633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.789429091831778e-07,
+      "loss": 0.7437,
+      "step": 33634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.786340826100213e-07,
+      "loss": 1.102,
+      "step": 33635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.783253238228726e-07,
+      "loss": 0.8453,
+      "step": 33636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.780166328239768e-07,
+      "loss": 0.5017,
+      "step": 33637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.777080096155797e-07,
+      "loss": 0.7725,
+      "step": 33638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.773994541999252e-07,
+      "loss": 0.7694,
+      "step": 33639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.770909665792569e-07,
+      "loss": 0.8448,
+      "step": 33640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.767825467558176e-07,
+      "loss": 0.5799,
+      "step": 33641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.76474194731852e-07,
+      "loss": 0.8488,
+      "step": 33642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.761659105096007e-07,
+      "loss": 0.9877,
+      "step": 33643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.758576940913075e-07,
+      "loss": 0.9053,
+      "step": 33644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.755495454792116e-07,
+      "loss": 0.7537,
+      "step": 33645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.752414646755545e-07,
+      "loss": 0.8453,
+      "step": 33646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.749334516825767e-07,
+      "loss": 0.8952,
+      "step": 33647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.746255065025187e-07,
+      "loss": 0.5514,
+      "step": 33648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.743176291376197e-07,
+      "loss": 0.8858,
+      "step": 33649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.740098195901213e-07,
+      "loss": 0.8462,
+      "step": 33650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.737020778622561e-07,
+      "loss": 0.7129,
+      "step": 33651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.733944039562645e-07,
+      "loss": 0.6174,
+      "step": 33652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.730867978743849e-07,
+      "loss": 0.9586,
+      "step": 33653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.727792596188542e-07,
+      "loss": 0.8843,
+      "step": 33654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.724717891919097e-07,
+      "loss": 0.9788,
+      "step": 33655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.72164386595785e-07,
+      "loss": 0.7408,
+      "step": 33656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.718570518327172e-07,
+      "loss": 0.5138,
+      "step": 33657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.715497849049402e-07,
+      "loss": 0.9492,
+      "step": 33658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.712425858146876e-07,
+      "loss": 0.6871,
+      "step": 33659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.709354545641989e-07,
+      "loss": 1.1111,
+      "step": 33660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.70628391155701e-07,
+      "loss": 0.8569,
+      "step": 33661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.7032139559143e-07,
+      "loss": 0.676,
+      "step": 33662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.700144678736176e-07,
+      "loss": 0.8529,
+      "step": 33663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.697076080044951e-07,
+      "loss": 0.8968,
+      "step": 33664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.694008159862952e-07,
+      "loss": 0.9305,
+      "step": 33665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.690940918212486e-07,
+      "loss": 0.8005,
+      "step": 33666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.687874355115876e-07,
+      "loss": 0.6469,
+      "step": 33667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.684808470595394e-07,
+      "loss": 0.9686,
+      "step": 33668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.681743264673346e-07,
+      "loss": 0.7005,
+      "step": 33669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.678678737372035e-07,
+      "loss": 0.7893,
+      "step": 33670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.675614888713721e-07,
+      "loss": 0.856,
+      "step": 33671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.67255171872071e-07,
+      "loss": 0.6609,
+      "step": 33672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.66948922741526e-07,
+      "loss": 0.8392,
+      "step": 33673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.666427414819654e-07,
+      "loss": 0.8805,
+      "step": 33674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.663366280956152e-07,
+      "loss": 0.7144,
+      "step": 33675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.660305825847014e-07,
+      "loss": 0.8574,
+      "step": 33676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.657246049514477e-07,
+      "loss": 0.8458,
+      "step": 33677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.654186951980835e-07,
+      "loss": 0.9159,
+      "step": 33678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.651128533268314e-07,
+      "loss": 0.8073,
+      "step": 33679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.648070793399153e-07,
+      "loss": 0.8867,
+      "step": 33680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.645013732395589e-07,
+      "loss": 0.9969,
+      "step": 33681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.641957350279838e-07,
+      "loss": 0.7918,
+      "step": 33682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.638901647074137e-07,
+      "loss": 0.8275,
+      "step": 33683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.635846622800713e-07,
+      "loss": 0.6634,
+      "step": 33684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.632792277481792e-07,
+      "loss": 0.8269,
+      "step": 33685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.629738611139569e-07,
+      "loss": 0.6853,
+      "step": 33686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.626685623796259e-07,
+      "loss": 0.876,
+      "step": 33687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.623633315474054e-07,
+      "loss": 0.9408,
+      "step": 33688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.62058168619516e-07,
+      "loss": 0.9831,
+      "step": 33689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.617530735981758e-07,
+      "loss": 0.6497,
+      "step": 33690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.614480464856054e-07,
+      "loss": 0.773,
+      "step": 33691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.611430872840208e-07,
+      "loss": 0.9711,
+      "step": 33692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.608381959956411e-07,
+      "loss": 0.654,
+      "step": 33693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.605333726226837e-07,
+      "loss": 0.8393,
+      "step": 33694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.602286171673622e-07,
+      "loss": 0.7393,
+      "step": 33695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.599239296318982e-07,
+      "loss": 0.8741,
+      "step": 33696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.596193100185033e-07,
+      "loss": 0.8207,
+      "step": 33697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.593147583293946e-07,
+      "loss": 0.8303,
+      "step": 33698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.590102745667859e-07,
+      "loss": 1.0009,
+      "step": 33699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.587058587328943e-07,
+      "loss": 0.5686,
+      "step": 33700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.584015108299269e-07,
+      "loss": 0.9801,
+      "step": 33701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.58097230860103e-07,
+      "loss": 0.6849,
+      "step": 33702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.577930188256331e-07,
+      "loss": 0.9421,
+      "step": 33703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.5748887472873e-07,
+      "loss": 0.8876,
+      "step": 33704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.571847985716063e-07,
+      "loss": 0.6849,
+      "step": 33705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.568807903564712e-07,
+      "loss": 1.0046,
+      "step": 33706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.565768500855363e-07,
+      "loss": 0.7632,
+      "step": 33707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.56272977761011e-07,
+      "loss": 0.7005,
+      "step": 33708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.559691733851071e-07,
+      "loss": 0.8989,
+      "step": 33709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.556654369600357e-07,
+      "loss": 0.6135,
+      "step": 33710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.553617684880009e-07,
+      "loss": 0.5336,
+      "step": 33711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.550581679712121e-07,
+      "loss": 0.8548,
+      "step": 33712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.547546354118772e-07,
+      "loss": 0.9908,
+      "step": 33713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.544511708122059e-07,
+      "loss": 1.0059,
+      "step": 33714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.54147774174404e-07,
+      "loss": 0.8879,
+      "step": 33715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.538444455006765e-07,
+      "loss": 0.705,
+      "step": 33716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.535411847932304e-07,
+      "loss": 0.6138,
+      "step": 33717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.532379920542709e-07,
+      "loss": 0.878,
+      "step": 33718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.529348672860026e-07,
+      "loss": 0.3822,
+      "step": 33719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.526318104906293e-07,
+      "loss": 0.8384,
+      "step": 33720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.523288216703561e-07,
+      "loss": 1.0246,
+      "step": 33721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.520259008273855e-07,
+      "loss": 0.8777,
+      "step": 33722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.517230479639203e-07,
+      "loss": 0.7458,
+      "step": 33723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.514202630821643e-07,
+      "loss": 0.8879,
+      "step": 33724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.511175461843178e-07,
+      "loss": 0.7404,
+      "step": 33725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.508148972725803e-07,
+      "loss": 0.9335,
+      "step": 33726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.505123163491578e-07,
+      "loss": 0.991,
+      "step": 33727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.502098034162474e-07,
+      "loss": 0.7092,
+      "step": 33728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.499073584760496e-07,
+      "loss": 0.8436,
+      "step": 33729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.496049815307648e-07,
+      "loss": 0.7653,
+      "step": 33730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.493026725825879e-07,
+      "loss": 0.7359,
+      "step": 33731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.490004316337229e-07,
+      "loss": 0.682,
+      "step": 33732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.486982586863644e-07,
+      "loss": 0.6062,
+      "step": 33733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.483961537427108e-07,
+      "loss": 0.9779,
+      "step": 33734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.480941168049593e-07,
+      "loss": 0.5186,
+      "step": 33735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.477921478753046e-07,
+      "loss": 0.9839,
+      "step": 33736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.474902469559452e-07,
+      "loss": 0.9556,
+      "step": 33737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.471884140490736e-07,
+      "loss": 0.6995,
+      "step": 33738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.468866491568871e-07,
+      "loss": 0.845,
+      "step": 33739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.465849522815815e-07,
+      "loss": 0.7089,
+      "step": 33740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.462833234253474e-07,
+      "loss": 0.829,
+      "step": 33741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.459817625903797e-07,
+      "loss": 0.6467,
+      "step": 33742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.456802697788722e-07,
+      "loss": 0.8978,
+      "step": 33743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.453788449930132e-07,
+      "loss": 0.6977,
+      "step": 33744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.450774882350009e-07,
+      "loss": 0.5836,
+      "step": 33745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.447761995070235e-07,
+      "loss": 0.84,
+      "step": 33746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.444749788112725e-07,
+      "loss": 0.5874,
+      "step": 33747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.441738261499386e-07,
+      "loss": 0.8588,
+      "step": 33748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.43872741525211e-07,
+      "loss": 0.8696,
+      "step": 33749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.435717249392803e-07,
+      "loss": 0.6435,
+      "step": 33750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.432707763943358e-07,
+      "loss": 0.7087,
+      "step": 33751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.429698958925634e-07,
+      "loss": 0.9925,
+      "step": 33752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.426690834361548e-07,
+      "loss": 0.6583,
+      "step": 33753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.42368339027295e-07,
+      "loss": 0.7666,
+      "step": 33754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.42067662668171e-07,
+      "loss": 1.1504,
+      "step": 33755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.417670543609689e-07,
+      "loss": 0.8711,
+      "step": 33756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.41466514107878e-07,
+      "loss": 0.7264,
+      "step": 33757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.411660419110821e-07,
+      "loss": 0.6619,
+      "step": 33758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.408656377727651e-07,
+      "loss": 0.7176,
+      "step": 33759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.405653016951152e-07,
+      "loss": 0.8906,
+      "step": 33760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.402650336803118e-07,
+      "loss": 0.9611,
+      "step": 33761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.399648337305386e-07,
+      "loss": 0.5124,
+      "step": 33762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.396647018479818e-07,
+      "loss": 0.8485,
+      "step": 33763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.393646380348228e-07,
+      "loss": 1.1691,
+      "step": 33764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.390646422932445e-07,
+      "loss": 0.9042,
+      "step": 33765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.387647146254272e-07,
+      "loss": 1.125,
+      "step": 33766
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.384648550335526e-07,
+      "loss": 0.8597,
+      "step": 33767
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.381650635198e-07,
+      "loss": 0.7029,
+      "step": 33768
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.378653400863499e-07,
+      "loss": 1.0031,
+      "step": 33769
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.375656847353861e-07,
+      "loss": 0.9113,
+      "step": 33770
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.372660974690825e-07,
+      "loss": 0.7649,
+      "step": 33771
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.369665782896195e-07,
+      "loss": 0.7959,
+      "step": 33772
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.366671271991753e-07,
+      "loss": 0.7741,
+      "step": 33773
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.363677441999261e-07,
+      "loss": 0.9053,
+      "step": 33774
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.360684292940522e-07,
+      "loss": 0.8584,
+      "step": 33775
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.357691824837287e-07,
+      "loss": 1.0188,
+      "step": 33776
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.354700037711303e-07,
+      "loss": 0.8499,
+      "step": 33777
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.351708931584354e-07,
+      "loss": 0.7792,
+      "step": 33778
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.348718506478168e-07,
+      "loss": 0.789,
+      "step": 33779
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.345728762414504e-07,
+      "loss": 0.6887,
+      "step": 33780
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.342739699415102e-07,
+      "loss": 0.7153,
+      "step": 33781
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.339751317501696e-07,
+      "loss": 0.8294,
+      "step": 33782
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.336763616696018e-07,
+      "loss": 0.9177,
+      "step": 33783
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.333776597019803e-07,
+      "loss": 0.7783,
+      "step": 33784
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.330790258494756e-07,
+      "loss": 0.6781,
+      "step": 33785
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.327804601142607e-07,
+      "loss": 0.9898,
+      "step": 33786
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.324819624985046e-07,
+      "loss": 0.7609,
+      "step": 33787
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.321835330043824e-07,
+      "loss": 0.8262,
+      "step": 33788
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.318851716340613e-07,
+      "loss": 0.8029,
+      "step": 33789
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.315868783897128e-07,
+      "loss": 0.7363,
+      "step": 33790
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.312886532735041e-07,
+      "loss": 0.7016,
+      "step": 33791
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.309904962876023e-07,
+      "loss": 0.989,
+      "step": 33792
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.306924074341802e-07,
+      "loss": 0.76,
+      "step": 33793
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.303943867154028e-07,
+      "loss": 0.654,
+      "step": 33794
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.300964341334382e-07,
+      "loss": 0.9493,
+      "step": 33795
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.297985496904524e-07,
+      "loss": 1.0089,
+      "step": 33796
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.295007333886128e-07,
+      "loss": 0.8424,
+      "step": 33797
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.292029852300841e-07,
+      "loss": 0.8464,
+      "step": 33798
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.289053052170302e-07,
+      "loss": 0.7825,
+      "step": 33799
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.286076933516205e-07,
+      "loss": 0.9107,
+      "step": 33800
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.283101496360156e-07,
+      "loss": 0.7425,
+      "step": 33801
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.280126740723791e-07,
+      "loss": 0.8658,
+      "step": 33802
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.27715266662875e-07,
+      "loss": 0.881,
+      "step": 33803
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.27417927409667e-07,
+      "loss": 0.9691,
+      "step": 33804
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.271206563149146e-07,
+      "loss": 0.936,
+      "step": 33805
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.268234533807838e-07,
+      "loss": 0.8852,
+      "step": 33806
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.265263186094328e-07,
+      "loss": 0.805,
+      "step": 33807
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.262292520030244e-07,
+      "loss": 0.8838,
+      "step": 33808
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.259322535637169e-07,
+      "loss": 0.6615,
+      "step": 33809
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.256353232936718e-07,
+      "loss": 0.7282,
+      "step": 33810
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.253384611950475e-07,
+      "loss": 0.7103,
+      "step": 33811
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.250416672700021e-07,
+      "loss": 0.7033,
+      "step": 33812
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.247449415206952e-07,
+      "loss": 0.6842,
+      "step": 33813
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.244482839492849e-07,
+      "loss": 0.7846,
+      "step": 33814
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.241516945579262e-07,
+      "loss": 0.837,
+      "step": 33815
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.238551733487785e-07,
+      "loss": 0.5128,
+      "step": 33816
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.235587203239968e-07,
+      "loss": 0.9915,
+      "step": 33817
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.23262335485737e-07,
+      "loss": 0.7473,
+      "step": 33818
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.229660188361552e-07,
+      "loss": 0.7355,
+      "step": 33819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.226697703774076e-07,
+      "loss": 0.5834,
+      "step": 33820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.223735901116457e-07,
+      "loss": 0.7477,
+      "step": 33821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.220774780410243e-07,
+      "loss": 0.9,
+      "step": 33822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.217814341676942e-07,
+      "loss": 0.7275,
+      "step": 33823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.214854584938135e-07,
+      "loss": 0.9896,
+      "step": 33824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.211895510215316e-07,
+      "loss": 0.6944,
+      "step": 33825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.208937117530012e-07,
+      "loss": 0.7021,
+      "step": 33826
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.205979406903728e-07,
+      "loss": 0.8977,
+      "step": 33827
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.203022378357981e-07,
+      "loss": 0.7265,
+      "step": 33828
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.200066031914276e-07,
+      "loss": 0.9656,
+      "step": 33829
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.197110367594095e-07,
+      "loss": 0.6509,
+      "step": 33830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.194155385418953e-07,
+      "loss": 0.9595,
+      "step": 33831
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.191201085410326e-07,
+      "loss": 0.6468,
+      "step": 33832
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.188247467589703e-07,
+      "loss": 0.5794,
+      "step": 33833
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.185294531978559e-07,
+      "loss": 0.9278,
+      "step": 33834
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.182342278598363e-07,
+      "loss": 0.5787,
+      "step": 33835
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.179390707470601e-07,
+      "loss": 0.478,
+      "step": 33836
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.176439818616731e-07,
+      "loss": 0.7156,
+      "step": 33837
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.173489612058204e-07,
+      "loss": 0.8158,
+      "step": 33838
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.170540087816478e-07,
+      "loss": 1.1166,
+      "step": 33839
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.167591245913029e-07,
+      "loss": 0.8985,
+      "step": 33840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.164643086369249e-07,
+      "loss": 1.1416,
+      "step": 33841
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.161695609206608e-07,
+      "loss": 0.5731,
+      "step": 33842
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.158748814446546e-07,
+      "loss": 0.6469,
+      "step": 33843
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.155802702110481e-07,
+      "loss": 1.0596,
+      "step": 33844
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.152857272219837e-07,
+      "loss": 0.8705,
+      "step": 33845
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.149912524796031e-07,
+      "loss": 0.6612,
+      "step": 33846
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.146968459860492e-07,
+      "loss": 0.8659,
+      "step": 33847
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.144025077434601e-07,
+      "loss": 0.7615,
+      "step": 33848
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.141082377539798e-07,
+      "loss": 0.6631,
+      "step": 33849
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.138140360197486e-07,
+      "loss": 0.8918,
+      "step": 33850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.135199025429028e-07,
+      "loss": 0.9,
+      "step": 33851
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.132258373255817e-07,
+      "loss": 0.7647,
+      "step": 33852
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.129318403699236e-07,
+      "loss": 0.4933,
+      "step": 33853
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.12637911678069e-07,
+      "loss": 0.8876,
+      "step": 33854
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.123440512521539e-07,
+      "loss": 0.7229,
+      "step": 33855
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.120502590943155e-07,
+      "loss": 0.6949,
+      "step": 33856
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.117565352066901e-07,
+      "loss": 0.7717,
+      "step": 33857
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.114628795914135e-07,
+      "loss": 0.791,
+      "step": 33858
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.111692922506219e-07,
+      "loss": 0.5648,
+      "step": 33859
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.108757731864489e-07,
+      "loss": 1.0285,
+      "step": 33860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.10582322401031e-07,
+      "loss": 0.9248,
+      "step": 33861
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.102889398965006e-07,
+      "loss": 0.6788,
+      "step": 33862
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.099956256749918e-07,
+      "loss": 0.8521,
+      "step": 33863
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.09702379738637e-07,
+      "loss": 0.7318,
+      "step": 33864
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.094092020895703e-07,
+      "loss": 0.9198,
+      "step": 33865
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.0911609272992e-07,
+      "loss": 0.7527,
+      "step": 33866
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.088230516618221e-07,
+      "loss": 0.7862,
+      "step": 33867
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.085300788874071e-07,
+      "loss": 0.9411,
+      "step": 33868
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.082371744088033e-07,
+      "loss": 0.6959,
+      "step": 33869
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.079443382281424e-07,
+      "loss": 0.907,
+      "step": 33870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.076515703475516e-07,
+      "loss": 0.7934,
+      "step": 33871
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.073588707691624e-07,
+      "loss": 0.902,
+      "step": 33872
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.070662394951021e-07,
+      "loss": 0.5914,
+      "step": 33873
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.067736765275001e-07,
+      "loss": 0.7314,
+      "step": 33874
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.064811818684823e-07,
+      "loss": 0.8104,
+      "step": 33875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.061887555201773e-07,
+      "loss": 0.6528,
+      "step": 33876
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.058963974847099e-07,
+      "loss": 0.971,
+      "step": 33877
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.056041077642061e-07,
+      "loss": 0.8541,
+      "step": 33878
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.053118863607943e-07,
+      "loss": 0.7685,
+      "step": 33879
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.050197332765983e-07,
+      "loss": 0.776,
+      "step": 33880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.047276485137421e-07,
+      "loss": 0.7918,
+      "step": 33881
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.044356320743483e-07,
+      "loss": 1.1837,
+      "step": 33882
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.04143683960542e-07,
+      "loss": 0.7799,
+      "step": 33883
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.038518041744456e-07,
+      "loss": 0.9555,
+      "step": 33884
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.035599927181834e-07,
+      "loss": 0.7384,
+      "step": 33885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.032682495938758e-07,
+      "loss": 0.8857,
+      "step": 33886
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.029765748036453e-07,
+      "loss": 0.7432,
+      "step": 33887
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.026849683496128e-07,
+      "loss": 1.0466,
+      "step": 33888
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.023934302338996e-07,
+      "loss": 1.0383,
+      "step": 33889
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.021019604586232e-07,
+      "loss": 1.0964,
+      "step": 33890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.018105590259061e-07,
+      "loss": 0.9729,
+      "step": 33891
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.015192259378666e-07,
+      "loss": 0.8161,
+      "step": 33892
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.01227961196622e-07,
+      "loss": 0.9643,
+      "step": 33893
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.009367648042919e-07,
+      "loss": 0.6058,
+      "step": 33894
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.006456367629932e-07,
+      "loss": 0.7039,
+      "step": 33895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.003545770748432e-07,
+      "loss": 0.6531,
+      "step": 33896
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.000635857419567e-07,
+      "loss": 0.6495,
+      "step": 33897
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.997726627664535e-07,
+      "loss": 0.9073,
+      "step": 33898
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.994818081504483e-07,
+      "loss": 0.7393,
+      "step": 33899
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.99191021896055e-07,
+      "loss": 0.6403,
+      "step": 33900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.989003040053887e-07,
+      "loss": 0.7752,
+      "step": 33901
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.986096544805608e-07,
+      "loss": 0.6631,
+      "step": 33902
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.983190733236888e-07,
+      "loss": 0.7272,
+      "step": 33903
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.980285605368851e-07,
+      "loss": 0.5741,
+      "step": 33904
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.977381161222617e-07,
+      "loss": 0.7334,
+      "step": 33905
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.974477400819312e-07,
+      "loss": 0.8377,
+      "step": 33906
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.97157432418004e-07,
+      "loss": 0.7992,
+      "step": 33907
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.968671931325931e-07,
+      "loss": 0.6491,
+      "step": 33908
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.965770222278056e-07,
+      "loss": 0.7098,
+      "step": 33909
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.962869197057585e-07,
+      "loss": 0.7685,
+      "step": 33910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.959968855685549e-07,
+      "loss": 0.7803,
+      "step": 33911
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.957069198183063e-07,
+      "loss": 0.6152,
+      "step": 33912
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.95417022457121e-07,
+      "loss": 0.851,
+      "step": 33913
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.951271934871073e-07,
+      "loss": 0.9545,
+      "step": 33914
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.948374329103723e-07,
+      "loss": 0.7187,
+      "step": 33915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.945477407290246e-07,
+      "loss": 0.8904,
+      "step": 33916
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.942581169451711e-07,
+      "loss": 0.5797,
+      "step": 33917
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.93968561560917e-07,
+      "loss": 0.5733,
+      "step": 33918
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.936790745783672e-07,
+      "loss": 0.7749,
+      "step": 33919
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.933896559996277e-07,
+      "loss": 0.6626,
+      "step": 33920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.931003058268037e-07,
+      "loss": 1.0051,
+      "step": 33921
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.928110240619988e-07,
+      "loss": 1.0315,
+      "step": 33922
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.925218107073172e-07,
+      "loss": 0.8305,
+      "step": 33923
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.922326657648614e-07,
+      "loss": 0.779,
+      "step": 33924
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.919435892367342e-07,
+      "loss": 0.7593,
+      "step": 33925
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.916545811250384e-07,
+      "loss": 0.7354,
+      "step": 33926
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.913656414318736e-07,
+      "loss": 0.5577,
+      "step": 33927
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.910767701593434e-07,
+      "loss": 0.7193,
+      "step": 33928
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.907879673095485e-07,
+      "loss": 0.7605,
+      "step": 33929
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.904992328845893e-07,
+      "loss": 0.8405,
+      "step": 33930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.902105668865632e-07,
+      "loss": 0.8544,
+      "step": 33931
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.899219693175717e-07,
+      "loss": 1.0937,
+      "step": 33932
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.896334401797099e-07,
+      "loss": 0.7483,
+      "step": 33933
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.893449794750816e-07,
+      "loss": 0.7168,
+      "step": 33934
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.890565872057807e-07,
+      "loss": 0.6502,
+      "step": 33935
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.887682633739056e-07,
+      "loss": 0.8483,
+      "step": 33936
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.884800079815522e-07,
+      "loss": 1.0259,
+      "step": 33937
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.881918210308179e-07,
+      "loss": 0.9107,
+      "step": 33938
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.879037025237977e-07,
+      "loss": 0.8639,
+      "step": 33939
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.876156524625864e-07,
+      "loss": 0.7484,
+      "step": 33940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.873276708492803e-07,
+      "loss": 0.6075,
+      "step": 33941
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.870397576859721e-07,
+      "loss": 0.7871,
+      "step": 33942
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.867519129747557e-07,
+      "loss": 0.7604,
+      "step": 33943
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.86464136717726e-07,
+      "loss": 0.5959,
+      "step": 33944
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.861764289169713e-07,
+      "loss": 0.649,
+      "step": 33945
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.858887895745891e-07,
+      "loss": 0.9808,
+      "step": 33946
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.856012186926696e-07,
+      "loss": 0.7327,
+      "step": 33947
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.853137162733036e-07,
+      "loss": 0.683,
+      "step": 33948
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.850262823185804e-07,
+      "loss": 0.5808,
+      "step": 33949
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.84738916830594e-07,
+      "loss": 1.065,
+      "step": 33950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.844516198114292e-07,
+      "loss": 0.7225,
+      "step": 33951
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.841643912631789e-07,
+      "loss": 0.7323,
+      "step": 33952
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.838772311879304e-07,
+      "loss": 1.0572,
+      "step": 33953
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.835901395877719e-07,
+      "loss": 0.9763,
+      "step": 33954
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.833031164647929e-07,
+      "loss": 0.829,
+      "step": 33955
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.830161618210783e-07,
+      "loss": 1.2021,
+      "step": 33956
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.827292756587166e-07,
+      "loss": 0.5991,
+      "step": 33957
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.824424579797916e-07,
+      "loss": 0.8402,
+      "step": 33958
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.821557087863916e-07,
+      "loss": 0.8128,
+      "step": 33959
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.818690280806038e-07,
+      "loss": 0.5802,
+      "step": 33960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.815824158645078e-07,
+      "loss": 0.737,
+      "step": 33961
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.812958721401906e-07,
+      "loss": 0.6658,
+      "step": 33962
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.810093969097342e-07,
+      "loss": 0.7556,
+      "step": 33963
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.807229901752254e-07,
+      "loss": 0.7327,
+      "step": 33964
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.80436651938745e-07,
+      "loss": 0.8767,
+      "step": 33965
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.801503822023747e-07,
+      "loss": 0.8663,
+      "step": 33966
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.798641809681971e-07,
+      "loss": 0.6,
+      "step": 33967
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.79578048238294e-07,
+      "loss": 0.7444,
+      "step": 33968
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.792919840147449e-07,
+      "loss": 0.9506,
+      "step": 33969
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.790059882996302e-07,
+      "loss": 0.8176,
+      "step": 33970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.787200610950305e-07,
+      "loss": 1.1705,
+      "step": 33971
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.784342024030243e-07,
+      "loss": 0.6347,
+      "step": 33972
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.781484122256919e-07,
+      "loss": 0.7264,
+      "step": 33973
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.778626905651097e-07,
+      "loss": 0.726,
+      "step": 33974
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.775770374233558e-07,
+      "loss": 0.3066,
+      "step": 33975
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.772914528025076e-07,
+      "loss": 0.8882,
+      "step": 33976
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.770059367046432e-07,
+      "loss": 0.709,
+      "step": 33977
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.767204891318378e-07,
+      "loss": 0.5116,
+      "step": 33978
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.764351100861688e-07,
+      "loss": 1.1144,
+      "step": 33979
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.761497995697108e-07,
+      "loss": 0.8283,
+      "step": 33980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.758645575845345e-07,
+      "loss": 0.6537,
+      "step": 33981
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.755793841327195e-07,
+      "loss": 0.8146,
+      "step": 33982
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.752942792163374e-07,
+      "loss": 0.6868,
+      "step": 33983
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.750092428374621e-07,
+      "loss": 0.9595,
+      "step": 33984
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.747242749981674e-07,
+      "loss": 0.854,
+      "step": 33985
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.744393757005229e-07,
+      "loss": 0.8715,
+      "step": 33986
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.741545449466024e-07,
+      "loss": 0.9536,
+      "step": 33987
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.738697827384754e-07,
+      "loss": 0.9519,
+      "step": 33988
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.735850890782158e-07,
+      "loss": 1.2376,
+      "step": 33989
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.73300463967893e-07,
+      "loss": 0.6582,
+      "step": 33990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.730159074095753e-07,
+      "loss": 0.5514,
+      "step": 33991
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 5.727314194053324e-07,
+      "loss": 0.7521,
+      "step": 33992
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.724469999572336e-07,
+      "loss": 0.9867,
+      "step": 33993
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.721626490673449e-07,
+      "loss": 0.4319,
+      "step": 33994
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.718783667377392e-07,
+      "loss": 0.8196,
+      "step": 33995
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.715941529704793e-07,
+      "loss": 0.9469,
+      "step": 33996
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.713100077676347e-07,
+      "loss": 0.765,
+      "step": 33997
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.710259311312715e-07,
+      "loss": 0.7606,
+      "step": 33998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.707419230634537e-07,
+      "loss": 0.7254,
+      "step": 33999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.704579835662482e-07,
+      "loss": 0.876,
+      "step": 34000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.701741126417193e-07,
+      "loss": 0.7582,
+      "step": 34001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.698903102919317e-07,
+      "loss": 0.9439,
+      "step": 34002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.696065765189484e-07,
+      "loss": 0.8518,
+      "step": 34003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.693229113248333e-07,
+      "loss": 0.7528,
+      "step": 34004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.690393147116491e-07,
+      "loss": 0.617,
+      "step": 34005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.687557866814575e-07,
+      "loss": 0.7352,
+      "step": 34006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.684723272363224e-07,
+      "loss": 0.7337,
+      "step": 34007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.681889363783033e-07,
+      "loss": 0.7818,
+      "step": 34008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.679056141094619e-07,
+      "loss": 1.0094,
+      "step": 34009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.676223604318598e-07,
+      "loss": 0.9026,
+      "step": 34010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.673391753475543e-07,
+      "loss": 0.4481,
+      "step": 34011
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.670560588586038e-07,
+      "loss": 0.6207,
+      "step": 34012
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.66773010967071e-07,
+      "loss": 0.9827,
+      "step": 34013
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.664900316750121e-07,
+      "loss": 0.8253,
+      "step": 34014
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.662071209844855e-07,
+      "loss": 0.8061,
+      "step": 34015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.659242788975483e-07,
+      "loss": 0.5162,
+      "step": 34016
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.656415054162579e-07,
+      "loss": 0.8572,
+      "step": 34017
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.653588005426691e-07,
+      "loss": 0.7351,
+      "step": 34018
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.650761642788371e-07,
+      "loss": 0.7329,
+      "step": 34019
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.647935966268225e-07,
+      "loss": 1.1161,
+      "step": 34020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.645110975886758e-07,
+      "loss": 0.9638,
+      "step": 34021
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.642286671664509e-07,
+      "loss": 0.9716,
+      "step": 34022
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.639463053622041e-07,
+      "loss": 0.774,
+      "step": 34023
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.636640121779858e-07,
+      "loss": 0.7644,
+      "step": 34024
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.63381787615852e-07,
+      "loss": 0.8216,
+      "step": 34025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.630996316778537e-07,
+      "loss": 0.9839,
+      "step": 34026
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.628175443660433e-07,
+      "loss": 0.7492,
+      "step": 34027
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.625355256824705e-07,
+      "loss": 0.8696,
+      "step": 34028
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.62253575629188e-07,
+      "loss": 0.8522,
+      "step": 34029
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.619716942082464e-07,
+      "loss": 0.9052,
+      "step": 34030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.616898814216931e-07,
+      "loss": 1.0131,
+      "step": 34031
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.614081372715807e-07,
+      "loss": 0.8518,
+      "step": 34032
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.611264617599555e-07,
+      "loss": 0.55,
+      "step": 34033
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.608448548888667e-07,
+      "loss": 0.9892,
+      "step": 34034
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.60563316660363e-07,
+      "loss": 0.5219,
+      "step": 34035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.602818470764903e-07,
+      "loss": 0.7237,
+      "step": 34036
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.60000446139296e-07,
+      "loss": 1.0199,
+      "step": 34037
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.597191138508273e-07,
+      "loss": 0.8218,
+      "step": 34038
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.594378502131292e-07,
+      "loss": 0.8631,
+      "step": 34039
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.591566552282502e-07,
+      "loss": 0.6519,
+      "step": 34040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.588755288982295e-07,
+      "loss": 0.6565,
+      "step": 34041
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.585944712251135e-07,
+      "loss": 0.9189,
+      "step": 34042
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.583134822109482e-07,
+      "loss": 0.687,
+      "step": 34043
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.580325618577764e-07,
+      "loss": 0.7484,
+      "step": 34044
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.577517101676399e-07,
+      "loss": 0.8725,
+      "step": 34045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.574709271425805e-07,
+      "loss": 0.8779,
+      "step": 34046
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.571902127846429e-07,
+      "loss": 0.7329,
+      "step": 34047
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.569095670958657e-07,
+      "loss": 0.6711,
+      "step": 34048
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.566289900782895e-07,
+      "loss": 0.8088,
+      "step": 34049
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.563484817339581e-07,
+      "loss": 0.5192,
+      "step": 34050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.56068042064909e-07,
+      "loss": 0.9214,
+      "step": 34051
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.557876710731813e-07,
+      "loss": 0.8041,
+      "step": 34052
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.555073687608136e-07,
+      "loss": 0.9919,
+      "step": 34053
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.552271351298466e-07,
+      "loss": 0.8346,
+      "step": 34054
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.54946970182314e-07,
+      "loss": 0.9011,
+      "step": 34055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.546668739202576e-07,
+      "loss": 0.8534,
+      "step": 34056
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.543868463457124e-07,
+      "loss": 0.5642,
+      "step": 34057
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.541068874607158e-07,
+      "loss": 0.64,
+      "step": 34058
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.538269972673027e-07,
+      "loss": 0.6196,
+      "step": 34059
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.535471757675082e-07,
+      "loss": 0.8431,
+      "step": 34060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.532674229633683e-07,
+      "loss": 0.792,
+      "step": 34061
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.52987738856916e-07,
+      "loss": 0.693,
+      "step": 34062
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.527081234501863e-07,
+      "loss": 0.8348,
+      "step": 34063
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.52428576745212e-07,
+      "loss": 0.8739,
+      "step": 34064
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.521490987440259e-07,
+      "loss": 0.9081,
+      "step": 34065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.518696894486608e-07,
+      "loss": 0.4184,
+      "step": 34066
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.515903488611473e-07,
+      "loss": 0.7822,
+      "step": 34067
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.513110769835206e-07,
+      "loss": 1.121,
+      "step": 34068
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.510318738178078e-07,
+      "loss": 1.022,
+      "step": 34069
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.507527393660416e-07,
+      "loss": 0.7264,
+      "step": 34070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.504736736302496e-07,
+      "loss": 0.6079,
+      "step": 34071
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.501946766124633e-07,
+      "loss": 0.7648,
+      "step": 34072
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.499157483147078e-07,
+      "loss": 0.7509,
+      "step": 34073
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.496368887390168e-07,
+      "loss": 0.686,
+      "step": 34074
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.493580978874158e-07,
+      "loss": 0.8707,
+      "step": 34075
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.490793757619318e-07,
+      "loss": 0.6269,
+      "step": 34076
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.488007223645919e-07,
+      "loss": 0.8842,
+      "step": 34077
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.485221376974237e-07,
+      "loss": 0.855,
+      "step": 34078
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.482436217624509e-07,
+      "loss": 0.7499,
+      "step": 34079
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.47965174561701e-07,
+      "loss": 0.8548,
+      "step": 34080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.476867960971977e-07,
+      "loss": 0.9686,
+      "step": 34081
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.47408486370965e-07,
+      "loss": 0.6943,
+      "step": 34082
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.47130245385028e-07,
+      "loss": 1.1077,
+      "step": 34083
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.468520731414095e-07,
+      "loss": 0.9245,
+      "step": 34084
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.465739696421312e-07,
+      "loss": 0.7544,
+      "step": 34085
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.462959348892183e-07,
+      "loss": 0.5969,
+      "step": 34086
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.460179688846912e-07,
+      "loss": 0.9105,
+      "step": 34087
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.457400716305705e-07,
+      "loss": 0.6402,
+      "step": 34088
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.454622431288781e-07,
+      "loss": 0.6913,
+      "step": 34089
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.451844833816366e-07,
+      "loss": 0.7384,
+      "step": 34090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.449067923908602e-07,
+      "loss": 0.9473,
+      "step": 34091
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.446291701585726e-07,
+      "loss": 0.969,
+      "step": 34092
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.443516166867912e-07,
+      "loss": 0.6577,
+      "step": 34093
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.440741319775356e-07,
+      "loss": 0.7908,
+      "step": 34094
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.437967160328228e-07,
+      "loss": 0.7,
+      "step": 34095
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.435193688546702e-07,
+      "loss": 0.9394,
+      "step": 34096
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.432420904450952e-07,
+      "loss": 0.7804,
+      "step": 34097
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.429648808061128e-07,
+      "loss": 0.7686,
+      "step": 34098
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.426877399397401e-07,
+      "loss": 0.6927,
+      "step": 34099
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.424106678479946e-07,
+      "loss": 0.9043,
+      "step": 34100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.421336645328878e-07,
+      "loss": 0.8762,
+      "step": 34101
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.418567299964351e-07,
+      "loss": 0.8165,
+      "step": 34102
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.415798642406489e-07,
+      "loss": 0.9093,
+      "step": 34103
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.413030672675456e-07,
+      "loss": 0.6248,
+      "step": 34104
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.41026339079137e-07,
+      "loss": 0.7744,
+      "step": 34105
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.407496796774358e-07,
+      "loss": 0.9033,
+      "step": 34106
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.404730890644527e-07,
+      "loss": 0.9153,
+      "step": 34107
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.401965672421994e-07,
+      "loss": 0.686,
+      "step": 34108
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.399201142126876e-07,
+      "loss": 0.8677,
+      "step": 34109
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.396437299779278e-07,
+      "loss": 0.845,
+      "step": 34110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.393674145399286e-07,
+      "loss": 0.8321,
+      "step": 34111
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.390911679007005e-07,
+      "loss": 0.726,
+      "step": 34112
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.388149900622508e-07,
+      "loss": 0.9472,
+      "step": 34113
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.385388810265902e-07,
+      "loss": 0.8553,
+      "step": 34114
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.382628407957258e-07,
+      "loss": 0.9961,
+      "step": 34115
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.379868693716628e-07,
+      "loss": 0.6934,
+      "step": 34116
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.377109667564118e-07,
+      "loss": 0.9133,
+      "step": 34117
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.374351329519778e-07,
+      "loss": 0.7108,
+      "step": 34118
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.371593679603648e-07,
+      "loss": 0.8767,
+      "step": 34119
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.368836717835824e-07,
+      "loss": 0.4636,
+      "step": 34120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.366080444236299e-07,
+      "loss": 0.8233,
+      "step": 34121
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.363324858825158e-07,
+      "loss": 0.9335,
+      "step": 34122
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.36056996162243e-07,
+      "loss": 0.8474,
+      "step": 34123
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.357815752648144e-07,
+      "loss": 0.7806,
+      "step": 34124
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.355062231922326e-07,
+      "loss": 0.9554,
+      "step": 34125
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.352309399465017e-07,
+      "loss": 0.5898,
+      "step": 34126
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.349557255296212e-07,
+      "loss": 0.6295,
+      "step": 34127
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.346805799435928e-07,
+      "loss": 0.8472,
+      "step": 34128
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.344055031904205e-07,
+      "loss": 0.9482,
+      "step": 34129
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.341304952721027e-07,
+      "loss": 0.9573,
+      "step": 34130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.338555561906389e-07,
+      "loss": 0.8356,
+      "step": 34131
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.335806859480274e-07,
+      "loss": 0.894,
+      "step": 34132
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.333058845462691e-07,
+      "loss": 0.729,
+      "step": 34133
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.330311519873609e-07,
+      "loss": 1.0837,
+      "step": 34134
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.327564882733017e-07,
+      "loss": 0.5999,
+      "step": 34135
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.324818934060894e-07,
+      "loss": 0.7393,
+      "step": 34136
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.322073673877193e-07,
+      "loss": 0.6747,
+      "step": 34137
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.319329102201898e-07,
+      "loss": 0.6996,
+      "step": 34138
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.316585219054949e-07,
+      "loss": 0.4611,
+      "step": 34139
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.313842024456306e-07,
+      "loss": 0.8268,
+      "step": 34140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.311099518425911e-07,
+      "loss": 0.7759,
+      "step": 34141
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.308357700983724e-07,
+      "loss": 0.6949,
+      "step": 34142
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.305616572149674e-07,
+      "loss": 0.7039,
+      "step": 34143
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.302876131943701e-07,
+      "loss": 0.6806,
+      "step": 34144
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.300136380385723e-07,
+      "loss": 0.7255,
+      "step": 34145
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.297397317495656e-07,
+      "loss": 0.8054,
+      "step": 34146
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.294658943293452e-07,
+      "loss": 0.6392,
+      "step": 34147
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.291921257798993e-07,
+      "loss": 0.9394,
+      "step": 34148
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.28918426103221e-07,
+      "loss": 0.9212,
+      "step": 34149
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.286447953013007e-07,
+      "loss": 0.7819,
+      "step": 34150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.28371233376126e-07,
+      "loss": 0.4837,
+      "step": 34151
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.28097740329685e-07,
+      "loss": 0.7438,
+      "step": 34152
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.278243161639718e-07,
+      "loss": 0.7953,
+      "step": 34153
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.275509608809715e-07,
+      "loss": 0.9963,
+      "step": 34154
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.272776744826724e-07,
+      "loss": 0.9786,
+      "step": 34155
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.270044569710619e-07,
+      "loss": 0.826,
+      "step": 34156
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.267313083481274e-07,
+      "loss": 0.8507,
+      "step": 34157
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.264582286158548e-07,
+      "loss": 0.8188,
+      "step": 34158
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.261852177762283e-07,
+      "loss": 0.7253,
+      "step": 34159
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.259122758312373e-07,
+      "loss": 0.5424,
+      "step": 34160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.256394027828637e-07,
+      "loss": 0.5918,
+      "step": 34161
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.253665986330925e-07,
+      "loss": 0.8557,
+      "step": 34162
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.250938633839065e-07,
+      "loss": 0.5791,
+      "step": 34163
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.248211970372896e-07,
+      "loss": 0.6686,
+      "step": 34164
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.24548599595226e-07,
+      "loss": 0.528,
+      "step": 34165
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.242760710596984e-07,
+      "loss": 0.7799,
+      "step": 34166
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.240036114326863e-07,
+      "loss": 0.9824,
+      "step": 34167
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.237312207161726e-07,
+      "loss": 0.8606,
+      "step": 34168
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.23458898912137e-07,
+      "loss": 1.0384,
+      "step": 34169
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.231866460225621e-07,
+      "loss": 1.0286,
+      "step": 34170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.229144620494253e-07,
+      "loss": 0.6876,
+      "step": 34171
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.226423469947073e-07,
+      "loss": 0.798,
+      "step": 34172
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.223703008603864e-07,
+      "loss": 0.8458,
+      "step": 34173
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.220983236484412e-07,
+      "loss": 0.6434,
+      "step": 34174
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.218264153608488e-07,
+      "loss": 0.7231,
+      "step": 34175
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.215545759995877e-07,
+      "loss": 0.4025,
+      "step": 34176
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.21282805566633e-07,
+      "loss": 0.5847,
+      "step": 34177
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.210111040639642e-07,
+      "loss": 0.7277,
+      "step": 34178
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.207394714935554e-07,
+      "loss": 0.6491,
+      "step": 34179
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.204679078573826e-07,
+      "loss": 0.574,
+      "step": 34180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.201964131574177e-07,
+      "loss": 0.8617,
+      "step": 34181
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.199249873956369e-07,
+      "loss": 0.801,
+      "step": 34182
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.196536305740152e-07,
+      "loss": 0.7505,
+      "step": 34183
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.193823426945255e-07,
+      "loss": 0.8223,
+      "step": 34184
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.191111237591406e-07,
+      "loss": 0.9141,
+      "step": 34185
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.188399737698313e-07,
+      "loss": 0.9281,
+      "step": 34186
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.185688927285715e-07,
+      "loss": 0.7819,
+      "step": 34187
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.182978806373307e-07,
+      "loss": 0.6644,
+      "step": 34188
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.180269374980796e-07,
+      "loss": 0.7004,
+      "step": 34189
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.177560633127932e-07,
+      "loss": 0.5781,
+      "step": 34190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.174852580834355e-07,
+      "loss": 0.5452,
+      "step": 34191
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.172145218119784e-07,
+      "loss": 0.8838,
+      "step": 34192
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.169438545003891e-07,
+      "loss": 0.8309,
+      "step": 34193
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.166732561506383e-07,
+      "loss": 0.7323,
+      "step": 34194
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.16402726764691e-07,
+      "loss": 0.7557,
+      "step": 34195
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.161322663445178e-07,
+      "loss": 0.9418,
+      "step": 34196
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.158618748920841e-07,
+      "loss": 0.9743,
+      "step": 34197
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.155915524093557e-07,
+      "loss": 0.7497,
+      "step": 34198
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.153212988982992e-07,
+      "loss": 0.7748,
+      "step": 34199
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.150511143608782e-07,
+      "loss": 0.7761,
+      "step": 34200
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.147809987990604e-07,
+      "loss": 0.8097,
+      "step": 34201
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.145109522148073e-07,
+      "loss": 1.0038,
+      "step": 34202
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.14240974610084e-07,
+      "loss": 0.9094,
+      "step": 34203
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.139710659868546e-07,
+      "loss": 0.5371,
+      "step": 34204
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.137012263470797e-07,
+      "loss": 0.7693,
+      "step": 34205
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.134314556927245e-07,
+      "loss": 0.5924,
+      "step": 34206
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.131617540257461e-07,
+      "loss": 0.9305,
+      "step": 34207
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.128921213481119e-07,
+      "loss": 1.0315,
+      "step": 34208
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.126225576617782e-07,
+      "loss": 0.6983,
+      "step": 34209
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.123530629687091e-07,
+      "loss": 0.6787,
+      "step": 34210
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.120836372708604e-07,
+      "loss": 0.6919,
+      "step": 34211
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.118142805701942e-07,
+      "loss": 0.7441,
+      "step": 34212
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.115449928686656e-07,
+      "loss": 0.8412,
+      "step": 34213
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.112757741682372e-07,
+      "loss": 0.7341,
+      "step": 34214
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.110066244708645e-07,
+      "loss": 0.8423,
+      "step": 34215
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.107375437785056e-07,
+      "loss": 0.6597,
+      "step": 34216
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.104685320931179e-07,
+      "loss": 0.901,
+      "step": 34217
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.101995894166556e-07,
+      "loss": 0.9593,
+      "step": 34218
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.099307157510769e-07,
+      "loss": 0.5393,
+      "step": 34219
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.096619110983348e-07,
+      "loss": 0.8463,
+      "step": 34220
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.093931754603865e-07,
+      "loss": 0.7215,
+      "step": 34221
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.091245088391838e-07,
+      "loss": 0.546,
+      "step": 34222
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.08855911236682e-07,
+      "loss": 0.8458,
+      "step": 34223
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.085873826548337e-07,
+      "loss": 0.9247,
+      "step": 34224
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.08318923095591e-07,
+      "loss": 0.5052,
+      "step": 34225
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.080505325609075e-07,
+      "loss": 0.8998,
+      "step": 34226
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.077822110527353e-07,
+      "loss": 0.8478,
+      "step": 34227
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.07513958573026e-07,
+      "loss": 0.5392,
+      "step": 34228
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.07245775123728e-07,
+      "loss": 0.6556,
+      "step": 34229
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.069776607067944e-07,
+      "loss": 0.7294,
+      "step": 34230
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.067096153241713e-07,
+      "loss": 0.6633,
+      "step": 34231
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.064416389778104e-07,
+      "loss": 0.5627,
+      "step": 34232
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.061737316696602e-07,
+      "loss": 0.5477,
+      "step": 34233
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.059058934016703e-07,
+      "loss": 0.8186,
+      "step": 34234
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.056381241757857e-07,
+      "loss": 1.057,
+      "step": 34235
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.05370423993955e-07,
+      "loss": 0.8316,
+      "step": 34236
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.051027928581254e-07,
+      "loss": 1.0059,
+      "step": 34237
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.04835230770242e-07,
+      "loss": 0.7984,
+      "step": 34238
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.045677377322511e-07,
+      "loss": 0.8592,
+      "step": 34239
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.043003137461011e-07,
+      "loss": 0.7635,
+      "step": 34240
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.040329588137304e-07,
+      "loss": 0.9371,
+      "step": 34241
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.037656729370876e-07,
+      "loss": 0.876,
+      "step": 34242
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.034984561181144e-07,
+      "loss": 0.7867,
+      "step": 34243
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.032313083587559e-07,
+      "loss": 0.9147,
+      "step": 34244
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.029642296609538e-07,
+      "loss": 0.802,
+      "step": 34245
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.0269722002665e-07,
+      "loss": 1.0,
+      "step": 34246
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.024302794577873e-07,
+      "loss": 0.9004,
+      "step": 34247
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.021634079563065e-07,
+      "loss": 0.7385,
+      "step": 34248
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.018966055241481e-07,
+      "loss": 0.8783,
+      "step": 34249
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.016298721632518e-07,
+      "loss": 0.922,
+      "step": 34250
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.013632078755582e-07,
+      "loss": 1.0553,
+      "step": 34251
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.010966126630057e-07,
+      "loss": 0.94,
+      "step": 34252
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.008300865275339e-07,
+      "loss": 0.8208,
+      "step": 34253
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.005636294710802e-07,
+      "loss": 0.9444,
+      "step": 34254
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.002972414955831e-07,
+      "loss": 0.8116,
+      "step": 34255
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 5.000309226029776e-07,
+      "loss": 0.7736,
+      "step": 34256
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.997646727952043e-07,
+      "loss": 0.7486,
+      "step": 34257
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.994984920741964e-07,
+      "loss": 0.934,
+      "step": 34258
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.992323804418908e-07,
+      "loss": 0.6015,
+      "step": 34259
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.98966337900224e-07,
+      "loss": 0.8763,
+      "step": 34260
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.987003644511257e-07,
+      "loss": 0.5659,
+      "step": 34261
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.984344600965352e-07,
+      "loss": 0.7168,
+      "step": 34262
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.981686248383844e-07,
+      "loss": 0.9494,
+      "step": 34263
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.979028586786061e-07,
+      "loss": 0.7267,
+      "step": 34264
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.976371616191344e-07,
+      "loss": 0.6593,
+      "step": 34265
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.973715336619001e-07,
+      "loss": 0.5035,
+      "step": 34266
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.971059748088347e-07,
+      "loss": 0.7123,
+      "step": 34267
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.968404850618691e-07,
+      "loss": 0.9503,
+      "step": 34268
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.96575064422935e-07,
+      "loss": 1.0623,
+      "step": 34269
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.963097128939653e-07,
+      "loss": 0.4541,
+      "step": 34270
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.960444304768841e-07,
+      "loss": 0.6842,
+      "step": 34271
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.957792171736242e-07,
+      "loss": 0.6974,
+      "step": 34272
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.95514072986113e-07,
+      "loss": 0.746,
+      "step": 34273
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.952489979162778e-07,
+      "loss": 0.8852,
+      "step": 34274
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.949839919660481e-07,
+      "loss": 0.8001,
+      "step": 34275
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.947190551373504e-07,
+      "loss": 0.6987,
+      "step": 34276
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.944541874321118e-07,
+      "loss": 0.6069,
+      "step": 34277
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.941893888522576e-07,
+      "loss": 0.8484,
+      "step": 34278
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.939246593997138e-07,
+      "loss": 0.7472,
+      "step": 34279
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.936599990764046e-07,
+      "loss": 0.9321,
+      "step": 34280
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.933954078842562e-07,
+      "loss": 0.8025,
+      "step": 34281
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.931308858251926e-07,
+      "loss": 0.6801,
+      "step": 34282
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.928664329011368e-07,
+      "loss": 0.6353,
+      "step": 34283
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.926020491140116e-07,
+      "loss": 0.8161,
+      "step": 34284
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.923377344657398e-07,
+      "loss": 0.8328,
+      "step": 34285
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.920734889582435e-07,
+      "loss": 0.9896,
+      "step": 34286
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.918093125934454e-07,
+      "loss": 0.66,
+      "step": 34287
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.915452053732661e-07,
+      "loss": 0.8545,
+      "step": 34288
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.912811672996253e-07,
+      "loss": 0.8705,
+      "step": 34289
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.91017198374445e-07,
+      "loss": 0.4281,
+      "step": 34290
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.907532985996432e-07,
+      "loss": 0.764,
+      "step": 34291
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.904894679771377e-07,
+      "loss": 0.6751,
+      "step": 34292
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.9022570650885e-07,
+      "loss": 1.0171,
+      "step": 34293
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.899620141966965e-07,
+      "loss": 1.0041,
+      "step": 34294
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.896983910425956e-07,
+      "loss": 0.7573,
+      "step": 34295
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.894348370484648e-07,
+      "loss": 0.7999,
+      "step": 34296
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.891713522162189e-07,
+      "loss": 0.7066,
+      "step": 34297
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.889079365477756e-07,
+      "loss": 0.7453,
+      "step": 34298
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.88644590045051e-07,
+      "loss": 0.8881,
+      "step": 34299
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.88381312709958e-07,
+      "loss": 0.7451,
+      "step": 34300
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.881181045444116e-07,
+      "loss": 0.9702,
+      "step": 34301
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.878549655503285e-07,
+      "loss": 1.0125,
+      "step": 34302
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.875918957296189e-07,
+      "loss": 0.9067,
+      "step": 34303
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.873288950841971e-07,
+      "loss": 0.7751,
+      "step": 34304
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.870659636159758e-07,
+      "loss": 0.5571,
+      "step": 34305
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.868031013268682e-07,
+      "loss": 0.9206,
+      "step": 34306
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.865403082187847e-07,
+      "loss": 0.934,
+      "step": 34307
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.862775842936373e-07,
+      "loss": 0.9923,
+      "step": 34308
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.860149295533356e-07,
+      "loss": 0.9139,
+      "step": 34309
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.85752343999788e-07,
+      "loss": 1.0304,
+      "step": 34310
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.854898276349074e-07,
+      "loss": 0.7515,
+      "step": 34311
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.852273804606011e-07,
+      "loss": 0.8742,
+      "step": 34312
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.849650024787778e-07,
+      "loss": 1.0289,
+      "step": 34313
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.847026936913457e-07,
+      "loss": 0.7427,
+      "step": 34314
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.844404541002123e-07,
+      "loss": 0.6677,
+      "step": 34315
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.841782837072851e-07,
+      "loss": 0.9597,
+      "step": 34316
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.839161825144678e-07,
+      "loss": 0.9376,
+      "step": 34317
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.836541505236714e-07,
+      "loss": 0.5197,
+      "step": 34318
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.833921877367986e-07,
+      "loss": 0.8487,
+      "step": 34319
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.831302941557537e-07,
+      "loss": 0.7835,
+      "step": 34320
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.828684697824426e-07,
+      "loss": 0.7623,
+      "step": 34321
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.826067146187674e-07,
+      "loss": 0.914,
+      "step": 34322
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.823450286666342e-07,
+      "loss": 0.8899,
+      "step": 34323
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.820834119279449e-07,
+      "loss": 1.0019,
+      "step": 34324
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.818218644046024e-07,
+      "loss": 0.9109,
+      "step": 34325
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.815603860985074e-07,
+      "loss": 1.1971,
+      "step": 34326
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.812989770115628e-07,
+      "loss": 0.9598,
+      "step": 34327
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.810376371456693e-07,
+      "loss": 0.6575,
+      "step": 34328
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.807763665027265e-07,
+      "loss": 0.6644,
+      "step": 34329
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.805151650846362e-07,
+      "loss": 0.7291,
+      "step": 34330
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.802540328932959e-07,
+      "loss": 0.7675,
+      "step": 34331
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.79992969930605e-07,
+      "loss": 0.7035,
+      "step": 34332
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.797319761984631e-07,
+      "loss": 0.8194,
+      "step": 34333
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.794710516987667e-07,
+      "loss": 0.8424,
+      "step": 34334
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.79210196433414e-07,
+      "loss": 0.6262,
+      "step": 34335
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.789494104043024e-07,
+      "loss": 0.6602,
+      "step": 34336
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.786886936133283e-07,
+      "loss": 0.7855,
+      "step": 34337
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.784280460623869e-07,
+      "loss": 0.9522,
+      "step": 34338
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.781674677533754e-07,
+      "loss": 0.3523,
+      "step": 34339
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.779069586881857e-07,
+      "loss": 0.6528,
+      "step": 34340
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.776465188687141e-07,
+      "loss": 0.744,
+      "step": 34341
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.773861482968556e-07,
+      "loss": 0.6567,
+      "step": 34342
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.771258469745022e-07,
+      "loss": 0.6229,
+      "step": 34343
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.768656149035467e-07,
+      "loss": 1.0647,
+      "step": 34344
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7660545208588204e-07,
+      "loss": 0.7153,
+      "step": 34345
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7634535852340013e-07,
+      "loss": 0.6477,
+      "step": 34346
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7608533421799053e-07,
+      "loss": 0.8375,
+      "step": 34347
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.758253791715484e-07,
+      "loss": 0.8655,
+      "step": 34348
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7556549338596213e-07,
+      "loss": 0.686,
+      "step": 34349
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7530567686312035e-07,
+      "loss": 0.5724,
+      "step": 34350
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7504592960491367e-07,
+      "loss": 0.7964,
+      "step": 34351
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7478625161323066e-07,
+      "loss": 0.8154,
+      "step": 34352
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7452664288995867e-07,
+      "loss": 0.7173,
+      "step": 34353
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.742671034369872e-07,
+      "loss": 0.6672,
+      "step": 34354
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.740076332562038e-07,
+      "loss": 0.7952,
+      "step": 34355
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7374823234949464e-07,
+      "loss": 0.8198,
+      "step": 34356
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7348890071874596e-07,
+      "loss": 0.8474,
+      "step": 34357
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.73229638365843e-07,
+      "loss": 0.7738,
+      "step": 34358
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7297044529267313e-07,
+      "loss": 0.8238,
+      "step": 34359
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7271132150112034e-07,
+      "loss": 0.9648,
+      "step": 34360
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7245226699306867e-07,
+      "loss": 0.7647,
+      "step": 34361
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7219328177040223e-07,
+      "loss": 0.8262,
+      "step": 34362
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7193436583500394e-07,
+      "loss": 0.5945,
+      "step": 34363
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7167551918875674e-07,
+      "loss": 0.8582,
+      "step": 34364
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7141674183354247e-07,
+      "loss": 0.7788,
+      "step": 34365
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7115803377124404e-07,
+      "loss": 0.8549,
+      "step": 34366
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7089939500374435e-07,
+      "loss": 0.8086,
+      "step": 34367
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.706408255329209e-07,
+      "loss": 0.7593,
+      "step": 34368
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.7038232536065765e-07,
+      "loss": 1.3968,
+      "step": 34369
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.701238944888309e-07,
+      "loss": 0.811,
+      "step": 34370
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 4.698655329193202e-07,
+      "loss": 0.6647,
+      "step": 34371
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6960724065400645e-07,
+      "loss": 0.8573,
+      "step": 34372
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.693490176947679e-07,
+      "loss": 1.0161,
+      "step": 34373
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6909086404348104e-07,
+      "loss": 0.6988,
+      "step": 34374
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.688327797020242e-07,
+      "loss": 0.9818,
+      "step": 34375
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.685747646722738e-07,
+      "loss": 0.8458,
+      "step": 34376
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.68316818956106e-07,
+      "loss": 0.6752,
+      "step": 34377
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6805894255539495e-07,
+      "loss": 0.8428,
+      "step": 34378
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.678011354720213e-07,
+      "loss": 0.8289,
+      "step": 34379
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6754339770785474e-07,
+      "loss": 0.6281,
+      "step": 34380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6728572926477145e-07,
+      "loss": 0.7602,
+      "step": 34381
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6702813014464444e-07,
+      "loss": 0.6013,
+      "step": 34382
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.667706003493455e-07,
+      "loss": 0.7094,
+      "step": 34383
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.665131398807521e-07,
+      "loss": 0.7779,
+      "step": 34384
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6625574874073267e-07,
+      "loss": 0.6527,
+      "step": 34385
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6599842693116124e-07,
+      "loss": 0.8037,
+      "step": 34386
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6574117445390644e-07,
+      "loss": 0.8862,
+      "step": 34387
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6548399131084223e-07,
+      "loss": 0.8505,
+      "step": 34388
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.65226877503836e-07,
+      "loss": 0.955,
+      "step": 34389
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6496983303475963e-07,
+      "loss": 0.5934,
+      "step": 34390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6471285790548047e-07,
+      "loss": 0.8284,
+      "step": 34391
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.644559521178693e-07,
+      "loss": 0.655,
+      "step": 34392
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.641991156737924e-07,
+      "loss": 0.6612,
+      "step": 34393
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6394234857511934e-07,
+      "loss": 0.9136,
+      "step": 34394
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.636856508237164e-07,
+      "loss": 0.864,
+      "step": 34395
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6342902242144883e-07,
+      "loss": 1.0618,
+      "step": 34396
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.631724633701851e-07,
+      "loss": 0.8048,
+      "step": 34397
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6291597367179143e-07,
+      "loss": 0.7969,
+      "step": 34398
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6265955332813194e-07,
+      "loss": 0.6539,
+      "step": 34399
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.624032023410707e-07,
+      "loss": 1.0092,
+      "step": 34400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.621469207124729e-07,
+      "loss": 0.6887,
+      "step": 34401
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6189070844419926e-07,
+      "loss": 0.7239,
+      "step": 34402
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6163456553811826e-07,
+      "loss": 0.8164,
+      "step": 34403
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.613784919960895e-07,
+      "loss": 0.7556,
+      "step": 34404
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.611224878199749e-07,
+      "loss": 0.8779,
+      "step": 34405
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.608665530116374e-07,
+      "loss": 0.9548,
+      "step": 34406
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.606106875729377e-07,
+      "loss": 0.6097,
+      "step": 34407
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6035489150573656e-07,
+      "loss": 0.6827,
+      "step": 34408
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.6009916481189356e-07,
+      "loss": 0.8429,
+      "step": 34409
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5984350749326835e-07,
+      "loss": 0.5612,
+      "step": 34410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5958791955172166e-07,
+      "loss": 0.85,
+      "step": 34411
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.593324009891098e-07,
+      "loss": 0.7114,
+      "step": 34412
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.590769518072935e-07,
+      "loss": 0.863,
+      "step": 34413
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.588215720081268e-07,
+      "loss": 0.8423,
+      "step": 34414
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.585662615934716e-07,
+      "loss": 0.6726,
+      "step": 34415
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5831102056518085e-07,
+      "loss": 0.8193,
+      "step": 34416
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.58055848925113e-07,
+      "loss": 0.8938,
+      "step": 34417
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.578007466751222e-07,
+      "loss": 0.8725,
+      "step": 34418
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.575457138170647e-07,
+      "loss": 0.7411,
+      "step": 34419
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5729075035279235e-07,
+      "loss": 0.9633,
+      "step": 34420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.570358562841626e-07,
+      "loss": 0.9368,
+      "step": 34421
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.567810316130283e-07,
+      "loss": 0.8467,
+      "step": 34422
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.565262763412415e-07,
+      "loss": 0.6097,
+      "step": 34423
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5627159047065496e-07,
+      "loss": 0.7096,
+      "step": 34424
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5601697400312175e-07,
+      "loss": 1.0137,
+      "step": 34425
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.557624269404937e-07,
+      "loss": 0.8103,
+      "step": 34426
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.555079492846193e-07,
+      "loss": 0.7104,
+      "step": 34427
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5525354103735154e-07,
+      "loss": 0.7698,
+      "step": 34428
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.549992022005423e-07,
+      "loss": 0.8741,
+      "step": 34429
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.547449327760367e-07,
+      "loss": 0.6914,
+      "step": 34430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5449073276568665e-07,
+      "loss": 0.5564,
+      "step": 34431
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.542366021713385e-07,
+      "loss": 0.5619,
+      "step": 34432
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5398254099484284e-07,
+      "loss": 0.8605,
+      "step": 34433
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.537285492380461e-07,
+      "loss": 1.0571,
+      "step": 34434
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5347462690279566e-07,
+      "loss": 1.0322,
+      "step": 34435
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.532207739909378e-07,
+      "loss": 1.0931,
+      "step": 34436
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.529669905043188e-07,
+      "loss": 0.7249,
+      "step": 34437
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.527132764447839e-07,
+      "loss": 1.0255,
+      "step": 34438
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.524596318141783e-07,
+      "loss": 0.6035,
+      "step": 34439
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5220605661434605e-07,
+      "loss": 0.745,
+      "step": 34440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.519525508471323e-07,
+      "loss": 0.6463,
+      "step": 34441
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.51699114514379e-07,
+      "loss": 0.7246,
+      "step": 34442
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5144574761793014e-07,
+      "loss": 1.0922,
+      "step": 34443
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5119245015962876e-07,
+      "loss": 0.6745,
+      "step": 34444
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5093922214131556e-07,
+      "loss": 0.8712,
+      "step": 34445
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5068606356483247e-07,
+      "loss": 0.7936,
+      "step": 34446
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.504329744320224e-07,
+      "loss": 0.6898,
+      "step": 34447
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.5017995474472277e-07,
+      "loss": 0.748,
+      "step": 34448
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4992700450477657e-07,
+      "loss": 1.0643,
+      "step": 34449
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.496741237140201e-07,
+      "loss": 0.7421,
+      "step": 34450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4942131237429407e-07,
+      "loss": 0.9735,
+      "step": 34451
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.491685704874371e-07,
+      "loss": 0.9387,
+      "step": 34452
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4891589805528766e-07,
+      "loss": 0.6685,
+      "step": 34453
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4866329507968207e-07,
+      "loss": 0.795,
+      "step": 34454
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4841076156245665e-07,
+      "loss": 0.7896,
+      "step": 34455
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4815829750544994e-07,
+      "loss": 0.7274,
+      "step": 34456
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4790590291049483e-07,
+      "loss": 0.7061,
+      "step": 34457
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4765357777943106e-07,
+      "loss": 0.7295,
+      "step": 34458
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.474013221140916e-07,
+      "loss": 0.9307,
+      "step": 34459
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4714913591630935e-07,
+      "loss": 1.004,
+      "step": 34460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.468970191879196e-07,
+      "loss": 1.0431,
+      "step": 34461
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.466449719307553e-07,
+      "loss": 0.67,
+      "step": 34462
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4639299414664827e-07,
+      "loss": 0.8281,
+      "step": 34463
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4614108583743486e-07,
+      "loss": 0.9887,
+      "step": 34464
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.458892470049425e-07,
+      "loss": 1.1813,
+      "step": 34465
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.456374776510064e-07,
+      "loss": 0.6376,
+      "step": 34466
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4538577777745393e-07,
+      "loss": 0.7203,
+      "step": 34467
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4513414738611814e-07,
+      "loss": 0.9841,
+      "step": 34468
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.448825864788275e-07,
+      "loss": 0.9433,
+      "step": 34469
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4463109505741177e-07,
+      "loss": 0.8092,
+      "step": 34470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4437967312369936e-07,
+      "loss": 0.6293,
+      "step": 34471
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4412832067951994e-07,
+      "loss": 0.8217,
+      "step": 34472
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4387703772669987e-07,
+      "loss": 0.8295,
+      "step": 34473
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.436258242670666e-07,
+      "loss": 0.6261,
+      "step": 34474
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.433746803024475e-07,
+      "loss": 0.9968,
+      "step": 34475
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4312360583466993e-07,
+      "loss": 0.5504,
+      "step": 34476
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4287260086555814e-07,
+      "loss": 1.2529,
+      "step": 34477
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.426216653969384e-07,
+      "loss": 1.0247,
+      "step": 34478
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.42370799430637e-07,
+      "loss": 0.9212,
+      "step": 34479
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.421200029684747e-07,
+      "loss": 0.8831,
+      "step": 34480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.418692760122767e-07,
+      "loss": 0.7616,
+      "step": 34481
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.416186185638671e-07,
+      "loss": 0.9163,
+      "step": 34482
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4136803062506896e-07,
+      "loss": 0.7789,
+      "step": 34483
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.4111751219770294e-07,
+      "loss": 0.9437,
+      "step": 34484
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.408670632835932e-07,
+      "loss": 0.9909,
+      "step": 34485
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.406166838845582e-07,
+      "loss": 0.7263,
+      "step": 34486
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.403663740024211e-07,
+      "loss": 0.7992,
+      "step": 34487
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.401161336390003e-07,
+      "loss": 0.8397,
+      "step": 34488
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.398659627961188e-07,
+      "loss": 0.6611,
+      "step": 34489
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.396158614755919e-07,
+      "loss": 0.5298,
+      "step": 34490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3936582967924134e-07,
+      "loss": 0.987,
+      "step": 34491
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3911586740888246e-07,
+      "loss": 0.8273,
+      "step": 34492
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3886597466633484e-07,
+      "loss": 0.8474,
+      "step": 34493
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3861615145341595e-07,
+      "loss": 0.9916,
+      "step": 34494
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3836639777194325e-07,
+      "loss": 0.7515,
+      "step": 34495
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3811671362373076e-07,
+      "loss": 0.765,
+      "step": 34496
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.37867099010596e-07,
+      "loss": 0.8195,
+      "step": 34497
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3761755393435303e-07,
+      "loss": 0.6527,
+      "step": 34498
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.373680783968182e-07,
+      "loss": 0.7978,
+      "step": 34499
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3711867239980335e-07,
+      "loss": 0.8686,
+      "step": 34500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.368693359451237e-07,
+      "loss": 0.5907,
+      "step": 34501
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.366200690345923e-07,
+      "loss": 0.9472,
+      "step": 34502
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3637087167002214e-07,
+      "loss": 1.0365,
+      "step": 34503
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.36121743853225e-07,
+      "loss": 0.5921,
+      "step": 34504
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.358726855860118e-07,
+      "loss": 1.2337,
+      "step": 34505
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3562369687019437e-07,
+      "loss": 0.9513,
+      "step": 34506
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.353747777075834e-07,
+      "loss": 0.9039,
+      "step": 34507
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3512592809999087e-07,
+      "loss": 0.5589,
+      "step": 34508
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.348771480492242e-07,
+      "loss": 0.8114,
+      "step": 34509
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3462843755709306e-07,
+      "loss": 0.3929,
+      "step": 34510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.343797966254049e-07,
+      "loss": 0.8316,
+      "step": 34511
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.341312252559693e-07,
+      "loss": 1.1963,
+      "step": 34512
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.338827234505949e-07,
+      "loss": 0.9098,
+      "step": 34513
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.336342912110869e-07,
+      "loss": 0.8082,
+      "step": 34514
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3338592853925277e-07,
+      "loss": 0.9644,
+      "step": 34515
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3313763543689877e-07,
+      "loss": 0.8295,
+      "step": 34516
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.328894119058302e-07,
+      "loss": 0.7284,
+      "step": 34517
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3264125794785114e-07,
+      "loss": 0.7825,
+      "step": 34518
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3239317356476906e-07,
+      "loss": 0.6274,
+      "step": 34519
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.321451587583858e-07,
+      "loss": 1.2158,
+      "step": 34520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.318972135305055e-07,
+      "loss": 0.7796,
+      "step": 34521
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3164933788293005e-07,
+      "loss": 0.8939,
+      "step": 34522
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.314015318174647e-07,
+      "loss": 1.0457,
+      "step": 34523
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3115379533590684e-07,
+      "loss": 0.9084,
+      "step": 34524
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3090612844006397e-07,
+      "loss": 0.9842,
+      "step": 34525
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.306585311317335e-07,
+      "loss": 0.9367,
+      "step": 34526
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.3041100341271737e-07,
+      "loss": 0.7082,
+      "step": 34527
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.301635452848152e-07,
+      "loss": 0.6839,
+      "step": 34528
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2991615674982666e-07,
+      "loss": 0.57,
+      "step": 34529
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.296688378095493e-07,
+      "loss": 0.7601,
+      "step": 34530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2942158846578377e-07,
+      "loss": 0.9411,
+      "step": 34531
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2917440872032754e-07,
+      "loss": 0.9294,
+      "step": 34532
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2892729857497704e-07,
+      "loss": 0.7703,
+      "step": 34533
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.286802580315308e-07,
+      "loss": 0.9254,
+      "step": 34534
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.28433287091784e-07,
+      "loss": 1.0146,
+      "step": 34535
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.28186385757533e-07,
+      "loss": 0.7323,
+      "step": 34536
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.279395540305742e-07,
+      "loss": 0.7512,
+      "step": 34537
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.276927919127016e-07,
+      "loss": 0.8122,
+      "step": 34538
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.274460994057117e-07,
+      "loss": 0.9894,
+      "step": 34539
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2719947651139515e-07,
+      "loss": 0.5011,
+      "step": 34540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2695292323154727e-07,
+      "loss": 0.9568,
+      "step": 34541
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2670643956795877e-07,
+      "loss": 0.5907,
+      "step": 34542
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2646002552242605e-07,
+      "loss": 0.6408,
+      "step": 34543
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2621368109673875e-07,
+      "loss": 1.0922,
+      "step": 34544
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2596740629268997e-07,
+      "loss": 0.6323,
+      "step": 34545
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2572120111206817e-07,
+      "loss": 0.8367,
+      "step": 34546
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.254750655566653e-07,
+      "loss": 0.8808,
+      "step": 34547
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2522899962827104e-07,
+      "loss": 0.8354,
+      "step": 34548
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.249830033286739e-07,
+      "loss": 0.8307,
+      "step": 34549
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2473707665966477e-07,
+      "loss": 0.6751,
+      "step": 34550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.244912196230311e-07,
+      "loss": 0.555,
+      "step": 34551
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2424543222056024e-07,
+      "loss": 0.6377,
+      "step": 34552
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2399971445403974e-07,
+      "loss": 1.1384,
+      "step": 34553
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2375406632525483e-07,
+      "loss": 1.0314,
+      "step": 34554
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.235084878359963e-07,
+      "loss": 0.653,
+      "step": 34555
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2326297898804604e-07,
+      "loss": 0.7403,
+      "step": 34556
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.230175397831926e-07,
+      "loss": 0.6941,
+      "step": 34557
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.227721702232179e-07,
+      "loss": 0.8854,
+      "step": 34558
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.225268703099095e-07,
+      "loss": 0.8237,
+      "step": 34559
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.222816400450458e-07,
+      "loss": 0.4071,
+      "step": 34560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.220364794304155e-07,
+      "loss": 0.8386,
+      "step": 34561
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2179138846779933e-07,
+      "loss": 0.7189,
+      "step": 34562
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.215463671589803e-07,
+      "loss": 0.6688,
+      "step": 34563
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2130141550573933e-07,
+      "loss": 0.7545,
+      "step": 34564
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.210565335098582e-07,
+      "loss": 1.0626,
+      "step": 34565
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.208117211731178e-07,
+      "loss": 0.7781,
+      "step": 34566
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2056697849729655e-07,
+      "loss": 0.9881,
+      "step": 34567
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.2032230548417763e-07,
+      "loss": 0.9387,
+      "step": 34568
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.200777021355396e-07,
+      "loss": 0.7408,
+      "step": 34569
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.198331684531598e-07,
+      "loss": 0.6852,
+      "step": 34570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1958870443881693e-07,
+      "loss": 0.8274,
+      "step": 34571
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.193443100942873e-07,
+      "loss": 0.8723,
+      "step": 34572
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1909998542135065e-07,
+      "loss": 0.7752,
+      "step": 34573
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.188557304217833e-07,
+      "loss": 0.7239,
+      "step": 34574
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.186115450973616e-07,
+      "loss": 0.76,
+      "step": 34575
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.183674294498596e-07,
+      "loss": 0.9485,
+      "step": 34576
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.18123383481055e-07,
+      "loss": 0.598,
+      "step": 34577
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.178794071927195e-07,
+      "loss": 0.8129,
+      "step": 34578
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.176355005866306e-07,
+      "loss": 0.7047,
+      "step": 34579
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.173916636645592e-07,
+      "loss": 0.7198,
+      "step": 34580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1714789642828044e-07,
+      "loss": 0.9243,
+      "step": 34581
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1690419887956633e-07,
+      "loss": 0.894,
+      "step": 34582
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1666057102018875e-07,
+      "loss": 0.9158,
+      "step": 34583
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1641701285192073e-07,
+      "loss": 0.8275,
+      "step": 34584
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.161735243765297e-07,
+      "loss": 0.8639,
+      "step": 34585
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1593010559579096e-07,
+      "loss": 1.0334,
+      "step": 34586
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1568675651147303e-07,
+      "loss": 0.8473,
+      "step": 34587
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.154434771253446e-07,
+      "loss": 0.7416,
+      "step": 34588
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.152002674391775e-07,
+      "loss": 0.7901,
+      "step": 34589
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1495712745473595e-07,
+      "loss": 1.0355,
+      "step": 34590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1471405717379177e-07,
+      "loss": 0.6669,
+      "step": 34591
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.144710565981114e-07,
+      "loss": 0.7408,
+      "step": 34592
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.142281257294611e-07,
+      "loss": 0.8756,
+      "step": 34593
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.139852645696085e-07,
+      "loss": 0.6492,
+      "step": 34594
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.13742473120321e-07,
+      "loss": 0.7049,
+      "step": 34595
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1349975138336165e-07,
+      "loss": 0.8281,
+      "step": 34596
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1325709936049563e-07,
+      "loss": 0.6572,
+      "step": 34597
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.130145170534894e-07,
+      "loss": 0.8253,
+      "step": 34598
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.127720044641081e-07,
+      "loss": 0.7586,
+      "step": 34599
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.125295615941116e-07,
+      "loss": 0.7425,
+      "step": 34600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1228718844526506e-07,
+      "loss": 0.6001,
+      "step": 34601
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1204488501933037e-07,
+      "loss": 0.9686,
+      "step": 34602
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.118026513180695e-07,
+      "loss": 0.9477,
+      "step": 34603
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1156048734324547e-07,
+      "loss": 1.0224,
+      "step": 34604
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1131839309661803e-07,
+      "loss": 0.8625,
+      "step": 34605
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.110763685799479e-07,
+      "loss": 0.881,
+      "step": 34606
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.10834413794996e-07,
+      "loss": 0.6502,
+      "step": 34607
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1059252874352085e-07,
+      "loss": 0.7018,
+      "step": 34608
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1035071342728107e-07,
+      "loss": 0.9416,
+      "step": 34609
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.1010896784803633e-07,
+      "loss": 1.0038,
+      "step": 34610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.098672920075442e-07,
+      "loss": 0.9455,
+      "step": 34611
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0962568590756204e-07,
+      "loss": 0.6058,
+      "step": 34612
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.093841495498474e-07,
+      "loss": 0.7977,
+      "step": 34613
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0914268293615554e-07,
+      "loss": 0.7388,
+      "step": 34614
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0890128606824176e-07,
+      "loss": 0.6551,
+      "step": 34615
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.086599589478646e-07,
+      "loss": 0.882,
+      "step": 34616
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.084187015767771e-07,
+      "loss": 0.4585,
+      "step": 34617
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.081775139567334e-07,
+      "loss": 0.6582,
+      "step": 34618
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0793639608948997e-07,
+      "loss": 0.842,
+      "step": 34619
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.076953479767964e-07,
+      "loss": 0.7577,
+      "step": 34620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.07454369620407e-07,
+      "loss": 0.9225,
+      "step": 34621
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.072134610220757e-07,
+      "loss": 0.8557,
+      "step": 34622
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.069726221835524e-07,
+      "loss": 0.8456,
+      "step": 34623
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0673185310659003e-07,
+      "loss": 0.512,
+      "step": 34624
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0649115379293945e-07,
+      "loss": 1.0197,
+      "step": 34625
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.062505242443504e-07,
+      "loss": 0.9356,
+      "step": 34626
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.060099644625726e-07,
+      "loss": 0.6795,
+      "step": 34627
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0576947444935454e-07,
+      "loss": 0.8252,
+      "step": 34628
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0552905420644826e-07,
+      "loss": 0.8399,
+      "step": 34629
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0528870373560016e-07,
+      "loss": 0.6512,
+      "step": 34630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.050484230385565e-07,
+      "loss": 0.7664,
+      "step": 34631
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.048082121170671e-07,
+      "loss": 0.861,
+      "step": 34632
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.045680709728772e-07,
+      "loss": 0.7213,
+      "step": 34633
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.043279996077343e-07,
+      "loss": 0.7488,
+      "step": 34634
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.040879980233836e-07,
+      "loss": 0.9644,
+      "step": 34635
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0384806622157157e-07,
+      "loss": 1.0098,
+      "step": 34636
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0360820420404123e-07,
+      "loss": 0.6891,
+      "step": 34637
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.033684119725378e-07,
+      "loss": 0.9557,
+      "step": 34638
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0312868952880446e-07,
+      "loss": 1.0792,
+      "step": 34639
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0288903687458526e-07,
+      "loss": 1.0045,
+      "step": 34640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0264945401162325e-07,
+      "loss": 0.9293,
+      "step": 34641
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.024099409416604e-07,
+      "loss": 0.7329,
+      "step": 34642
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0217049766643754e-07,
+      "loss": 0.5634,
+      "step": 34643
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.019311241876966e-07,
+      "loss": 0.9889,
+      "step": 34644
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.016918205071785e-07,
+      "loss": 0.8121,
+      "step": 34645
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.014525866266228e-07,
+      "loss": 0.3466,
+      "step": 34646
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.012134225477704e-07,
+      "loss": 1.0464,
+      "step": 34647
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0097432827235995e-07,
+      "loss": 0.8411,
+      "step": 34648
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.007353038021311e-07,
+      "loss": 0.8929,
+      "step": 34649
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.004963491388203e-07,
+      "loss": 0.5183,
+      "step": 34650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.002574642841639e-07,
+      "loss": 0.965,
+      "step": 34651
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.0001864923990385e-07,
+      "loss": 0.8526,
+      "step": 34652
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9977990400777324e-07,
+      "loss": 0.8659,
+      "step": 34653
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.995412285895095e-07,
+      "loss": 0.502,
+      "step": 34654
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.99302622986848e-07,
+      "loss": 0.9262,
+      "step": 34655
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.990640872015239e-07,
+      "loss": 0.8905,
+      "step": 34656
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.988256212352726e-07,
+      "loss": 0.7122,
+      "step": 34657
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9858722508982595e-07,
+      "loss": 0.7038,
+      "step": 34658
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.983488987669215e-07,
+      "loss": 0.6692,
+      "step": 34659
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.98110642268289e-07,
+      "loss": 0.9628,
+      "step": 34660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9787245559566254e-07,
+      "loss": 0.8682,
+      "step": 34661
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9763433875077306e-07,
+      "loss": 0.8855,
+      "step": 34662
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9739629173535355e-07,
+      "loss": 0.619,
+      "step": 34663
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9715831455113264e-07,
+      "loss": 0.7961,
+      "step": 34664
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.969204071998445e-07,
+      "loss": 0.8414,
+      "step": 34665
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.966825696832177e-07,
+      "loss": 0.7175,
+      "step": 34666
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9644480200298206e-07,
+      "loss": 0.6189,
+      "step": 34667
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9620710416086617e-07,
+      "loss": 0.856,
+      "step": 34668
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9596947615859747e-07,
+      "loss": 0.6322,
+      "step": 34669
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.957319179979069e-07,
+      "loss": 0.9476,
+      "step": 34670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9549442968051964e-07,
+      "loss": 0.6447,
+      "step": 34671
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.952570112081633e-07,
+      "loss": 0.8879,
+      "step": 34672
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.950196625825642e-07,
+      "loss": 1.0089,
+      "step": 34673
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9478238380544877e-07,
+      "loss": 0.5721,
+      "step": 34674
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.945451748785434e-07,
+      "loss": 0.7979,
+      "step": 34675
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9430803580357e-07,
+      "loss": 0.7036,
+      "step": 34676
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.940709665822573e-07,
+      "loss": 0.5557,
+      "step": 34677
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9383396721632715e-07,
+      "loss": 0.6781,
+      "step": 34678
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.935970377075038e-07,
+      "loss": 0.7025,
+      "step": 34679
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.93360178057508e-07,
+      "loss": 0.7248,
+      "step": 34680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.931233882680652e-07,
+      "loss": 1.0199,
+      "step": 34681
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9288666834089387e-07,
+      "loss": 0.9569,
+      "step": 34682
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.926500182777182e-07,
+      "loss": 0.9967,
+      "step": 34683
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.924134380802591e-07,
+      "loss": 0.6,
+      "step": 34684
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.921769277502363e-07,
+      "loss": 0.8962,
+      "step": 34685
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.919404872893695e-07,
+      "loss": 0.9841,
+      "step": 34686
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.917041166993796e-07,
+      "loss": 0.9931,
+      "step": 34687
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9146781598198294e-07,
+      "loss": 0.8629,
+      "step": 34688
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9123158513890035e-07,
+      "loss": 0.8378,
+      "step": 34689
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9099542417184723e-07,
+      "loss": 0.8576,
+      "step": 34690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.907593330825443e-07,
+      "loss": 0.698,
+      "step": 34691
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9052331187270474e-07,
+      "loss": 0.6997,
+      "step": 34692
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.9028736054404824e-07,
+      "loss": 0.8662,
+      "step": 34693
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.900514790982879e-07,
+      "loss": 0.7858,
+      "step": 34694
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8981566753714116e-07,
+      "loss": 0.8881,
+      "step": 34695
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8957992586232117e-07,
+      "loss": 0.3434,
+      "step": 34696
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.893442540755443e-07,
+      "loss": 0.6247,
+      "step": 34697
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.891086521785237e-07,
+      "loss": 0.7486,
+      "step": 34698
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.888731201729723e-07,
+      "loss": 0.6383,
+      "step": 34699
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8863765806060107e-07,
+      "loss": 0.5566,
+      "step": 34700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8840226584312523e-07,
+      "loss": 0.7759,
+      "step": 34701
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.881669435222546e-07,
+      "loss": 0.8168,
+      "step": 34702
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.879316910997022e-07,
+      "loss": 0.5731,
+      "step": 34703
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.876965085771778e-07,
+      "loss": 0.8116,
+      "step": 34704
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.874613959563922e-07,
+      "loss": 0.6085,
+      "step": 34705
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.872263532390541e-07,
+      "loss": 0.8333,
+      "step": 34706
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8699138042687324e-07,
+      "loss": 0.999,
+      "step": 34707
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.867564775215582e-07,
+      "loss": 0.6653,
+      "step": 34708
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.865216445248199e-07,
+      "loss": 0.8725,
+      "step": 34709
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8628688143836244e-07,
+      "loss": 0.6818,
+      "step": 34710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.860521882638946e-07,
+      "loss": 0.9526,
+      "step": 34711
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.858175650031204e-07,
+      "loss": 0.9306,
+      "step": 34712
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.855830116577508e-07,
+      "loss": 0.7699,
+      "step": 34713
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8534852822948886e-07,
+      "loss": 0.9102,
+      "step": 34714
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8511411472004104e-07,
+      "loss": 0.7511,
+      "step": 34715
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.848797711311103e-07,
+      "loss": 0.8256,
+      "step": 34716
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.846454974644009e-07,
+      "loss": 0.7307,
+      "step": 34717
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8441129372161825e-07,
+      "loss": 0.9234,
+      "step": 34718
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.841771599044641e-07,
+      "loss": 0.9049,
+      "step": 34719
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8394309601464064e-07,
+      "loss": 0.7818,
+      "step": 34720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8370910205385193e-07,
+      "loss": 0.8606,
+      "step": 34721
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8347517802379773e-07,
+      "loss": 0.8268,
+      "step": 34722
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.832413239261812e-07,
+      "loss": 0.9901,
+      "step": 34723
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8300753976270087e-07,
+      "loss": 0.8661,
+      "step": 34724
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.827738255350555e-07,
+      "loss": 0.8028,
+      "step": 34725
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.825401812449492e-07,
+      "loss": 0.6046,
+      "step": 34726
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.823066068940784e-07,
+      "loss": 1.1316,
+      "step": 34727
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8207310248414287e-07,
+      "loss": 0.7575,
+      "step": 34728
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.818396680168401e-07,
+      "loss": 0.8945,
+      "step": 34729
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.816063034938644e-07,
+      "loss": 0.6776,
+      "step": 34730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8137300891691765e-07,
+      "loss": 1.0444,
+      "step": 34731
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8113978428769405e-07,
+      "loss": 0.8019,
+      "step": 34732
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8090662960789113e-07,
+      "loss": 0.7707,
+      "step": 34733
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8067354487920207e-07,
+      "loss": 0.5588,
+      "step": 34734
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.8044053010332426e-07,
+      "loss": 0.7616,
+      "step": 34735
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.802075852819509e-07,
+      "loss": 1.0449,
+      "step": 34736
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.799747104167739e-07,
+      "loss": 0.8229,
+      "step": 34737
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7974190550949197e-07,
+      "loss": 0.9616,
+      "step": 34738
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.795091705617948e-07,
+      "loss": 1.0268,
+      "step": 34739
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.792765055753755e-07,
+      "loss": 0.5762,
+      "step": 34740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.79043910551925e-07,
+      "loss": 0.6577,
+      "step": 34741
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.788113854931363e-07,
+      "loss": 0.7342,
+      "step": 34742
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7857893040069706e-07,
+      "loss": 1.1052,
+      "step": 34743
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.783465452763024e-07,
+      "loss": 0.8125,
+      "step": 34744
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7811423012164005e-07,
+      "loss": 0.6108,
+      "step": 34745
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.778819849383997e-07,
+      "loss": 0.7882,
+      "step": 34746
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.776498097282699e-07,
+      "loss": 0.6863,
+      "step": 34747
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.774177044929395e-07,
+      "loss": 0.7082,
+      "step": 34748
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7718566923409584e-07,
+      "loss": 0.8541,
+      "step": 34749
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.7695370395342546e-07,
+      "loss": 0.6924,
+      "step": 34750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.767218086526181e-07,
+      "loss": 0.7857,
+      "step": 34751
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7648998333335797e-07,
+      "loss": 0.8997,
+      "step": 34752
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7625822799733035e-07,
+      "loss": 0.9931,
+      "step": 34753
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7602654264622284e-07,
+      "loss": 0.7917,
+      "step": 34754
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.757949272817174e-07,
+      "loss": 0.6971,
+      "step": 34755
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7556338190550045e-07,
+      "loss": 1.032,
+      "step": 34756
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7533190651925623e-07,
+      "loss": 0.8323,
+      "step": 34757
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7510050112466666e-07,
+      "loss": 0.8149,
+      "step": 34758
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7486916572341713e-07,
+      "loss": 0.6665,
+      "step": 34759
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.746379003171863e-07,
+      "loss": 0.666,
+      "step": 34760
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.744067049076561e-07,
+      "loss": 0.7659,
+      "step": 34761
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7417557949650963e-07,
+      "loss": 1.0263,
+      "step": 34762
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.739445240854289e-07,
+      "loss": 0.9868,
+      "step": 34763
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.737135386760915e-07,
+      "loss": 0.8425,
+      "step": 34764
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.734826232701783e-07,
+      "loss": 0.7356,
+      "step": 34765
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7325177786936897e-07,
+      "loss": 0.8982,
+      "step": 34766
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.730210024753411e-07,
+      "loss": 0.5877,
+      "step": 34767
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.727902970897723e-07,
+      "loss": 0.6264,
+      "step": 34768
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7255966171434453e-07,
+      "loss": 1.0306,
+      "step": 34769
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.723290963507309e-07,
+      "loss": 0.8631,
+      "step": 34770
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.720986010006089e-07,
+      "loss": 0.6881,
+      "step": 34771
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7186817566565503e-07,
+      "loss": 0.7116,
+      "step": 34772
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.716378203475435e-07,
+      "loss": 1.0481,
+      "step": 34773
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7140753504795177e-07,
+      "loss": 0.94,
+      "step": 34774
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.711773197685553e-07,
+      "loss": 0.8044,
+      "step": 34775
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.709471745110249e-07,
+      "loss": 0.7366,
+      "step": 34776
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7071709927703704e-07,
+      "loss": 0.8072,
+      "step": 34777
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.704870940682637e-07,
+      "loss": 1.0853,
+      "step": 34778
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.7025715888637683e-07,
+      "loss": 1.0378,
+      "step": 34779
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.700272937330496e-07,
+      "loss": 1.2267,
+      "step": 34780
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6979749860995284e-07,
+      "loss": 0.7469,
+      "step": 34781
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.695677735187586e-07,
+      "loss": 0.7653,
+      "step": 34782
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6933811846113665e-07,
+      "loss": 1.0238,
+      "step": 34783
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6910853343875785e-07,
+      "loss": 1.0043,
+      "step": 34784
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.688790184532909e-07,
+      "loss": 0.8992,
+      "step": 34785
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.686495735064044e-07,
+      "loss": 0.942,
+      "step": 34786
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.684201985997682e-07,
+      "loss": 0.9137,
+      "step": 34787
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.681908937350509e-07,
+      "loss": 0.695,
+      "step": 34788
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.67961658913919e-07,
+      "loss": 0.7482,
+      "step": 34789
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.677324941380389e-07,
+      "loss": 0.7628,
+      "step": 34790
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6750339940907707e-07,
+      "loss": 0.6452,
+      "step": 34791
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.672743747287011e-07,
+      "loss": 0.8081,
+      "step": 34792
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6704542009857516e-07,
+      "loss": 0.8293,
+      "step": 34793
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6681653552036456e-07,
+      "loss": 0.6592,
+      "step": 34794
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.665877209957347e-07,
+      "loss": 0.8403,
+      "step": 34795
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6635897652634754e-07,
+      "loss": 0.6401,
+      "step": 34796
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.661303021138685e-07,
+      "loss": 0.8958,
+      "step": 34797
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.659016977599583e-07,
+      "loss": 0.8807,
+      "step": 34798
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6567316346628354e-07,
+      "loss": 0.7443,
+      "step": 34799
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.654446992345018e-07,
+      "loss": 0.7451,
+      "step": 34800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.65216305066276e-07,
+      "loss": 0.5348,
+      "step": 34801
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.649879809632673e-07,
+      "loss": 0.7362,
+      "step": 34802
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6475972692713633e-07,
+      "loss": 0.7471,
+      "step": 34803
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6453154295954087e-07,
+      "loss": 1.0646,
+      "step": 34804
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.643034290621439e-07,
+      "loss": 0.8447,
+      "step": 34805
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6407538523660193e-07,
+      "loss": 1.1131,
+      "step": 34806
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6384741148457357e-07,
+      "loss": 0.8282,
+      "step": 34807
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.636195078077176e-07,
+      "loss": 0.7501,
+      "step": 34808
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6339167420768926e-07,
+      "loss": 0.5057,
+      "step": 34809
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6316391068614844e-07,
+      "loss": 0.7202,
+      "step": 34810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6293621724474925e-07,
+      "loss": 0.9108,
+      "step": 34811
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6270859388514713e-07,
+      "loss": 0.6946,
+      "step": 34812
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.624810406089996e-07,
+      "loss": 0.8197,
+      "step": 34813
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6225355741795977e-07,
+      "loss": 0.5549,
+      "step": 34814
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.620261443136819e-07,
+      "loss": 0.7499,
+      "step": 34815
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.617988012978191e-07,
+      "loss": 0.8778,
+      "step": 34816
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.615715283720278e-07,
+      "loss": 0.8665,
+      "step": 34817
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.613443255379578e-07,
+      "loss": 0.7375,
+      "step": 34818
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6111719279726343e-07,
+      "loss": 0.7696,
+      "step": 34819
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.608901301515944e-07,
+      "loss": 0.7671,
+      "step": 34820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.606631376026015e-07,
+      "loss": 0.7384,
+      "step": 34821
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6043621515193696e-07,
+      "loss": 0.868,
+      "step": 34822
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6020936280125154e-07,
+      "loss": 0.8861,
+      "step": 34823
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.599825805521939e-07,
+      "loss": 0.5255,
+      "step": 34824
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5975586840641396e-07,
+      "loss": 0.8688,
+      "step": 34825
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.595292263655592e-07,
+      "loss": 0.6124,
+      "step": 34826
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.593026544312783e-07,
+      "loss": 0.8908,
+      "step": 34827
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5907615260522e-07,
+      "loss": 0.9718,
+      "step": 34828
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.588497208890296e-07,
+      "loss": 0.9533,
+      "step": 34829
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5862335928435465e-07,
+      "loss": 0.5552,
+      "step": 34830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5839706779284167e-07,
+      "loss": 0.7728,
+      "step": 34831
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.58170846416136e-07,
+      "loss": 0.7588,
+      "step": 34832
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5794469515588294e-07,
+      "loss": 0.8995,
+      "step": 34833
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5771861401372453e-07,
+      "loss": 0.9843,
+      "step": 34834
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.574926029913095e-07,
+      "loss": 1.0273,
+      "step": 34835
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.572666620902787e-07,
+      "loss": 0.8435,
+      "step": 34836
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.570407913122753e-07,
+      "loss": 0.6266,
+      "step": 34837
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.568149906589413e-07,
+      "loss": 0.8106,
+      "step": 34838
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5658926013192205e-07,
+      "loss": 0.5284,
+      "step": 34839
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5636359973285405e-07,
+      "loss": 0.7798,
+      "step": 34840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.561380094633815e-07,
+      "loss": 0.6788,
+      "step": 34841
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5591248932514423e-07,
+      "loss": 0.7913,
+      "step": 34842
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5568703931978197e-07,
+      "loss": 0.7134,
+      "step": 34843
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5546165944893574e-07,
+      "loss": 0.7167,
+      "step": 34844
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5523634971424194e-07,
+      "loss": 1.0074,
+      "step": 34845
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5501111011734144e-07,
+      "loss": 0.6685,
+      "step": 34846
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.547859406598697e-07,
+      "loss": 0.7808,
+      "step": 34847
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5456084134346646e-07,
+      "loss": 0.6557,
+      "step": 34848
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5433581216977043e-07,
+      "loss": 0.9257,
+      "step": 34849
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.541108531404125e-07,
+      "loss": 0.9899,
+      "step": 34850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.538859642570325e-07,
+      "loss": 1.067,
+      "step": 34851
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.536611455212635e-07,
+      "loss": 0.6561,
+      "step": 34852
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.534363969347432e-07,
+      "loss": 0.6,
+      "step": 34853
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5321171849910576e-07,
+      "loss": 0.8617,
+      "step": 34854
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.529871102159821e-07,
+      "loss": 0.8822,
+      "step": 34855
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.527625720870087e-07,
+      "loss": 0.8086,
+      "step": 34856
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.525381041138165e-07,
+      "loss": 0.5294,
+      "step": 34857
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.523137062980386e-07,
+      "loss": 0.7156,
+      "step": 34858
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.520893786413071e-07,
+      "loss": 0.8543,
+      "step": 34859
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5186512114525283e-07,
+      "loss": 1.0923,
+      "step": 34860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.516409338115068e-07,
+      "loss": 0.7344,
+      "step": 34861
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.514168166416998e-07,
+      "loss": 0.7503,
+      "step": 34862
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5119276963746063e-07,
+      "loss": 0.4682,
+      "step": 34863
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5096879280041907e-07,
+      "loss": 0.711,
+      "step": 34864
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.507448861322027e-07,
+      "loss": 0.732,
+      "step": 34865
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.5052104963444243e-07,
+      "loss": 0.9646,
+      "step": 34866
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.502972833087648e-07,
+      "loss": 0.8154,
+      "step": 34867
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.500735871567962e-07,
+      "loss": 0.8406,
+      "step": 34868
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4984996118016533e-07,
+      "loss": 0.6266,
+      "step": 34869
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4962640538049543e-07,
+      "loss": 0.5913,
+      "step": 34870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.494029197594151e-07,
+      "loss": 0.7175,
+      "step": 34871
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.491795043185486e-07,
+      "loss": 1.0435,
+      "step": 34872
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.489561590595203e-07,
+      "loss": 0.953,
+      "step": 34873
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4873288398395434e-07,
+      "loss": 0.8505,
+      "step": 34874
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.485096790934739e-07,
+      "loss": 0.6946,
+      "step": 34875
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.482865443897032e-07,
+      "loss": 0.7897,
+      "step": 34876
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4806347987426434e-07,
+      "loss": 1.0047,
+      "step": 34877
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4784048554878045e-07,
+      "loss": 0.6889,
+      "step": 34878
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.476175614148725e-07,
+      "loss": 0.6268,
+      "step": 34879
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.473947074741613e-07,
+      "loss": 0.7004,
+      "step": 34880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.471719237282678e-07,
+      "loss": 0.7723,
+      "step": 34881
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.469492101788119e-07,
+      "loss": 0.6341,
+      "step": 34882
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4672656682741224e-07,
+      "loss": 0.7138,
+      "step": 34883
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4650399367568974e-07,
+      "loss": 0.7603,
+      "step": 34884
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4628149072526207e-07,
+      "loss": 0.6427,
+      "step": 34885
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4605905797774785e-07,
+      "loss": 1.0243,
+      "step": 34886
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.458366954347636e-07,
+      "loss": 0.9495,
+      "step": 34887
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4561440309792804e-07,
+      "loss": 0.7286,
+      "step": 34888
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.453921809688554e-07,
+      "loss": 0.7306,
+      "step": 34889
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.451700290491633e-07,
+      "loss": 0.9231,
+      "step": 34890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.449479473404671e-07,
+      "loss": 0.8339,
+      "step": 34891
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4472593584438217e-07,
+      "loss": 0.5705,
+      "step": 34892
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.445039945625217e-07,
+      "loss": 0.8009,
+      "step": 34893
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4428212349650105e-07,
+      "loss": 0.5029,
+      "step": 34894
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.440603226479311e-07,
+      "loss": 1.1034,
+      "step": 34895
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.438385920184284e-07,
+      "loss": 0.9756,
+      "step": 34896
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4361693160960277e-07,
+      "loss": 0.655,
+      "step": 34897
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.433953414230684e-07,
+      "loss": 0.6028,
+      "step": 34898
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.431738214604363e-07,
+      "loss": 0.9521,
+      "step": 34899
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.429523717233152e-07,
+      "loss": 0.601,
+      "step": 34900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4273099221331487e-07,
+      "loss": 0.5279,
+      "step": 34901
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4250968293204956e-07,
+      "loss": 1.1115,
+      "step": 34902
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.422884438811258e-07,
+      "loss": 0.8351,
+      "step": 34903
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4206727506215343e-07,
+      "loss": 0.96,
+      "step": 34904
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4184617647674e-07,
+      "loss": 0.686,
+      "step": 34905
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4162514812649315e-07,
+      "loss": 0.9475,
+      "step": 34906
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4140419001302163e-07,
+      "loss": 0.4312,
+      "step": 34907
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4118330213792963e-07,
+      "loss": 0.9909,
+      "step": 34908
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.40962484502827e-07,
+      "loss": 0.7094,
+      "step": 34909
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4074173710931804e-07,
+      "loss": 0.9969,
+      "step": 34910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4052105995900696e-07,
+      "loss": 0.872,
+      "step": 34911
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4030045305349923e-07,
+      "loss": 0.8986,
+      "step": 34912
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.4007991639440016e-07,
+      "loss": 0.7014,
+      "step": 34913
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.398594499833108e-07,
+      "loss": 1.0182,
+      "step": 34914
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3963905382183747e-07,
+      "loss": 0.6849,
+      "step": 34915
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3941872791158124e-07,
+      "loss": 0.893,
+      "step": 34916
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.391984722541464e-07,
+      "loss": 0.6619,
+      "step": 34917
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3897828685113154e-07,
+      "loss": 0.5477,
+      "step": 34918
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3875817170413995e-07,
+      "loss": 0.7085,
+      "step": 34919
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3853812681477136e-07,
+      "loss": 0.5439,
+      "step": 34920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.383181521846268e-07,
+      "loss": 0.599,
+      "step": 34921
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.380982478153061e-07,
+      "loss": 0.5603,
+      "step": 34922
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.378784137084068e-07,
+      "loss": 0.7767,
+      "step": 34923
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3765864986552986e-07,
+      "loss": 0.8012,
+      "step": 34924
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3743895628827073e-07,
+      "loss": 0.9479,
+      "step": 34925
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.372193329782281e-07,
+      "loss": 0.7236,
+      "step": 34926
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.369997799370006e-07,
+      "loss": 0.8141,
+      "step": 34927
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.367802971661838e-07,
+      "loss": 0.5738,
+      "step": 34928
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.365608846673751e-07,
+      "loss": 1.2125,
+      "step": 34929
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3634154244216786e-07,
+      "loss": 0.9332,
+      "step": 34930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.361222704921574e-07,
+      "loss": 1.0721,
+      "step": 34931
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3590306881893797e-07,
+      "loss": 0.6379,
+      "step": 34932
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.35683937424105e-07,
+      "loss": 1.0065,
+      "step": 34933
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3546487630925275e-07,
+      "loss": 0.8616,
+      "step": 34934
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3524588547597327e-07,
+      "loss": 0.8086,
+      "step": 34935
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.350269649258575e-07,
+      "loss": 0.5028,
+      "step": 34936
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.348081146605009e-07,
+      "loss": 0.6989,
+      "step": 34937
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.34589334681491e-07,
+      "loss": 0.5843,
+      "step": 34938
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3437062499042216e-07,
+      "loss": 1.0455,
+      "step": 34939
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3415198558888305e-07,
+      "loss": 0.8448,
+      "step": 34940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3393341647846353e-07,
+      "loss": 0.9681,
+      "step": 34941
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.337149176607546e-07,
+      "loss": 0.7509,
+      "step": 34942
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.334964891373438e-07,
+      "loss": 0.7267,
+      "step": 34943
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3327813090981876e-07,
+      "loss": 0.8841,
+      "step": 34944
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3305984297976936e-07,
+      "loss": 1.0898,
+      "step": 34945
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3284162534878315e-07,
+      "loss": 0.9779,
+      "step": 34946
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.326234780184456e-07,
+      "loss": 0.6113,
+      "step": 34947
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3240540099034323e-07,
+      "loss": 1.0061,
+      "step": 34948
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3218739426606473e-07,
+      "loss": 0.9916,
+      "step": 34949
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3196945784718993e-07,
+      "loss": 0.8363,
+      "step": 34950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.317515917353076e-07,
+      "loss": 0.8799,
+      "step": 34951
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3153379593200195e-07,
+      "loss": 0.6894,
+      "step": 34952
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3131607043885626e-07,
+      "loss": 0.6527,
+      "step": 34953
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3109841525745367e-07,
+      "loss": 0.8369,
+      "step": 34954
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3088083038937734e-07,
+      "loss": 0.701,
+      "step": 34955
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3066331583620827e-07,
+      "loss": 0.8024,
+      "step": 34956
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3044587159952955e-07,
+      "loss": 0.8605,
+      "step": 34957
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3022849768092223e-07,
+      "loss": 0.8168,
+      "step": 34958
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.3001119408196946e-07,
+      "loss": 0.8123,
+      "step": 34959
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.297939608042466e-07,
+      "loss": 0.9769,
+      "step": 34960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.295767978493369e-07,
+      "loss": 0.8283,
+      "step": 34961
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.293597052188169e-07,
+      "loss": 0.6615,
+      "step": 34962
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2914268291426855e-07,
+      "loss": 0.7069,
+      "step": 34963
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.289257309372684e-07,
+      "loss": 0.8823,
+      "step": 34964
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.287088492893942e-07,
+      "loss": 0.4835,
+      "step": 34965
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.284920379722234e-07,
+      "loss": 0.7907,
+      "step": 34966
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2827529698733263e-07,
+      "loss": 0.9897,
+      "step": 34967
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2805862633629726e-07,
+      "loss": 0.7015,
+      "step": 34968
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.278420260206949e-07,
+      "loss": 0.5862,
+      "step": 34969
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.276254960420977e-07,
+      "loss": 0.8972,
+      "step": 34970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.274090364020832e-07,
+      "loss": 1.1286,
+      "step": 34971
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2719264710222356e-07,
+      "loss": 1.0964,
+      "step": 34972
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2697632814409407e-07,
+      "loss": 0.6546,
+      "step": 34973
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.267600795292658e-07,
+      "loss": 0.9159,
+      "step": 34974
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.265439012593119e-07,
+      "loss": 0.9325,
+      "step": 34975
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2632779333580555e-07,
+      "loss": 0.7932,
+      "step": 34976
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2611175576031886e-07,
+      "loss": 0.9933,
+      "step": 34977
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.258957885344205e-07,
+      "loss": 0.6559,
+      "step": 34978
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2567989165968373e-07,
+      "loss": 0.6166,
+      "step": 34979
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.25464065137675e-07,
+      "loss": 0.7951,
+      "step": 34980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2524830896996764e-07,
+      "loss": 0.6026,
+      "step": 34981
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.25032623158128e-07,
+      "loss": 0.7882,
+      "step": 34982
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2481700770372605e-07,
+      "loss": 0.9433,
+      "step": 34983
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2460146260832937e-07,
+      "loss": 0.8804,
+      "step": 34984
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.243859878735045e-07,
+      "loss": 1.0104,
+      "step": 34985
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.241705835008202e-07,
+      "loss": 0.7171,
+      "step": 34986
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2395524949183965e-07,
+      "loss": 0.6886,
+      "step": 34987
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.237399858481327e-07,
+      "loss": 0.9332,
+      "step": 34988
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.235247925712648e-07,
+      "loss": 0.6994,
+      "step": 34989
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2330966966279684e-07,
+      "loss": 0.6512,
+      "step": 34990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2309461712429656e-07,
+      "loss": 0.6969,
+      "step": 34991
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.22879634957326e-07,
+      "loss": 0.9596,
+      "step": 34992
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2266472316344945e-07,
+      "loss": 0.9096,
+      "step": 34993
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2244988174422895e-07,
+      "loss": 0.9908,
+      "step": 34994
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2223511070122893e-07,
+      "loss": 1.1846,
+      "step": 34995
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2202041003601026e-07,
+      "loss": 0.6963,
+      "step": 34996
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2180577975013284e-07,
+      "loss": 0.8546,
+      "step": 34997
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.215912198451587e-07,
+      "loss": 1.0344,
+      "step": 34998
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.213767303226478e-07,
+      "loss": 0.7501,
+      "step": 34999
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.211623111841611e-07,
+      "loss": 0.86,
+      "step": 35000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.209479624312561e-07,
+      "loss": 0.5906,
+      "step": 35001
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.2073368406549175e-07,
+      "loss": 0.7558,
+      "step": 35002
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.205194760884278e-07,
+      "loss": 0.9001,
+      "step": 35003
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.203053385016208e-07,
+      "loss": 0.6942,
+      "step": 35004
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.200912713066273e-07,
+      "loss": 0.5722,
+      "step": 35005
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1987727450500606e-07,
+      "loss": 1.0816,
+      "step": 35006
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1966334809831243e-07,
+      "loss": 1.1225,
+      "step": 35007
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1944949208810194e-07,
+      "loss": 0.6419,
+      "step": 35008
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1923570647593106e-07,
+      "loss": 1.0129,
+      "step": 35009
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.190219912633519e-07,
+      "loss": 0.5197,
+      "step": 35010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.188083464519176e-07,
+      "loss": 0.8204,
+      "step": 35011
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1859477204318703e-07,
+      "loss": 0.8597,
+      "step": 35012
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1838126803870884e-07,
+      "loss": 0.8063,
+      "step": 35013
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.181678344400374e-07,
+      "loss": 0.9949,
+      "step": 35014
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.179544712487259e-07,
+      "loss": 0.6162,
+      "step": 35015
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.177411784663231e-07,
+      "loss": 0.7531,
+      "step": 35016
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1752795609438224e-07,
+      "loss": 0.7488,
+      "step": 35017
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.17314804134452e-07,
+      "loss": 0.8437,
+      "step": 35018
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.171017225880868e-07,
+      "loss": 0.6605,
+      "step": 35019
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1688871145683087e-07,
+      "loss": 0.7799,
+      "step": 35020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1667577074223635e-07,
+      "loss": 0.55,
+      "step": 35021
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1646290044585083e-07,
+      "loss": 0.8846,
+      "step": 35022
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.162501005692209e-07,
+      "loss": 0.7767,
+      "step": 35023
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1603737111389756e-07,
+      "loss": 0.9764,
+      "step": 35024
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.158247120814251e-07,
+      "loss": 0.6288,
+      "step": 35025
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1561212347335e-07,
+      "loss": 0.8307,
+      "step": 35026
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1539960529122007e-07,
+      "loss": 0.4855,
+      "step": 35027
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1518715753657945e-07,
+      "loss": 0.6412,
+      "step": 35028
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.149747802109737e-07,
+      "loss": 0.8801,
+      "step": 35029
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.14762473315946e-07,
+      "loss": 0.7222,
+      "step": 35030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.145502368530418e-07,
+      "loss": 0.8588,
+      "step": 35031
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1433807082380305e-07,
+      "loss": 1.0379,
+      "step": 35032
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.141259752297743e-07,
+      "loss": 0.909,
+      "step": 35033
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1391395007249635e-07,
+      "loss": 0.7819,
+      "step": 35034
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1370199535351253e-07,
+      "loss": 0.7977,
+      "step": 35035
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1349011107436157e-07,
+      "loss": 0.9273,
+      "step": 35036
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1327829723658776e-07,
+      "loss": 0.6631,
+      "step": 35037
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.130665538417299e-07,
+      "loss": 0.6935,
+      "step": 35038
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1285488089132787e-07,
+      "loss": 0.9095,
+      "step": 35039
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1264327838692153e-07,
+      "loss": 0.9411,
+      "step": 35040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1243174633004637e-07,
+      "loss": 0.8475,
+      "step": 35041
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.122202847222455e-07,
+      "loss": 0.9078,
+      "step": 35042
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1200889356505336e-07,
+      "loss": 0.819,
+      "step": 35043
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1179757286000976e-07,
+      "loss": 0.8827,
+      "step": 35044
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.115863226086491e-07,
+      "loss": 0.7818,
+      "step": 35045
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.113751428125089e-07,
+      "loss": 0.5442,
+      "step": 35046
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.111640334731247e-07,
+      "loss": 0.652,
+      "step": 35047
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.109529945920298e-07,
+      "loss": 0.9202,
+      "step": 35048
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.107420261707628e-07,
+      "loss": 0.8006,
+      "step": 35049
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.1053112821085586e-07,
+      "loss": 0.77,
+      "step": 35050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.103203007138411e-07,
+      "loss": 0.6042,
+      "step": 35051
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.10109543681254e-07,
+      "loss": 0.9147,
+      "step": 35052
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0989885711462666e-07,
+      "loss": 0.6473,
+      "step": 35053
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.096882410154889e-07,
+      "loss": 0.6825,
+      "step": 35054
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.094776953853762e-07,
+      "loss": 0.8095,
+      "step": 35055
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0926722022581733e-07,
+      "loss": 0.8511,
+      "step": 35056
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.090568155383433e-07,
+      "loss": 0.79,
+      "step": 35057
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0884648132448515e-07,
+      "loss": 0.6371,
+      "step": 35058
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0863621758577043e-07,
+      "loss": 0.5911,
+      "step": 35059
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0842602432373027e-07,
+      "loss": 0.6717,
+      "step": 35060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0821590153989223e-07,
+      "loss": 0.7092,
+      "step": 35061
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.080058492357829e-07,
+      "loss": 0.8581,
+      "step": 35062
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.077958674129333e-07,
+      "loss": 0.9998,
+      "step": 35063
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0758595607286666e-07,
+      "loss": 0.7311,
+      "step": 35064
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0737611521711284e-07,
+      "loss": 0.5869,
+      "step": 35065
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.07166344847194e-07,
+      "loss": 0.7234,
+      "step": 35066
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0695664496463997e-07,
+      "loss": 0.7928,
+      "step": 35067
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.067470155709729e-07,
+      "loss": 0.9049,
+      "step": 35068
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.065374566677193e-07,
+      "loss": 0.7329,
+      "step": 35069
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.063279682564002e-07,
+      "loss": 1.047,
+      "step": 35070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.061185503385411e-07,
+      "loss": 0.7059,
+      "step": 35071
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.059092029156629e-07,
+      "loss": 0.9468,
+      "step": 35072
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.056999259892912e-07,
+      "loss": 0.7768,
+      "step": 35073
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.054907195609458e-07,
+      "loss": 0.809,
+      "step": 35074
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0528158363214764e-07,
+      "loss": 1.0069,
+      "step": 35075
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0507251820442007e-07,
+      "loss": 0.8613,
+      "step": 35076
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.048635232792807e-07,
+      "loss": 0.6878,
+      "step": 35077
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0465459885825054e-07,
+      "loss": 0.6585,
+      "step": 35078
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.044457449428484e-07,
+      "loss": 0.7812,
+      "step": 35079
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0423696153459417e-07,
+      "loss": 0.6641,
+      "step": 35080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0402824863500435e-07,
+      "loss": 0.7105,
+      "step": 35081
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0381960624559893e-07,
+      "loss": 0.9341,
+      "step": 35082
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.036110343678933e-07,
+      "loss": 0.6016,
+      "step": 35083
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.034025330034052e-07,
+      "loss": 0.9498,
+      "step": 35084
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0319410215365e-07,
+      "loss": 0.9295,
+      "step": 35085
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.029857418201454e-07,
+      "loss": 0.6517,
+      "step": 35086
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0277745200440466e-07,
+      "loss": 0.5453,
+      "step": 35087
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0256923270794326e-07,
+      "loss": 0.9185,
+      "step": 35088
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0236108393227657e-07,
+      "loss": 0.4486,
+      "step": 35089
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0215300567891457e-07,
+      "loss": 1.0193,
+      "step": 35090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0194499794937383e-07,
+      "loss": 0.9479,
+      "step": 35091
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.017370607451664e-07,
+      "loss": 0.5903,
+      "step": 35092
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0152919406780336e-07,
+      "loss": 1.066,
+      "step": 35093
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.013213979187979e-07,
+      "loss": 0.8044,
+      "step": 35094
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0111367229965993e-07,
+      "loss": 0.6491,
+      "step": 35095
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.009060172119005e-07,
+      "loss": 0.6749,
+      "step": 35096
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.006984326570284e-07,
+      "loss": 0.9293,
+      "step": 35097
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0049091863655453e-07,
+      "loss": 0.737,
+      "step": 35098
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.002834751519901e-07,
+      "loss": 0.802,
+      "step": 35099
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.0007610220483927e-07,
+      "loss": 0.6861,
+      "step": 35100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.998687997966132e-07,
+      "loss": 0.9205,
+      "step": 35101
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.996615679288173e-07,
+      "loss": 1.0628,
+      "step": 35102
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9945440660296034e-07,
+      "loss": 0.7563,
+      "step": 35103
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.992473158205478e-07,
+      "loss": 0.6029,
+      "step": 35104
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.990402955830862e-07,
+      "loss": 0.9424,
+      "step": 35105
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.988333458920811e-07,
+      "loss": 0.7253,
+      "step": 35106
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9862646674903686e-07,
+      "loss": 0.9424,
+      "step": 35107
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.984196581554588e-07,
+      "loss": 0.7515,
+      "step": 35108
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.982129201128503e-07,
+      "loss": 0.7039,
+      "step": 35109
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9800625262271455e-07,
+      "loss": 0.9526,
+      "step": 35110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9779965568655477e-07,
+      "loss": 0.8153,
+      "step": 35111
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9759312930587315e-07,
+      "loss": 0.4673,
+      "step": 35112
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9738667348217175e-07,
+      "loss": 0.9115,
+      "step": 35113
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9718028821695166e-07,
+      "loss": 0.6489,
+      "step": 35114
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.969739735117128e-07,
+      "loss": 0.694,
+      "step": 35115
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9676772936795826e-07,
+      "loss": 0.8975,
+      "step": 35116
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9656155578718703e-07,
+      "loss": 0.9031,
+      "step": 35117
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.963554527708967e-07,
+      "loss": 0.8403,
+      "step": 35118
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9614942032058834e-07,
+      "loss": 0.8125,
+      "step": 35119
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.959434584377563e-07,
+      "loss": 0.6314,
+      "step": 35120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9573756712390264e-07,
+      "loss": 0.8905,
+      "step": 35121
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9553174638052297e-07,
+      "loss": 0.8712,
+      "step": 35122
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.953259962091148e-07,
+      "loss": 0.884,
+      "step": 35123
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.951203166111727e-07,
+      "loss": 0.914,
+      "step": 35124
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.949147075881942e-07,
+      "loss": 0.8313,
+      "step": 35125
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.947091691416726e-07,
+      "loss": 1.0034,
+      "step": 35126
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9450370127310336e-07,
+      "loss": 0.7514,
+      "step": 35127
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9429830398398194e-07,
+      "loss": 0.8386,
+      "step": 35128
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9409297727580166e-07,
+      "loss": 0.8474,
+      "step": 35129
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9388772115005457e-07,
+      "loss": 0.9581,
+      "step": 35130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2.9368253560823403e-07,
+      "loss": 0.6845,
+      "step": 35131
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9347742065183094e-07,
+      "loss": 0.9379,
+      "step": 35132
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.932723762823375e-07,
+      "loss": 0.9342,
+      "step": 35133
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.93067402501247e-07,
+      "loss": 0.8977,
+      "step": 35134
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9286249931004704e-07,
+      "loss": 0.7991,
+      "step": 35135
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.926576667102299e-07,
+      "loss": 0.9706,
+      "step": 35136
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9245290470328315e-07,
+      "loss": 0.904,
+      "step": 35137
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9224821329069786e-07,
+      "loss": 0.6288,
+      "step": 35138
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9204359247396064e-07,
+      "loss": 0.8533,
+      "step": 35139
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9183904225456137e-07,
+      "loss": 0.6997,
+      "step": 35140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.916345626339867e-07,
+      "loss": 0.8687,
+      "step": 35141
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9143015361372205e-07,
+      "loss": 0.6754,
+      "step": 35142
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.912258151952574e-07,
+      "loss": 0.6617,
+      "step": 35143
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9102154738007593e-07,
+      "loss": 0.6428,
+      "step": 35144
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9081735016966205e-07,
+      "loss": 0.8736,
+      "step": 35145
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.906132235655046e-07,
+      "loss": 1.0944,
+      "step": 35146
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9040916756908566e-07,
+      "loss": 0.7387,
+      "step": 35147
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.9020518218188963e-07,
+      "loss": 0.5675,
+      "step": 35148
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.900012674054009e-07,
+      "loss": 0.6919,
+      "step": 35149
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.897974232411005e-07,
+      "loss": 0.9359,
+      "step": 35150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8959364969046943e-07,
+      "loss": 0.5294,
+      "step": 35151
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8938994675499434e-07,
+      "loss": 0.7139,
+      "step": 35152
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8918631443615286e-07,
+      "loss": 0.7764,
+      "step": 35153
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.889827527354261e-07,
+      "loss": 0.9614,
+      "step": 35154
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8877926165429615e-07,
+      "loss": 0.944,
+      "step": 35155
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8857584119424187e-07,
+      "loss": 0.8763,
+      "step": 35156
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.883724913567421e-07,
+      "loss": 1.0391,
+      "step": 35157
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8816921214327554e-07,
+      "loss": 0.9868,
+      "step": 35158
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.879660035553222e-07,
+      "loss": 1.0524,
+      "step": 35159
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.877628655943576e-07,
+      "loss": 0.7928,
+      "step": 35160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.875597982618605e-07,
+      "loss": 0.7236,
+      "step": 35161
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8735680155930643e-07,
+      "loss": 0.994,
+      "step": 35162
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.87153875488172e-07,
+      "loss": 0.9311,
+      "step": 35163
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.869510200499337e-07,
+      "loss": 0.7154,
+      "step": 35164
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.867482352460671e-07,
+      "loss": 0.8901,
+      "step": 35165
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8654552107804544e-07,
+      "loss": 0.8617,
+      "step": 35166
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.863428775473431e-07,
+      "loss": 0.9239,
+      "step": 35167
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.861403046554334e-07,
+      "loss": 1.0424,
+      "step": 35168
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.859378024037918e-07,
+      "loss": 0.6806,
+      "step": 35169
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8573537079388814e-07,
+      "loss": 0.9967,
+      "step": 35170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.855330098271958e-07,
+      "loss": 1.0745,
+      "step": 35171
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8533071950518574e-07,
+      "loss": 0.9614,
+      "step": 35172
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8512849982933024e-07,
+      "loss": 0.8944,
+      "step": 35173
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8492635080109797e-07,
+      "loss": 1.0765,
+      "step": 35174
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.847242724219612e-07,
+      "loss": 0.6312,
+      "step": 35175
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8452226469338653e-07,
+      "loss": 0.7254,
+      "step": 35176
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8432032761684715e-07,
+      "loss": 0.7037,
+      "step": 35177
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.841184611938075e-07,
+      "loss": 0.9126,
+      "step": 35178
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.839166654257397e-07,
+      "loss": 0.7947,
+      "step": 35179
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8371494031410704e-07,
+      "loss": 0.5596,
+      "step": 35180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.835132858603784e-07,
+      "loss": 1.1666,
+      "step": 35181
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8331170206602034e-07,
+      "loss": 0.6933,
+      "step": 35182
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.831101889324983e-07,
+      "loss": 0.6585,
+      "step": 35183
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.829087464612779e-07,
+      "loss": 0.6014,
+      "step": 35184
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.827073746538256e-07,
+      "loss": 0.6435,
+      "step": 35185
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8250607351160256e-07,
+      "loss": 0.834,
+      "step": 35186
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.823048430360753e-07,
+      "loss": 0.6857,
+      "step": 35187
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8210368322870497e-07,
+      "loss": 0.6647,
+      "step": 35188
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.819025940909592e-07,
+      "loss": 0.6597,
+      "step": 35189
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8170157562429466e-07,
+      "loss": 0.7209,
+      "step": 35190
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8150062783017574e-07,
+      "loss": 0.738,
+      "step": 35191
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8129975071006344e-07,
+      "loss": 0.7062,
+      "step": 35192
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8109894426541775e-07,
+      "loss": 0.9711,
+      "step": 35193
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8089820849769966e-07,
+      "loss": 0.9025,
+      "step": 35194
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.806975434083703e-07,
+      "loss": 0.8233,
+      "step": 35195
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8049694899888737e-07,
+      "loss": 0.7155,
+      "step": 35196
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8029642527070966e-07,
+      "loss": 1.0846,
+      "step": 35197
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.8009597222529607e-07,
+      "loss": 0.6473,
+      "step": 35198
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7989558986410426e-07,
+      "loss": 0.7413,
+      "step": 35199
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.796952781885909e-07,
+      "loss": 0.7721,
+      "step": 35200
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.794950372002125e-07,
+      "loss": 0.8092,
+      "step": 35201
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7929486690042586e-07,
+      "loss": 0.8438,
+      "step": 35202
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.790947672906863e-07,
+      "loss": 0.6183,
+      "step": 35203
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.788947383724494e-07,
+      "loss": 0.9363,
+      "step": 35204
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7869478014716953e-07,
+      "loss": 0.8112,
+      "step": 35205
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.784948926162989e-07,
+      "loss": 0.8982,
+      "step": 35206
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.782950757812952e-07,
+      "loss": 0.5545,
+      "step": 35207
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7809532964360843e-07,
+      "loss": 0.9837,
+      "step": 35208
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.778956542046929e-07,
+      "loss": 0.8629,
+      "step": 35209
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.776960494659986e-07,
+      "loss": 0.8091,
+      "step": 35210
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7749651542897883e-07,
+      "loss": 0.7983,
+      "step": 35211
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.772970520950835e-07,
+      "loss": 0.7131,
+      "step": 35212
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.770976594657637e-07,
+      "loss": 0.8304,
+      "step": 35213
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7689833754247055e-07,
+      "loss": 0.9229,
+      "step": 35214
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7669908632665164e-07,
+      "loss": 1.0298,
+      "step": 35215
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7649990581975705e-07,
+      "loss": 0.72,
+      "step": 35216
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7630079602323447e-07,
+      "loss": 0.7421,
+      "step": 35217
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.761017569385327e-07,
+      "loss": 0.6557,
+      "step": 35218
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.759027885670984e-07,
+      "loss": 0.497,
+      "step": 35219
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.757038909103793e-07,
+      "loss": 0.7733,
+      "step": 35220
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.755050639698209e-07,
+      "loss": 0.8556,
+      "step": 35221
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7530630774686986e-07,
+      "loss": 0.8281,
+      "step": 35222
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.751076222429705e-07,
+      "loss": 0.7515,
+      "step": 35223
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.749090074595684e-07,
+      "loss": 0.8442,
+      "step": 35224
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.74710463398109e-07,
+      "loss": 0.807,
+      "step": 35225
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.745119900600346e-07,
+      "loss": 0.9183,
+      "step": 35226
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.743135874467884e-07,
+      "loss": 0.6215,
+      "step": 35227
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.741152555598159e-07,
+      "loss": 0.8091,
+      "step": 35228
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7391699440055596e-07,
+      "loss": 0.5911,
+      "step": 35229
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7371880397045083e-07,
+      "loss": 0.9599,
+      "step": 35230
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.735206842709437e-07,
+      "loss": 0.7337,
+      "step": 35231
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7332263530347347e-07,
+      "loss": 0.7318,
+      "step": 35232
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.731246570694812e-07,
+      "loss": 0.9581,
+      "step": 35233
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.72926749570408e-07,
+      "loss": 0.7419,
+      "step": 35234
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7272891280769044e-07,
+      "loss": 0.6272,
+      "step": 35235
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.725311467827685e-07,
+      "loss": 0.7237,
+      "step": 35236
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.723334514970799e-07,
+      "loss": 0.8242,
+      "step": 35237
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7213582695206355e-07,
+      "loss": 0.652,
+      "step": 35238
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.71938273149156e-07,
+      "loss": 0.7527,
+      "step": 35239
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7174079008979283e-07,
+      "loss": 0.645,
+      "step": 35240
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.715433777754106e-07,
+      "loss": 0.8642,
+      "step": 35241
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.713460362074449e-07,
+      "loss": 0.7465,
+      "step": 35242
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7114876538733235e-07,
+      "loss": 0.7887,
+      "step": 35243
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.709515653165051e-07,
+      "loss": 0.7579,
+      "step": 35244
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7075443599639986e-07,
+      "loss": 1.0,
+      "step": 35245
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7055737742844866e-07,
+      "loss": 0.7516,
+      "step": 35246
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7036038961408383e-07,
+      "loss": 0.7608,
+      "step": 35247
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.7016347255473863e-07,
+      "loss": 0.7917,
+      "step": 35248
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.699666262518463e-07,
+      "loss": 0.8872,
+      "step": 35249
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.697698507068358e-07,
+      "loss": 0.8035,
+      "step": 35250
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6957314592114036e-07,
+      "loss": 0.6843,
+      "step": 35251
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6937651189618997e-07,
+      "loss": 1.0699,
+      "step": 35252
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6917994863341346e-07,
+      "loss": 1.0033,
+      "step": 35253
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.689834561342408e-07,
+      "loss": 0.7945,
+      "step": 35254
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6878703440009977e-07,
+      "loss": 0.6542,
+      "step": 35255
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.685906834324214e-07,
+      "loss": 0.8674,
+      "step": 35256
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6839440323263243e-07,
+      "loss": 0.8634,
+      "step": 35257
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6819819380216163e-07,
+      "loss": 0.731,
+      "step": 35258
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6800205514243225e-07,
+      "loss": 1.0522,
+      "step": 35259
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.678059872548722e-07,
+      "loss": 0.9228,
+      "step": 35260
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.676099901409079e-07,
+      "loss": 0.7078,
+      "step": 35261
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.674140638019651e-07,
+      "loss": 0.5715,
+      "step": 35262
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.672182082394681e-07,
+      "loss": 0.9888,
+      "step": 35263
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.670224234548413e-07,
+      "loss": 0.8171,
+      "step": 35264
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6682670944950804e-07,
+      "loss": 0.8422,
+      "step": 35265
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6663106622489164e-07,
+      "loss": 1.0928,
+      "step": 35266
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.664354937824132e-07,
+      "loss": 0.9277,
+      "step": 35267
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6623999212350037e-07,
+      "loss": 0.6778,
+      "step": 35268
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.660445612495688e-07,
+      "loss": 0.8679,
+      "step": 35269
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6584920116204283e-07,
+      "loss": 0.879,
+      "step": 35270
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.656539118623425e-07,
+      "loss": 0.629,
+      "step": 35271
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.654586933518877e-07,
+      "loss": 0.8744,
+      "step": 35272
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.652635456320973e-07,
+      "loss": 0.8016,
+      "step": 35273
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.650684687043925e-07,
+      "loss": 0.6468,
+      "step": 35274
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6487346257019095e-07,
+      "loss": 0.8438,
+      "step": 35275
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6467852723091046e-07,
+      "loss": 0.5349,
+      "step": 35276
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.644836626879688e-07,
+      "loss": 0.8199,
+      "step": 35277
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.642888689427825e-07,
+      "loss": 0.9889,
+      "step": 35278
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.640941459967694e-07,
+      "loss": 0.6951,
+      "step": 35279
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.638994938513451e-07,
+      "loss": 0.8023,
+      "step": 35280
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.637049125079238e-07,
+      "loss": 0.8357,
+      "step": 35281
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.635104019679224e-07,
+      "loss": 1.0623,
+      "step": 35282
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6331596223275414e-07,
+      "loss": 0.9842,
+      "step": 35283
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6312159330383336e-07,
+      "loss": 0.87,
+      "step": 35284
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.629272951825734e-07,
+      "loss": 0.8664,
+      "step": 35285
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6273306787038764e-07,
+      "loss": 0.82,
+      "step": 35286
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6253891136868823e-07,
+      "loss": 1.2001,
+      "step": 35287
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6234482567888853e-07,
+      "loss": 0.7321,
+      "step": 35288
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6215081080239847e-07,
+      "loss": 1.1238,
+      "step": 35289
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6195686674062804e-07,
+      "loss": 0.933,
+      "step": 35290
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6176299349498725e-07,
+      "loss": 0.8177,
+      "step": 35291
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.615691910668894e-07,
+      "loss": 0.8203,
+      "step": 35292
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.613754594577411e-07,
+      "loss": 0.5923,
+      "step": 35293
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6118179866895245e-07,
+      "loss": 0.9067,
+      "step": 35294
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.609882087019311e-07,
+      "loss": 0.891,
+      "step": 35295
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.607946895580848e-07,
+      "loss": 0.9141,
+      "step": 35296
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6060124123882035e-07,
+      "loss": 1.1743,
+      "step": 35297
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.604078637455465e-07,
+      "loss": 0.7746,
+      "step": 35298
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.6021455707966767e-07,
+      "loss": 0.6005,
+      "step": 35299
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.600213212425895e-07,
+      "loss": 0.9169,
+      "step": 35300
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.598281562357186e-07,
+      "loss": 0.7445,
+      "step": 35301
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.596350620604593e-07,
+      "loss": 0.75,
+      "step": 35302
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5944203871821393e-07,
+      "loss": 0.7596,
+      "step": 35303
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5924908621038913e-07,
+      "loss": 0.6882,
+      "step": 35304
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.59056204538386e-07,
+      "loss": 1.0246,
+      "step": 35305
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.588633937036078e-07,
+      "loss": 0.736,
+      "step": 35306
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.586706537074579e-07,
+      "loss": 0.7423,
+      "step": 35307
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.584779845513363e-07,
+      "loss": 1.0358,
+      "step": 35308
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.58285386236643e-07,
+      "loss": 0.9529,
+      "step": 35309
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.580928587647824e-07,
+      "loss": 0.832,
+      "step": 35310
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.579004021371512e-07,
+      "loss": 0.8548,
+      "step": 35311
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5770801635515043e-07,
+      "loss": 0.847,
+      "step": 35312
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.575157014201779e-07,
+      "loss": 0.6986,
+      "step": 35313
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.573234573336336e-07,
+      "loss": 0.7484,
+      "step": 35314
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5713128409691537e-07,
+      "loss": 0.7714,
+      "step": 35315
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5693918171141976e-07,
+      "loss": 0.8979,
+      "step": 35316
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5674715017854454e-07,
+      "loss": 0.8266,
+      "step": 35317
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.565551894996876e-07,
+      "loss": 0.8807,
+      "step": 35318
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.563632996762422e-07,
+      "loss": 0.9946,
+      "step": 35319
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5617148070960494e-07,
+      "loss": 0.4967,
+      "step": 35320
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5597973260116927e-07,
+      "loss": 0.9893,
+      "step": 35321
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5578805535233175e-07,
+      "loss": 1.0155,
+      "step": 35322
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.555964489644858e-07,
+      "loss": 0.7172,
+      "step": 35323
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5540491343902576e-07,
+      "loss": 0.8848,
+      "step": 35324
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5521344877734165e-07,
+      "loss": 0.8415,
+      "step": 35325
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.55022054980828e-07,
+      "loss": 0.9404,
+      "step": 35326
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5483073205087695e-07,
+      "loss": 0.4527,
+      "step": 35327
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.546394799888785e-07,
+      "loss": 0.6348,
+      "step": 35328
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.544482987962238e-07,
+      "loss": 0.8068,
+      "step": 35329
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.542571884743028e-07,
+      "loss": 0.6968,
+      "step": 35330
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.540661490245067e-07,
+      "loss": 1.1439,
+      "step": 35331
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5387518044822313e-07,
+      "loss": 0.916,
+      "step": 35332
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.536842827468411e-07,
+      "loss": 0.9155,
+      "step": 35333
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.534934559217495e-07,
+      "loss": 0.8865,
+      "step": 35334
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.53302699974336e-07,
+      "loss": 1.0064,
+      "step": 35335
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5311201490598847e-07,
+      "loss": 0.8538,
+      "step": 35336
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5292140071809136e-07,
+      "loss": 0.8191,
+      "step": 35337
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.527308574120346e-07,
+      "loss": 0.8139,
+      "step": 35338
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.525403849891983e-07,
+      "loss": 0.7199,
+      "step": 35339
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.523499834509724e-07,
+      "loss": 0.9552,
+      "step": 35340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.521596527987402e-07,
+      "loss": 0.6497,
+      "step": 35341
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.519693930338851e-07,
+      "loss": 0.5877,
+      "step": 35342
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5177920415779045e-07,
+      "loss": 0.7499,
+      "step": 35343
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5158908617184066e-07,
+      "loss": 1.0256,
+      "step": 35344
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.513990390774168e-07,
+      "loss": 0.7954,
+      "step": 35345
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.512090628759012e-07,
+      "loss": 0.5838,
+      "step": 35346
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.510191575686771e-07,
+      "loss": 1.0757,
+      "step": 35347
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5082932315712573e-07,
+      "loss": 0.6608,
+      "step": 35348
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5063955964262475e-07,
+      "loss": 0.9999,
+      "step": 35349
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.504498670265554e-07,
+      "loss": 0.8647,
+      "step": 35350
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.5026024531029645e-07,
+      "loss": 1.0248,
+      "step": 35351
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.50070694495228e-07,
+      "loss": 0.6071,
+      "step": 35352
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.498812145827278e-07,
+      "loss": 0.7109,
+      "step": 35353
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4969180557417594e-07,
+      "loss": 0.8088,
+      "step": 35354
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.495024674709468e-07,
+      "loss": 0.8747,
+      "step": 35355
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.493132002744181e-07,
+      "loss": 0.911,
+      "step": 35356
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.491240039859677e-07,
+      "loss": 0.698,
+      "step": 35357
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4893487860697005e-07,
+      "loss": 0.6457,
+      "step": 35358
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.487458241387997e-07,
+      "loss": 0.8244,
+      "step": 35359
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4855684058283423e-07,
+      "loss": 0.6462,
+      "step": 35360
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4836792794044495e-07,
+      "loss": 0.9508,
+      "step": 35361
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4817908621300735e-07,
+      "loss": 0.928,
+      "step": 35362
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4799031540189476e-07,
+      "loss": 0.8343,
+      "step": 35363
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4780161550847834e-07,
+      "loss": 0.8687,
+      "step": 35364
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.476129865341326e-07,
+      "loss": 0.9094,
+      "step": 35365
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.474244284802285e-07,
+      "loss": 0.8693,
+      "step": 35366
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4723594134813623e-07,
+      "loss": 0.8631,
+      "step": 35367
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4704752513922903e-07,
+      "loss": 0.7513,
+      "step": 35368
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4685917985487363e-07,
+      "loss": 1.1451,
+      "step": 35369
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4667090549644e-07,
+      "loss": 0.8988,
+      "step": 35370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4648270206530045e-07,
+      "loss": 0.7254,
+      "step": 35371
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.462945695628216e-07,
+      "loss": 0.6384,
+      "step": 35372
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4610650799037126e-07,
+      "loss": 0.8031,
+      "step": 35373
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.459185173493184e-07,
+      "loss": 0.921,
+      "step": 35374
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4573059764102846e-07,
+      "loss": 0.6965,
+      "step": 35375
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.455427488668694e-07,
+      "loss": 0.6271,
+      "step": 35376
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.453549710282044e-07,
+      "loss": 0.8339,
+      "step": 35377
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.451672641264047e-07,
+      "loss": 0.7372,
+      "step": 35378
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4497962816283027e-07,
+      "loss": 0.8307,
+      "step": 35379
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4479206313884784e-07,
+      "loss": 0.6665,
+      "step": 35380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4460456905582075e-07,
+      "loss": 0.8054,
+      "step": 35381
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4441714591511124e-07,
+      "loss": 0.8148,
+      "step": 35382
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.442297937180838e-07,
+      "loss": 0.6785,
+      "step": 35383
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4404251246610165e-07,
+      "loss": 0.6724,
+      "step": 35384
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.43855302160525e-07,
+      "loss": 0.9031,
+      "step": 35385
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.436681628027171e-07,
+      "loss": 0.757,
+      "step": 35386
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4348109439403687e-07,
+      "loss": 0.9113,
+      "step": 35387
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.432940969358455e-07,
+      "loss": 0.4685,
+      "step": 35388
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4310717042950295e-07,
+      "loss": 0.7821,
+      "step": 35389
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4292031487636923e-07,
+      "loss": 0.9525,
+      "step": 35390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4273353027780224e-07,
+      "loss": 0.8315,
+      "step": 35391
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.425468166351597e-07,
+      "loss": 1.0545,
+      "step": 35392
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.423601739498016e-07,
+      "loss": 0.5132,
+      "step": 35393
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.421736022230825e-07,
+      "loss": 0.8328,
+      "step": 35394
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4198710145636015e-07,
+      "loss": 0.9062,
+      "step": 35395
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.418006716509924e-07,
+      "loss": 0.6642,
+      "step": 35396
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.416143128083337e-07,
+      "loss": 0.8775,
+      "step": 35397
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4142802492974073e-07,
+      "loss": 0.6875,
+      "step": 35398
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.412418080165646e-07,
+      "loss": 0.8926,
+      "step": 35399
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4105566207016207e-07,
+      "loss": 0.8478,
+      "step": 35400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.4086958709188645e-07,
+      "loss": 0.7106,
+      "step": 35401
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.406835830830923e-07,
+      "loss": 0.7588,
+      "step": 35402
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.404976500451295e-07,
+      "loss": 0.8203,
+      "step": 35403
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.403117879793515e-07,
+      "loss": 0.7127,
+      "step": 35404
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.401259968871106e-07,
+      "loss": 0.6592,
+      "step": 35405
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.399402767697578e-07,
+      "loss": 0.8351,
+      "step": 35406
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.397546276286411e-07,
+      "loss": 0.838,
+      "step": 35407
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3956904946511486e-07,
+      "loss": 0.8283,
+      "step": 35408
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.393835422805257e-07,
+      "loss": 0.7903,
+      "step": 35409
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3919810607622273e-07,
+      "loss": 0.8741,
+      "step": 35410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3901274085355585e-07,
+      "loss": 0.9351,
+      "step": 35411
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.388274466138718e-07,
+      "loss": 0.6483,
+      "step": 35412
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3864222335851727e-07,
+      "loss": 1.013,
+      "step": 35413
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.384570710888423e-07,
+      "loss": 0.9671,
+      "step": 35414
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3827198980619025e-07,
+      "loss": 0.9851,
+      "step": 35415
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3808697951190895e-07,
+      "loss": 0.9019,
+      "step": 35416
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3790204020734286e-07,
+      "loss": 0.8441,
+      "step": 35417
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3771717189383757e-07,
+      "loss": 0.6495,
+      "step": 35418
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.375323745727376e-07,
+      "loss": 0.3962,
+      "step": 35419
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3734764824538514e-07,
+      "loss": 0.7933,
+      "step": 35420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3716299291312583e-07,
+      "loss": 0.6498,
+      "step": 35421
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.369784085773008e-07,
+      "loss": 0.7342,
+      "step": 35422
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3679389523925233e-07,
+      "loss": 0.6186,
+      "step": 35423
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3660945290032378e-07,
+      "loss": 0.7401,
+      "step": 35424
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3642508156185517e-07,
+      "loss": 0.9664,
+      "step": 35425
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3624078122518657e-07,
+      "loss": 0.9615,
+      "step": 35426
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3605655189166022e-07,
+      "loss": 0.91,
+      "step": 35427
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3587239356261615e-07,
+      "loss": 0.8046,
+      "step": 35428
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3568830623939108e-07,
+      "loss": 0.6479,
+      "step": 35429
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3550428992332508e-07,
+      "loss": 0.9214,
+      "step": 35430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.353203446157548e-07,
+      "loss": 0.5935,
+      "step": 35431
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3513647031802144e-07,
+      "loss": 0.7508,
+      "step": 35432
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3495266703145836e-07,
+      "loss": 0.7632,
+      "step": 35433
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3476893475740448e-07,
+      "loss": 0.482,
+      "step": 35434
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.345852734971965e-07,
+      "loss": 1.0182,
+      "step": 35435
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3440168325216782e-07,
+      "loss": 0.8839,
+      "step": 35436
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3421816402365406e-07,
+      "loss": 0.6637,
+      "step": 35437
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3403471581299076e-07,
+      "loss": 1.0637,
+      "step": 35438
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3385133862151133e-07,
+      "loss": 0.5654,
+      "step": 35439
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3366803245055025e-07,
+      "loss": 0.9736,
+      "step": 35440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.334847973014387e-07,
+      "loss": 0.7032,
+      "step": 35441
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3330163317551113e-07,
+      "loss": 0.813,
+      "step": 35442
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3311854007409874e-07,
+      "loss": 0.7068,
+      "step": 35443
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3293551799853152e-07,
+      "loss": 0.805,
+      "step": 35444
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.327525669501418e-07,
+      "loss": 1.0174,
+      "step": 35445
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3256968693026182e-07,
+      "loss": 0.9511,
+      "step": 35446
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.323868779402183e-07,
+      "loss": 0.7471,
+      "step": 35447
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.322041399813435e-07,
+      "loss": 0.9798,
+      "step": 35448
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3202147305496302e-07,
+      "loss": 1.091,
+      "step": 35449
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3183887716240805e-07,
+      "loss": 0.5958,
+      "step": 35450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3165635230500637e-07,
+      "loss": 0.8327,
+      "step": 35451
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3147389848408364e-07,
+      "loss": 0.9871,
+      "step": 35452
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3129151570096653e-07,
+      "loss": 0.613,
+      "step": 35453
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.31109203956984e-07,
+      "loss": 0.8308,
+      "step": 35454
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3092696325345943e-07,
+      "loss": 0.9102,
+      "step": 35455
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3074479359171732e-07,
+      "loss": 1.1799,
+      "step": 35456
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3056269497308548e-07,
+      "loss": 0.8884,
+      "step": 35457
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.303806673988873e-07,
+      "loss": 0.9546,
+      "step": 35458
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3019871087044509e-07,
+      "loss": 0.566,
+      "step": 35459
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.3001682538908333e-07,
+      "loss": 0.9244,
+      "step": 35460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2983501095612314e-07,
+      "loss": 1.017,
+      "step": 35461
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2965326757288686e-07,
+      "loss": 0.9404,
+      "step": 35462
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2947159524069896e-07,
+      "loss": 0.9622,
+      "step": 35463
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2928999396087726e-07,
+      "loss": 0.8139,
+      "step": 35464
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2910846373474404e-07,
+      "loss": 0.6615,
+      "step": 35465
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2892700456361939e-07,
+      "loss": 0.8409,
+      "step": 35466
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2874561644882332e-07,
+      "loss": 0.8165,
+      "step": 35467
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2856429939167257e-07,
+      "loss": 0.8358,
+      "step": 35468
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2838305339348833e-07,
+      "loss": 0.9035,
+      "step": 35469
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.282018784555884e-07,
+      "loss": 0.6543,
+      "step": 35470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.280207745792884e-07,
+      "loss": 0.6379,
+      "step": 35471
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2783974176590616e-07,
+      "loss": 0.6768,
+      "step": 35472
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2765878001675956e-07,
+      "loss": 0.8939,
+      "step": 35473
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2747788933316195e-07,
+      "loss": 0.6524,
+      "step": 35474
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2729706971643117e-07,
+      "loss": 0.7982,
+      "step": 35475
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2711632116788173e-07,
+      "loss": 0.5012,
+      "step": 35476
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2693564368882815e-07,
+      "loss": 0.4999,
+      "step": 35477
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.267550372805838e-07,
+      "loss": 0.8344,
+      "step": 35478
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2657450194446206e-07,
+      "loss": 0.726,
+      "step": 35479
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2639403768177416e-07,
+      "loss": 0.6716,
+      "step": 35480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2621364449383455e-07,
+      "loss": 0.6884,
+      "step": 35481
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2603332238195553e-07,
+      "loss": 0.7607,
+      "step": 35482
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2585307134744828e-07,
+      "loss": 0.7751,
+      "step": 35483
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2567289139162175e-07,
+      "loss": 0.808,
+      "step": 35484
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.254927825157871e-07,
+      "loss": 0.8034,
+      "step": 35485
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2531274472125551e-07,
+      "loss": 0.572,
+      "step": 35486
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2513277800933374e-07,
+      "loss": 0.9309,
+      "step": 35487
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2495288238133295e-07,
+      "loss": 0.731,
+      "step": 35488
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2477305783856097e-07,
+      "loss": 0.7297,
+      "step": 35489
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.245933043823234e-07,
+      "loss": 0.9357,
+      "step": 35490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2441362201392924e-07,
+      "loss": 0.5968,
+      "step": 35491
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.242340107346841e-07,
+      "loss": 0.6698,
+      "step": 35492
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.240544705458958e-07,
+      "loss": 0.7131,
+      "step": 35493
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2387500144886887e-07,
+      "loss": 0.8998,
+      "step": 35494
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2369560344490782e-07,
+      "loss": 0.7272,
+      "step": 35495
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2351627653531937e-07,
+      "loss": 0.9006,
+      "step": 35496
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2333702072140473e-07,
+      "loss": 0.8564,
+      "step": 35497
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.231578360044695e-07,
+      "loss": 0.841,
+      "step": 35498
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2297872238581707e-07,
+      "loss": 0.6426,
+      "step": 35499
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2279967986674754e-07,
+      "loss": 0.7622,
+      "step": 35500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2262070844856543e-07,
+      "loss": 0.7084,
+      "step": 35501
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2244180813257187e-07,
+      "loss": 1.0745,
+      "step": 35502
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2226297892006588e-07,
+      "loss": 0.5147,
+      "step": 35503
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.220842208123497e-07,
+      "loss": 0.7288,
+      "step": 35504
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2190553381072234e-07,
+      "loss": 0.7554,
+      "step": 35505
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2172691791648494e-07,
+      "loss": 0.575,
+      "step": 35506
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2154837313093424e-07,
+      "loss": 0.7036,
+      "step": 35507
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2136989945537146e-07,
+      "loss": 0.6005,
+      "step": 35508
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2119149689109109e-07,
+      "loss": 0.9584,
+      "step": 35509
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.210131654393921e-07,
+      "loss": 0.9464,
+      "step": 35510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2.2083490510157125e-07,
+      "loss": 0.8013,
+      "step": 35511
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2065671587892523e-07,
+      "loss": 0.9906,
+      "step": 35512
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.204785977727497e-07,
+      "loss": 0.8545,
+      "step": 35513
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2030055078434032e-07,
+      "loss": 0.9259,
+      "step": 35514
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.2012257491499046e-07,
+      "loss": 0.7849,
+      "step": 35515
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1994467016599574e-07,
+      "loss": 0.7932,
+      "step": 35516
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.197668365386485e-07,
+      "loss": 0.7699,
+      "step": 35517
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1958907403424435e-07,
+      "loss": 0.7721,
+      "step": 35518
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1941138265407335e-07,
+      "loss": 0.4015,
+      "step": 35519
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1923376239942895e-07,
+      "loss": 0.7474,
+      "step": 35520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1905621327160342e-07,
+      "loss": 0.6715,
+      "step": 35521
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1887873527188686e-07,
+      "loss": 0.8274,
+      "step": 35522
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1870132840156931e-07,
+      "loss": 0.992,
+      "step": 35523
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1852399266194312e-07,
+      "loss": 0.9932,
+      "step": 35524
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1834672805429723e-07,
+      "loss": 0.816,
+      "step": 35525
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1816953457991952e-07,
+      "loss": 0.7029,
+      "step": 35526
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1799241224009892e-07,
+      "loss": 0.7273,
+      "step": 35527
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1781536103612443e-07,
+      "loss": 0.8718,
+      "step": 35528
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.176383809692817e-07,
+      "loss": 0.5425,
+      "step": 35529
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1746147204085966e-07,
+      "loss": 0.8198,
+      "step": 35530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1728463425214508e-07,
+      "loss": 0.7109,
+      "step": 35531
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1710786760442138e-07,
+      "loss": 0.885,
+      "step": 35532
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.169311720989764e-07,
+      "loss": 0.6802,
+      "step": 35533
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.167545477370947e-07,
+      "loss": 0.8797,
+      "step": 35534
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1657799452005856e-07,
+      "loss": 0.663,
+      "step": 35535
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1640151244915585e-07,
+      "loss": 0.9189,
+      "step": 35536
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1622510152566668e-07,
+      "loss": 0.9105,
+      "step": 35537
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1604876175087664e-07,
+      "loss": 0.9839,
+      "step": 35538
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1587249312606474e-07,
+      "loss": 1.0166,
+      "step": 35539
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.156962956525155e-07,
+      "loss": 0.7101,
+      "step": 35540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1552016933150788e-07,
+      "loss": 0.8295,
+      "step": 35541
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1534411416432422e-07,
+      "loss": 0.9402,
+      "step": 35542
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1516813015224568e-07,
+      "loss": 0.7632,
+      "step": 35543
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1499221729655128e-07,
+      "loss": 0.7396,
+      "step": 35544
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1481637559851999e-07,
+      "loss": 1.0374,
+      "step": 35545
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1464060505942962e-07,
+      "loss": 1.049,
+      "step": 35546
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1446490568056033e-07,
+      "loss": 1.0073,
+      "step": 35547
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1428927746318884e-07,
+      "loss": 0.8847,
+      "step": 35548
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1411372040859302e-07,
+      "loss": 0.8619,
+      "step": 35549
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.139382345180474e-07,
+      "loss": 0.6625,
+      "step": 35550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1376281979283098e-07,
+      "loss": 0.9381,
+      "step": 35551
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1358747623421715e-07,
+      "loss": 0.8299,
+      "step": 35552
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.134122038434816e-07,
+      "loss": 0.8918,
+      "step": 35553
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1323700262189995e-07,
+      "loss": 0.9631,
+      "step": 35554
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1306187257074673e-07,
+      "loss": 0.907,
+      "step": 35555
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1288681369129316e-07,
+      "loss": 0.8596,
+      "step": 35556
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1271182598481377e-07,
+      "loss": 0.8238,
+      "step": 35557
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1253690945258088e-07,
+      "loss": 0.5412,
+      "step": 35558
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1236206409586568e-07,
+      "loss": 0.9643,
+      "step": 35559
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1218728991594163e-07,
+      "loss": 0.8669,
+      "step": 35560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1201258691407766e-07,
+      "loss": 0.5265,
+      "step": 35561
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1183795509154615e-07,
+      "loss": 1.0814,
+      "step": 35562
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1166339444961492e-07,
+      "loss": 1.109,
+      "step": 35563
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.114889049895552e-07,
+      "loss": 0.8185,
+      "step": 35564
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1131448671263378e-07,
+      "loss": 0.7094,
+      "step": 35565
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.111401396201207e-07,
+      "loss": 0.9672,
+      "step": 35566
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1096586371328498e-07,
+      "loss": 0.585,
+      "step": 35567
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1079165899339228e-07,
+      "loss": 0.7084,
+      "step": 35568
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1061752546170933e-07,
+      "loss": 0.9003,
+      "step": 35569
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1044346311950291e-07,
+      "loss": 0.6348,
+      "step": 35570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.102694719680376e-07,
+      "loss": 0.7817,
+      "step": 35571
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.1009555200858012e-07,
+      "loss": 0.8739,
+      "step": 35572
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0992170324239613e-07,
+      "loss": 0.8991,
+      "step": 35573
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0974792567074797e-07,
+      "loss": 0.8982,
+      "step": 35574
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0957421929490018e-07,
+      "loss": 0.9697,
+      "step": 35575
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0940058411611618e-07,
+      "loss": 0.9903,
+      "step": 35576
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0922702013565833e-07,
+      "loss": 0.8288,
+      "step": 35577
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.090535273547889e-07,
+      "loss": 0.8741,
+      "step": 35578
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0888010577476914e-07,
+      "loss": 0.6191,
+      "step": 35579
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0870675539686024e-07,
+      "loss": 0.9712,
+      "step": 35580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0853347622232345e-07,
+      "loss": 0.559,
+      "step": 35581
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0836026825241885e-07,
+      "loss": 0.9114,
+      "step": 35582
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0818713148840543e-07,
+      "loss": 0.7928,
+      "step": 35583
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0801406593154105e-07,
+      "loss": 0.6112,
+      "step": 35584
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0784107158308698e-07,
+      "loss": 0.7182,
+      "step": 35585
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0766814844430106e-07,
+      "loss": 0.8349,
+      "step": 35586
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0749529651643896e-07,
+      "loss": 0.9725,
+      "step": 35587
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0732251580075858e-07,
+      "loss": 0.8775,
+      "step": 35588
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0714980629851554e-07,
+      "loss": 0.9519,
+      "step": 35589
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.069771680109678e-07,
+      "loss": 0.703,
+      "step": 35590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.068046009393687e-07,
+      "loss": 0.7447,
+      "step": 35591
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.06632105084974e-07,
+      "loss": 0.9682,
+      "step": 35592
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.064596804490393e-07,
+      "loss": 1.0061,
+      "step": 35593
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0628732703281696e-07,
+      "loss": 0.6391,
+      "step": 35594
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0611504483756038e-07,
+      "loss": 0.7463,
+      "step": 35595
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0594283386452196e-07,
+      "loss": 0.5142,
+      "step": 35596
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.057706941149551e-07,
+      "loss": 0.7551,
+      "step": 35597
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0559862559011322e-07,
+      "loss": 0.8963,
+      "step": 35598
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0542662829124426e-07,
+      "loss": 0.7326,
+      "step": 35599
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0525470221960054e-07,
+      "loss": 1.0954,
+      "step": 35600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0508284737643326e-07,
+      "loss": 0.9671,
+      "step": 35601
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.049110637629892e-07,
+      "loss": 0.6281,
+      "step": 35602
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.047393513805207e-07,
+      "loss": 0.5101,
+      "step": 35603
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0456771023027566e-07,
+      "loss": 0.4524,
+      "step": 35604
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0439614031350196e-07,
+      "loss": 1.0289,
+      "step": 35605
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.042246416314475e-07,
+      "loss": 0.8803,
+      "step": 35606
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0405321418535907e-07,
+      "loss": 0.7784,
+      "step": 35607
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0388185797648341e-07,
+      "loss": 0.693,
+      "step": 35608
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0371057300606621e-07,
+      "loss": 0.7089,
+      "step": 35609
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0353935927535428e-07,
+      "loss": 0.5106,
+      "step": 35610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0336821678559216e-07,
+      "loss": 0.7602,
+      "step": 35611
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0319714553802438e-07,
+      "loss": 1.2619,
+      "step": 35612
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0302614553389443e-07,
+      "loss": 0.785,
+      "step": 35613
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0285521677444465e-07,
+      "loss": 0.6856,
+      "step": 35614
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0268435926092177e-07,
+      "loss": 0.9569,
+      "step": 35615
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.02513572994566e-07,
+      "loss": 0.6923,
+      "step": 35616
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0234285797661846e-07,
+      "loss": 0.6959,
+      "step": 35617
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0217221420832378e-07,
+      "loss": 0.7027,
+      "step": 35618
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0200164169091875e-07,
+      "loss": 0.7915,
+      "step": 35619
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0183114042564566e-07,
+      "loss": 0.9286,
+      "step": 35620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.016607104137458e-07,
+      "loss": 0.8678,
+      "step": 35621
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0149035165645703e-07,
+      "loss": 0.9251,
+      "step": 35622
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0132006415501836e-07,
+      "loss": 0.823,
+      "step": 35623
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0114984791066884e-07,
+      "loss": 0.6571,
+      "step": 35624
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.009797029246452e-07,
+      "loss": 0.786,
+      "step": 35625
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0080962919818537e-07,
+      "loss": 0.6822,
+      "step": 35626
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0063962673252502e-07,
+      "loss": 0.8357,
+      "step": 35627
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0046969552890317e-07,
+      "loss": 0.9269,
+      "step": 35628
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.0029983558855327e-07,
+      "loss": 0.7524,
+      "step": 35629
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2.001300469127121e-07,
+      "loss": 0.5688,
+      "step": 35630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9996032950261201e-07,
+      "loss": 0.7341,
+      "step": 35631
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.997906833594887e-07,
+      "loss": 0.7662,
+      "step": 35632
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9962110848457673e-07,
+      "loss": 0.729,
+      "step": 35633
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9945160487910842e-07,
+      "loss": 0.9147,
+      "step": 35634
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9928217254431503e-07,
+      "loss": 0.8079,
+      "step": 35635
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9911281148143113e-07,
+      "loss": 0.7618,
+      "step": 35636
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9894352169168795e-07,
+      "loss": 0.769,
+      "step": 35637
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9877430317631453e-07,
+      "loss": 0.6831,
+      "step": 35638
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9860515593654316e-07,
+      "loss": 0.891,
+      "step": 35639
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9843607997360403e-07,
+      "loss": 0.7049,
+      "step": 35640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9826707528872613e-07,
+      "loss": 0.7354,
+      "step": 35641
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9809814188313737e-07,
+      "loss": 0.9137,
+      "step": 35642
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9792927975806898e-07,
+      "loss": 0.8453,
+      "step": 35643
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9776048891474552e-07,
+      "loss": 0.6745,
+      "step": 35644
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9759176935439718e-07,
+      "loss": 0.8649,
+      "step": 35645
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9742312107824956e-07,
+      "loss": 0.8985,
+      "step": 35646
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9725454408753065e-07,
+      "loss": 1.2076,
+      "step": 35647
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9708603838346608e-07,
+      "loss": 0.9882,
+      "step": 35648
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9691760396727823e-07,
+      "loss": 0.8782,
+      "step": 35649
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.96749240840195e-07,
+      "loss": 0.7591,
+      "step": 35650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9658094900343983e-07,
+      "loss": 1.0455,
+      "step": 35651
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9641272845823621e-07,
+      "loss": 0.7677,
+      "step": 35652
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9624457920580876e-07,
+      "loss": 0.8184,
+      "step": 35653
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9607650124737976e-07,
+      "loss": 0.7609,
+      "step": 35654
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.959084945841705e-07,
+      "loss": 0.924,
+      "step": 35655
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.957405592174033e-07,
+      "loss": 0.8564,
+      "step": 35656
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9557269514829835e-07,
+      "loss": 0.7937,
+      "step": 35657
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9540490237808018e-07,
+      "loss": 0.629,
+      "step": 35658
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.952371809079645e-07,
+      "loss": 0.8433,
+      "step": 35659
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9506953073917368e-07,
+      "loss": 1.0472,
+      "step": 35660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.949019518729256e-07,
+      "loss": 0.8279,
+      "step": 35661
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9473444431043932e-07,
+      "loss": 0.541,
+      "step": 35662
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.945670080529316e-07,
+      "loss": 0.8129,
+      "step": 35663
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9439964310162373e-07,
+      "loss": 0.6334,
+      "step": 35664
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9423234945772918e-07,
+      "loss": 0.677,
+      "step": 35665
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9406512712246696e-07,
+      "loss": 0.6391,
+      "step": 35666
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9389797609705163e-07,
+      "loss": 0.8622,
+      "step": 35667
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9373089638269893e-07,
+      "loss": 0.6909,
+      "step": 35668
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9356388798062453e-07,
+      "loss": 0.7004,
+      "step": 35669
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9339695089204192e-07,
+      "loss": 0.6848,
+      "step": 35670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9323008511816566e-07,
+      "loss": 0.7687,
+      "step": 35671
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9306329066021033e-07,
+      "loss": 0.6659,
+      "step": 35672
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9289656751938613e-07,
+      "loss": 0.8944,
+      "step": 35673
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9272991569690868e-07,
+      "loss": 0.6787,
+      "step": 35674
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9256333519398706e-07,
+      "loss": 0.97,
+      "step": 35675
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9239682601183475e-07,
+      "loss": 0.5596,
+      "step": 35676
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9223038815166183e-07,
+      "loss": 0.6776,
+      "step": 35677
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.920640216146785e-07,
+      "loss": 0.9401,
+      "step": 35678
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9189772640209493e-07,
+      "loss": 0.8313,
+      "step": 35679
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9173150251512008e-07,
+      "loss": 0.7843,
+      "step": 35680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.915653499549619e-07,
+      "loss": 1.0047,
+      "step": 35681
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9139926872283166e-07,
+      "loss": 0.718,
+      "step": 35682
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9123325881993394e-07,
+      "loss": 0.6398,
+      "step": 35683
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9106732024747777e-07,
+      "loss": 0.7289,
+      "step": 35684
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9090145300666885e-07,
+      "loss": 0.6148,
+      "step": 35685
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.90735657098714e-07,
+      "loss": 0.9884,
+      "step": 35686
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9056993252481892e-07,
+      "loss": 0.981,
+      "step": 35687
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.904042792861882e-07,
+      "loss": 1.035,
+      "step": 35688
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9023869738402756e-07,
+      "loss": 0.9245,
+      "step": 35689
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9007318681953934e-07,
+      "loss": 1.0725,
+      "step": 35690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8990774759392928e-07,
+      "loss": 0.7368,
+      "step": 35691
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.897423797083997e-07,
+      "loss": 0.9107,
+      "step": 35692
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.895770831641508e-07,
+      "loss": 0.5165,
+      "step": 35693
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8941185796238826e-07,
+      "loss": 0.7326,
+      "step": 35694
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8924670410431223e-07,
+      "loss": 0.8226,
+      "step": 35695
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8908162159112397e-07,
+      "loss": 0.6697,
+      "step": 35696
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8891661042402365e-07,
+      "loss": 0.8921,
+      "step": 35697
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.887516706042114e-07,
+      "loss": 0.7214,
+      "step": 35698
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8858680213288517e-07,
+      "loss": 0.6665,
+      "step": 35699
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.884220050112462e-07,
+      "loss": 0.6832,
+      "step": 35700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8825727924049242e-07,
+      "loss": 0.8784,
+      "step": 35701
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8809262482182067e-07,
+      "loss": 0.9125,
+      "step": 35702
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8792804175642888e-07,
+      "loss": 0.7246,
+      "step": 35703
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8776353004551384e-07,
+      "loss": 0.6528,
+      "step": 35704
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8759908969027242e-07,
+      "loss": 0.8739,
+      "step": 35705
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8743472069189915e-07,
+      "loss": 0.9709,
+      "step": 35706
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8727042305159093e-07,
+      "loss": 0.7117,
+      "step": 35707
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.871061967705423e-07,
+      "loss": 0.8933,
+      "step": 35708
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.869420418499468e-07,
+      "loss": 0.8267,
+      "step": 35709
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8677795829099898e-07,
+      "loss": 0.8558,
+      "step": 35710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8661394609489013e-07,
+      "loss": 0.6358,
+      "step": 35711
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8645000526281487e-07,
+      "loss": 1.1152,
+      "step": 35712
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8628613579596556e-07,
+      "loss": 0.7501,
+      "step": 35713
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8612233769553346e-07,
+      "loss": 0.9093,
+      "step": 35714
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8595861096270874e-07,
+      "loss": 0.9205,
+      "step": 35715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8579495559868376e-07,
+      "loss": 0.7325,
+      "step": 35716
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.856313716046465e-07,
+      "loss": 0.8891,
+      "step": 35717
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.854678589817882e-07,
+      "loss": 0.7586,
+      "step": 35718
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8530441773129792e-07,
+      "loss": 0.8613,
+      "step": 35719
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.851410478543636e-07,
+      "loss": 0.7072,
+      "step": 35720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8497774935217427e-07,
+      "loss": 0.8652,
+      "step": 35721
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8481452222591567e-07,
+      "loss": 0.7536,
+      "step": 35722
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8465136647677572e-07,
+      "loss": 0.5999,
+      "step": 35723
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8448828210594126e-07,
+      "loss": 0.8364,
+      "step": 35724
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8432526911459803e-07,
+      "loss": 0.8638,
+      "step": 35725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8416232750393172e-07,
+      "loss": 0.7646,
+      "step": 35726
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.839994572751258e-07,
+      "loss": 0.8499,
+      "step": 35727
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8383665842936826e-07,
+      "loss": 1.0305,
+      "step": 35728
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8367393096783815e-07,
+      "loss": 0.6739,
+      "step": 35729
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8351127489172227e-07,
+      "loss": 1.063,
+      "step": 35730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8334869020220304e-07,
+      "loss": 0.863,
+      "step": 35731
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.831861769004617e-07,
+      "loss": 1.0336,
+      "step": 35732
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8302373498768068e-07,
+      "loss": 0.8576,
+      "step": 35733
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8286136446504121e-07,
+      "loss": 0.8017,
+      "step": 35734
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.826990653337235e-07,
+      "loss": 0.781,
+      "step": 35735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.825368375949088e-07,
+      "loss": 0.8664,
+      "step": 35736
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.823746812497762e-07,
+      "loss": 0.3701,
+      "step": 35737
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8221259629950694e-07,
+      "loss": 1.1324,
+      "step": 35738
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8205058274527566e-07,
+      "loss": 0.6703,
+      "step": 35739
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.818886405882636e-07,
+      "loss": 1.1612,
+      "step": 35740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8172676982964765e-07,
+      "loss": 0.7483,
+      "step": 35741
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8156497047060462e-07,
+      "loss": 0.8313,
+      "step": 35742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8140324251231134e-07,
+      "loss": 0.797,
+      "step": 35743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8124158595594466e-07,
+      "loss": 0.7739,
+      "step": 35744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8108000080267918e-07,
+      "loss": 0.8041,
+      "step": 35745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8091848705369064e-07,
+      "loss": 0.8326,
+      "step": 35746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8075704471015254e-07,
+      "loss": 0.8718,
+      "step": 35747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8059567377324062e-07,
+      "loss": 0.6538,
+      "step": 35748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8043437424412614e-07,
+      "loss": 0.9955,
+      "step": 35749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8027314612398483e-07,
+      "loss": 1.0688,
+      "step": 35750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8011198941398688e-07,
+      "loss": 0.792,
+      "step": 35751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7995090411530468e-07,
+      "loss": 0.9163,
+      "step": 35752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7978989022911065e-07,
+      "loss": 0.7671,
+      "step": 35753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.796289477565738e-07,
+      "loss": 0.614,
+      "step": 35754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7946807669886657e-07,
+      "loss": 0.5035,
+      "step": 35755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.79307277057158e-07,
+      "loss": 0.673,
+      "step": 35756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7914654883261717e-07,
+      "loss": 1.272,
+      "step": 35757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7898589202641426e-07,
+      "loss": 0.9937,
+      "step": 35758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.788253066397161e-07,
+      "loss": 0.7348,
+      "step": 35759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7866479267369062e-07,
+      "loss": 0.4868,
+      "step": 35760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7850435012950474e-07,
+      "loss": 0.8012,
+      "step": 35761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7834397900832745e-07,
+      "loss": 0.89,
+      "step": 35762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.781836793113223e-07,
+      "loss": 0.5913,
+      "step": 35763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7802345103965612e-07,
+      "loss": 0.9196,
+      "step": 35764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7786329419449466e-07,
+      "loss": 0.7352,
+      "step": 35765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7770320877700143e-07,
+      "loss": 0.836,
+      "step": 35766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7754319478834103e-07,
+      "loss": 1.2262,
+      "step": 35767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7738325222967924e-07,
+      "loss": 0.7414,
+      "step": 35768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7722338110217507e-07,
+      "loss": 0.7062,
+      "step": 35769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7706358140699432e-07,
+      "loss": 1.1086,
+      "step": 35770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7690385314529823e-07,
+      "loss": 0.823,
+      "step": 35771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.767441963182459e-07,
+      "loss": 0.8085,
+      "step": 35772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7658461092700308e-07,
+      "loss": 0.974,
+      "step": 35773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7642509697272657e-07,
+      "loss": 0.9002,
+      "step": 35774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7626565445657883e-07,
+      "loss": 0.8999,
+      "step": 35775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.761062833797167e-07,
+      "loss": 0.8229,
+      "step": 35776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7594698374330143e-07,
+      "loss": 0.8213,
+      "step": 35777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7578775554849105e-07,
+      "loss": 0.8122,
+      "step": 35778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7562859879644233e-07,
+      "loss": 0.8763,
+      "step": 35779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7546951348831442e-07,
+      "loss": 0.8508,
+      "step": 35780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7531049962526192e-07,
+      "loss": 0.6686,
+      "step": 35781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7515155720844278e-07,
+      "loss": 0.8741,
+      "step": 35782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7499268623901278e-07,
+      "loss": 0.7826,
+      "step": 35783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7483388671812763e-07,
+      "loss": 0.7634,
+      "step": 35784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7467515864693974e-07,
+      "loss": 0.7136,
+      "step": 35785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7451650202660596e-07,
+      "loss": 0.9486,
+      "step": 35786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7435791685827984e-07,
+      "loss": 0.7824,
+      "step": 35787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.741994031431149e-07,
+      "loss": 0.5711,
+      "step": 35788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.740409608822613e-07,
+      "loss": 0.739,
+      "step": 35789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7388259007687368e-07,
+      "loss": 0.9706,
+      "step": 35790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7372429072810226e-07,
+      "loss": 0.7555,
+      "step": 35791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7356606283710053e-07,
+      "loss": 0.7174,
+      "step": 35792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7340790640501647e-07,
+      "loss": 0.8729,
+      "step": 35793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.732498214330014e-07,
+      "loss": 0.5193,
+      "step": 35794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7309180792220547e-07,
+      "loss": 0.7185,
+      "step": 35795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.729338658737767e-07,
+      "loss": 0.5412,
+      "step": 35796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7277599528886303e-07,
+      "loss": 0.8799,
+      "step": 35797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7261819616861574e-07,
+      "loss": 0.8815,
+      "step": 35798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7246046851417952e-07,
+      "loss": 0.974,
+      "step": 35799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.723028123267023e-07,
+      "loss": 0.7897,
+      "step": 35800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.721452276073299e-07,
+      "loss": 0.7205,
+      "step": 35801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7198771435720908e-07,
+      "loss": 0.874,
+      "step": 35802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.718302725774834e-07,
+      "loss": 0.8275,
+      "step": 35803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7167290226930088e-07,
+      "loss": 0.9222,
+      "step": 35804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.715156034338039e-07,
+      "loss": 1.0019,
+      "step": 35805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7135837607213712e-07,
+      "loss": 0.7939,
+      "step": 35806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7120122018544405e-07,
+      "loss": 0.8834,
+      "step": 35807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.710441357748671e-07,
+      "loss": 0.8563,
+      "step": 35808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7088712284154762e-07,
+      "loss": 0.4841,
+      "step": 35809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7073018138663022e-07,
+      "loss": 0.6773,
+      "step": 35810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7057331141125288e-07,
+      "loss": 0.8307,
+      "step": 35811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7041651291655915e-07,
+      "loss": 0.5335,
+      "step": 35812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7025978590368696e-07,
+      "loss": 0.6614,
+      "step": 35813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.7010313037377769e-07,
+      "loss": 0.7399,
+      "step": 35814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6994654632796925e-07,
+      "loss": 0.7542,
+      "step": 35815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6979003376740298e-07,
+      "loss": 0.8864,
+      "step": 35816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6963359269321356e-07,
+      "loss": 0.7144,
+      "step": 35817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6947722310654223e-07,
+      "loss": 0.589,
+      "step": 35818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6932092500852258e-07,
+      "loss": 0.7168,
+      "step": 35819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.691646984002937e-07,
+      "loss": 0.8812,
+      "step": 35820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6900854328298911e-07,
+      "loss": 0.8998,
+      "step": 35821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6885245965774677e-07,
+      "loss": 0.7809,
+      "step": 35822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6869644752570247e-07,
+      "loss": 0.8615,
+      "step": 35823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.685405068879875e-07,
+      "loss": 0.8897,
+      "step": 35824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6838463774573875e-07,
+      "loss": 0.5805,
+      "step": 35825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6822884010008866e-07,
+      "loss": 0.7209,
+      "step": 35826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6807311395217074e-07,
+      "loss": 0.9961,
+      "step": 35827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.679174593031163e-07,
+      "loss": 0.9453,
+      "step": 35828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6776187615405781e-07,
+      "loss": 0.4962,
+      "step": 35829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6760636450612656e-07,
+      "loss": 0.5417,
+      "step": 35830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6745092436045495e-07,
+      "loss": 1.0399,
+      "step": 35831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.67295555718171e-07,
+      "loss": 0.8252,
+      "step": 35832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6714025858040496e-07,
+      "loss": 0.6443,
+      "step": 35833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6698503294828805e-07,
+      "loss": 0.8796,
+      "step": 35834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6682987882294722e-07,
+      "loss": 0.6448,
+      "step": 35835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6667479620551264e-07,
+      "loss": 0.9442,
+      "step": 35836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6651978509711008e-07,
+      "loss": 0.8153,
+      "step": 35837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6636484549886867e-07,
+      "loss": 0.8684,
+      "step": 35838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.662099774119119e-07,
+      "loss": 0.848,
+      "step": 35839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6605518083737005e-07,
+      "loss": 0.7226,
+      "step": 35840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.659004557763666e-07,
+      "loss": 0.9034,
+      "step": 35841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6574580223002623e-07,
+      "loss": 0.618,
+      "step": 35842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6559122019947583e-07,
+      "loss": 0.7308,
+      "step": 35843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.654367096858367e-07,
+      "loss": 0.6088,
+      "step": 35844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.652822706902346e-07,
+      "loss": 0.7579,
+      "step": 35845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6512790321379203e-07,
+      "loss": 1.0749,
+      "step": 35846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6497360725763135e-07,
+      "loss": 0.6786,
+      "step": 35847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.64819382822875e-07,
+      "loss": 0.7245,
+      "step": 35848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6466522991064436e-07,
+      "loss": 0.9408,
+      "step": 35849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6451114852206073e-07,
+      "loss": 0.9585,
+      "step": 35850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.643571386582421e-07,
+      "loss": 0.936,
+      "step": 35851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.64203200320312e-07,
+      "loss": 0.862,
+      "step": 35852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6404933350938844e-07,
+      "loss": 0.8941,
+      "step": 35853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6389553822659054e-07,
+      "loss": 0.8396,
+      "step": 35854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6374181447303738e-07,
+      "loss": 0.7313,
+      "step": 35855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6358816224984476e-07,
+      "loss": 1.0793,
+      "step": 35856
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6343458155813174e-07,
+      "loss": 0.8092,
+      "step": 35857
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6328107239901524e-07,
+      "loss": 0.9229,
+      "step": 35858
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6312763477360993e-07,
+      "loss": 0.912,
+      "step": 35859
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.629742686830338e-07,
+      "loss": 0.9797,
+      "step": 35860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6282097412840147e-07,
+      "loss": 0.6583,
+      "step": 35861
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6266775111082657e-07,
+      "loss": 0.914,
+      "step": 35862
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6251459963142478e-07,
+      "loss": 0.7459,
+      "step": 35863
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6236151969130754e-07,
+      "loss": 0.977,
+      "step": 35864
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6220851129159164e-07,
+      "loss": 0.677,
+      "step": 35865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6205557443338737e-07,
+      "loss": 0.5717,
+      "step": 35866
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6190270911780714e-07,
+      "loss": 0.4292,
+      "step": 35867
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.617499153459645e-07,
+      "loss": 0.8276,
+      "step": 35868
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.615971931189675e-07,
+      "loss": 0.9115,
+      "step": 35869
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.614445424379285e-07,
+      "loss": 0.6171,
+      "step": 35870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.612919633039567e-07,
+      "loss": 0.5131,
+      "step": 35871
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6113945571816335e-07,
+      "loss": 0.8916,
+      "step": 35872
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6098701968165543e-07,
+      "loss": 0.6607,
+      "step": 35873
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.608346551955442e-07,
+      "loss": 0.8354,
+      "step": 35874
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6068236226093437e-07,
+      "loss": 0.9054,
+      "step": 35875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6053014087893504e-07,
+      "loss": 0.6847,
+      "step": 35876
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.603779910506531e-07,
+      "loss": 0.8418,
+      "step": 35877
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6022591277719656e-07,
+      "loss": 0.74,
+      "step": 35878
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.6007390605966898e-07,
+      "loss": 0.8216,
+      "step": 35879
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5992197089917727e-07,
+      "loss": 0.6531,
+      "step": 35880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5977010729682496e-07,
+      "loss": 0.6312,
+      "step": 35881
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5961831525371673e-07,
+      "loss": 0.9629,
+      "step": 35882
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5946659477095728e-07,
+      "loss": 1.1783,
+      "step": 35883
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5931494584964901e-07,
+      "loss": 0.7578,
+      "step": 35884
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5916336849089664e-07,
+      "loss": 0.8811,
+      "step": 35885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5901186269580037e-07,
+      "loss": 0.7688,
+      "step": 35886
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5886042846546268e-07,
+      "loss": 0.7948,
+      "step": 35887
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5870906580098378e-07,
+      "loss": 1.1019,
+      "step": 35888
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5855777470346723e-07,
+      "loss": 0.8635,
+      "step": 35889
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5840655517400995e-07,
+      "loss": 0.6099,
+      "step": 35890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.5825540721371325e-07,
+      "loss": 0.8081,
+      "step": 35891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5810433082367737e-07,
+      "loss": 0.7775,
+      "step": 35892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5795332600499813e-07,
+      "loss": 1.0993,
+      "step": 35893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5780239275877573e-07,
+      "loss": 0.6629,
+      "step": 35894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.576515310861071e-07,
+      "loss": 0.937,
+      "step": 35895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.575007409880902e-07,
+      "loss": 0.7432,
+      "step": 35896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5735002246582086e-07,
+      "loss": 0.7735,
+      "step": 35897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5719937552039488e-07,
+      "loss": 0.7876,
+      "step": 35898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5704880015290912e-07,
+      "loss": 0.6911,
+      "step": 35899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5689829636445498e-07,
+      "loss": 0.9436,
+      "step": 35900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5674786415613152e-07,
+      "loss": 0.8399,
+      "step": 35901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5659750352903013e-07,
+      "loss": 0.7133,
+      "step": 35902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5644721448424438e-07,
+      "loss": 0.7608,
+      "step": 35903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5629699702286783e-07,
+      "loss": 0.8135,
+      "step": 35904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5614685114599292e-07,
+      "loss": 0.9792,
+      "step": 35905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5599677685471103e-07,
+      "loss": 0.9194,
+      "step": 35906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5584677415011351e-07,
+      "loss": 0.8724,
+      "step": 35907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5569684303329168e-07,
+      "loss": 0.8708,
+      "step": 35908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.555469835053358e-07,
+      "loss": 0.9884,
+      "step": 35909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.55397195567335e-07,
+      "loss": 0.9524,
+      "step": 35910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5524747922037953e-07,
+      "loss": 1.0715,
+      "step": 35911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5509783446555738e-07,
+      "loss": 0.8933,
+      "step": 35912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5494826130395657e-07,
+      "loss": 0.8155,
+      "step": 35913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5479875973666626e-07,
+      "loss": 1.2046,
+      "step": 35914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.546493297647711e-07,
+      "loss": 0.796,
+      "step": 35915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5449997138936134e-07,
+      "loss": 0.8937,
+      "step": 35916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5435068461151948e-07,
+      "loss": 0.7869,
+      "step": 35917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.542014694323335e-07,
+      "loss": 0.5963,
+      "step": 35918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5405232585288809e-07,
+      "loss": 0.8458,
+      "step": 35919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5390325387426685e-07,
+      "loss": 0.7111,
+      "step": 35920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5375425349755446e-07,
+      "loss": 0.8648,
+      "step": 35921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.536053247238356e-07,
+      "loss": 0.9168,
+      "step": 35922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5345646755419052e-07,
+      "loss": 0.5732,
+      "step": 35923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5330768198970503e-07,
+      "loss": 0.9505,
+      "step": 35924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5315896803145824e-07,
+      "loss": 0.9407,
+      "step": 35925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5301032568053375e-07,
+      "loss": 0.9069,
+      "step": 35926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.528617549380107e-07,
+      "loss": 0.8242,
+      "step": 35927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5271325580497265e-07,
+      "loss": 0.9973,
+      "step": 35928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5256482828249542e-07,
+      "loss": 0.931,
+      "step": 35929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5241647237166035e-07,
+      "loss": 0.8537,
+      "step": 35930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.522681880735466e-07,
+      "loss": 0.826,
+      "step": 35931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5211997538923218e-07,
+      "loss": 0.9787,
+      "step": 35932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5197183431979513e-07,
+      "loss": 0.9001,
+      "step": 35933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5182376486631233e-07,
+      "loss": 1.1216,
+      "step": 35934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5167576702986076e-07,
+      "loss": 0.6793,
+      "step": 35935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5152784081151617e-07,
+      "loss": 0.6274,
+      "step": 35936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.513799862123555e-07,
+      "loss": 0.5795,
+      "step": 35937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5123220323345343e-07,
+      "loss": 1.0608,
+      "step": 35938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.510844918758836e-07,
+      "loss": 0.5954,
+      "step": 35939
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5093685214072173e-07,
+      "loss": 0.5591,
+      "step": 35940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5078928402904035e-07,
+      "loss": 1.0305,
+      "step": 35941
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5064178754191305e-07,
+      "loss": 0.8228,
+      "step": 35942
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5049436268041118e-07,
+      "loss": 0.9186,
+      "step": 35943
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5034700944560942e-07,
+      "loss": 0.669,
+      "step": 35944
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5019972783857805e-07,
+      "loss": 0.9265,
+      "step": 35945
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.5005251786038843e-07,
+      "loss": 0.7442,
+      "step": 35946
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.499053795121097e-07,
+      "loss": 0.9122,
+      "step": 35947
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4975831279481435e-07,
+      "loss": 1.0196,
+      "step": 35948
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4961131770956928e-07,
+      "loss": 0.6084,
+      "step": 35949
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4946439425744475e-07,
+      "loss": 0.863,
+      "step": 35950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.493175424395088e-07,
+      "loss": 0.6113,
+      "step": 35951
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4917076225682946e-07,
+      "loss": 0.5048,
+      "step": 35952
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.490240537104748e-07,
+      "loss": 0.7471,
+      "step": 35953
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4887741680151058e-07,
+      "loss": 0.6289,
+      "step": 35954
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4873085153100485e-07,
+      "loss": 0.7615,
+      "step": 35955
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4858435790002013e-07,
+      "loss": 0.738,
+      "step": 35956
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4843793590962553e-07,
+      "loss": 0.8765,
+      "step": 35957
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4829158556088464e-07,
+      "loss": 0.6519,
+      "step": 35958
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4814530685486107e-07,
+      "loss": 0.8573,
+      "step": 35959
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.479990997926184e-07,
+      "loss": 0.7702,
+      "step": 35960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4785296437522024e-07,
+      "loss": 1.0662,
+      "step": 35961
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4770690060372905e-07,
+      "loss": 0.7837,
+      "step": 35962
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4756090847920846e-07,
+      "loss": 0.9324,
+      "step": 35963
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4741498800271758e-07,
+      "loss": 0.9113,
+      "step": 35964
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4726913917532004e-07,
+      "loss": 0.9954,
+      "step": 35965
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4712336199807607e-07,
+      "loss": 0.9422,
+      "step": 35966
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4697765647204376e-07,
+      "loss": 0.8057,
+      "step": 35967
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4683202259828444e-07,
+      "loss": 0.6201,
+      "step": 35968
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4668646037785726e-07,
+      "loss": 0.4117,
+      "step": 35969
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4654096981182031e-07,
+      "loss": 0.7982,
+      "step": 35970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4639555090123159e-07,
+      "loss": 0.6243,
+      "step": 35971
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4625020364714805e-07,
+      "loss": 0.9155,
+      "step": 35972
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4610492805062772e-07,
+      "loss": 0.9141,
+      "step": 35973
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.459597241127264e-07,
+      "loss": 0.9247,
+      "step": 35974
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4581459183449997e-07,
+      "loss": 0.6124,
+      "step": 35975
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4566953121700534e-07,
+      "loss": 0.8881,
+      "step": 35976
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4552454226129498e-07,
+      "loss": 0.7445,
+      "step": 35977
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4537962496842583e-07,
+      "loss": 0.5841,
+      "step": 35978
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4523477933944817e-07,
+      "loss": 0.9864,
+      "step": 35979
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4509000537541895e-07,
+      "loss": 0.8311,
+      "step": 35980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4494530307738953e-07,
+      "loss": 0.759,
+      "step": 35981
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4480067244641239e-07,
+      "loss": 0.7598,
+      "step": 35982
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4465611348353893e-07,
+      "loss": 0.943,
+      "step": 35983
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4451162618982052e-07,
+      "loss": 1.0228,
+      "step": 35984
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4436721056630853e-07,
+      "loss": 0.8512,
+      "step": 35985
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4422286661405105e-07,
+      "loss": 0.842,
+      "step": 35986
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4407859433410054e-07,
+      "loss": 0.7587,
+      "step": 35987
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4393439372750616e-07,
+      "loss": 0.6693,
+      "step": 35988
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4379026479531377e-07,
+      "loss": 1.0243,
+      "step": 35989
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4364620753857362e-07,
+      "loss": 0.7547,
+      "step": 35990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4350222195833264e-07,
+      "loss": 0.4902,
+      "step": 35991
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4335830805563666e-07,
+      "loss": 0.664,
+      "step": 35992
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4321446583153486e-07,
+      "loss": 0.4805,
+      "step": 35993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4307069528707196e-07,
+      "loss": 0.9845,
+      "step": 35994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4292699642329376e-07,
+      "loss": 0.971,
+      "step": 35995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4278336924124502e-07,
+      "loss": 0.5811,
+      "step": 35996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4263981374197044e-07,
+      "loss": 1.0933,
+      "step": 35997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4249632992651363e-07,
+      "loss": 0.8541,
+      "step": 35998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.423529177959182e-07,
+      "loss": 0.8536,
+      "step": 35999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4220957735122663e-07,
+      "loss": 0.7782,
+      "step": 36000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4206630859348147e-07,
+      "loss": 1.0296,
+      "step": 36001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4192311152372518e-07,
+      "loss": 0.896,
+      "step": 36002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4177998614299914e-07,
+      "loss": 0.7074,
+      "step": 36003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4163693245234254e-07,
+      "loss": 1.006,
+      "step": 36004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.414939504527979e-07,
+      "loss": 0.7811,
+      "step": 36005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4135104014540434e-07,
+      "loss": 0.8596,
+      "step": 36006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4120820153119997e-07,
+      "loss": 0.5989,
+      "step": 36007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4106543461122613e-07,
+      "loss": 0.8355,
+      "step": 36008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4092273938651758e-07,
+      "loss": 0.6458,
+      "step": 36009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4078011585811348e-07,
+      "loss": 0.772,
+      "step": 36010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4063756402705187e-07,
+      "loss": 0.5396,
+      "step": 36011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4049508389436861e-07,
+      "loss": 0.8613,
+      "step": 36012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.4035267546109954e-07,
+      "loss": 0.5336,
+      "step": 36013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.402103387282816e-07,
+      "loss": 1.0552,
+      "step": 36014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.400680736969484e-07,
+      "loss": 1.04,
+      "step": 36015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.399258803681347e-07,
+      "loss": 0.8649,
+      "step": 36016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.397837587428752e-07,
+      "loss": 0.6366,
+      "step": 36017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.396417088222046e-07,
+      "loss": 0.9376,
+      "step": 36018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3949973060715327e-07,
+      "loss": 0.8608,
+      "step": 36019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3935782409875474e-07,
+      "loss": 0.5766,
+      "step": 36020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.392159892980405e-07,
+      "loss": 0.5692,
+      "step": 36021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3907422620604295e-07,
+      "loss": 0.966,
+      "step": 36022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3893253482379353e-07,
+      "loss": 0.8765,
+      "step": 36023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3879091515232146e-07,
+      "loss": 0.9805,
+      "step": 36024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3864936719265586e-07,
+      "loss": 0.8013,
+      "step": 36025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3850789094582818e-07,
+      "loss": 0.5272,
+      "step": 36026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3836648641286644e-07,
+      "loss": 1.0027,
+      "step": 36027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3822515359479872e-07,
+      "loss": 0.6825,
+      "step": 36028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3808389249265198e-07,
+      "loss": 0.8374,
+      "step": 36029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3794270310745538e-07,
+      "loss": 0.7024,
+      "step": 36030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.378015854402337e-07,
+      "loss": 0.7402,
+      "step": 36031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3766053949201386e-07,
+      "loss": 0.9484,
+      "step": 36032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3751956526382282e-07,
+      "loss": 0.6694,
+      "step": 36033
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.373786627566842e-07,
+      "loss": 0.9657,
+      "step": 36034
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3723783197162167e-07,
+      "loss": 0.8632,
+      "step": 36035
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.370970729096621e-07,
+      "loss": 0.784,
+      "step": 36036
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369563855718281e-07,
+      "loss": 0.9804,
+      "step": 36037
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3681576995914325e-07,
+      "loss": 0.7893,
+      "step": 36038
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3667522607262894e-07,
+      "loss": 0.9433,
+      "step": 36039
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365347539133066e-07,
+      "loss": 0.7416,
+      "step": 36040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3639435348219988e-07,
+      "loss": 0.6837,
+      "step": 36041
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.36254024780329e-07,
+      "loss": 0.7814,
+      "step": 36042
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.361137678087132e-07,
+      "loss": 0.8483,
+      "step": 36043
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.359735825683739e-07,
+      "loss": 0.967,
+      "step": 36044
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3583346906033024e-07,
+      "loss": 0.9845,
+      "step": 36045
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.356934272856003e-07,
+      "loss": 0.8111,
+      "step": 36046
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.355534572452033e-07,
+      "loss": 0.8236,
+      "step": 36047
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.354135589401584e-07,
+      "loss": 0.6412,
+      "step": 36048
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3527373237148033e-07,
+      "loss": 0.9984,
+      "step": 36049
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.351339775401872e-07,
+      "loss": 0.9563,
+      "step": 36050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3499429444729595e-07,
+      "loss": 0.8596,
+      "step": 36051
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3485468309382022e-07,
+      "loss": 0.8594,
+      "step": 36052
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3471514348077697e-07,
+      "loss": 0.7226,
+      "step": 36053
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3457567560918206e-07,
+      "loss": 1.2047,
+      "step": 36054
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.344362794800469e-07,
+      "loss": 0.6625,
+      "step": 36055
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3429695509438845e-07,
+      "loss": 0.7451,
+      "step": 36056
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3415770245321703e-07,
+      "loss": 0.9461,
+      "step": 36057
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3401852155754624e-07,
+      "loss": 0.9439,
+      "step": 36058
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3387941240838864e-07,
+      "loss": 0.7139,
+      "step": 36059
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3374037500675452e-07,
+      "loss": 0.9798,
+      "step": 36060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3360140935365752e-07,
+      "loss": 0.6699,
+      "step": 36061
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3346251545010568e-07,
+      "loss": 0.8184,
+      "step": 36062
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3332369329711047e-07,
+      "loss": 0.7191,
+      "step": 36063
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3318494289568108e-07,
+      "loss": 0.6899,
+      "step": 36064
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3304626424682666e-07,
+      "loss": 0.6839,
+      "step": 36065
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.329076573515553e-07,
+      "loss": 0.8435,
+      "step": 36066
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3276912221087513e-07,
+      "loss": 0.8759,
+      "step": 36067
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3263065882579417e-07,
+      "loss": 0.7599,
+      "step": 36068
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3249226719731835e-07,
+      "loss": 0.7786,
+      "step": 36069
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3235394732645456e-07,
+      "loss": 0.5924,
+      "step": 36070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3221569921420764e-07,
+      "loss": 0.9496,
+      "step": 36071
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.320775228615856e-07,
+      "loss": 1.0078,
+      "step": 36072
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.319394182695899e-07,
+      "loss": 0.9054,
+      "step": 36073
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3180138543922861e-07,
+      "loss": 0.7916,
+      "step": 36074
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3166342437150204e-07,
+      "loss": 0.6879,
+      "step": 36075
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3152553506741495e-07,
+      "loss": 0.8032,
+      "step": 36076
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3138771752797098e-07,
+      "loss": 0.7593,
+      "step": 36077
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3124997175417044e-07,
+      "loss": 0.8297,
+      "step": 36078
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3111229774701695e-07,
+      "loss": 0.8574,
+      "step": 36079
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3097469550750974e-07,
+      "loss": 0.8291,
+      "step": 36080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3083716503665134e-07,
+      "loss": 1.1335,
+      "step": 36081
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3069970633544095e-07,
+      "loss": 1.0999,
+      "step": 36082
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3056231940487663e-07,
+      "loss": 0.6985,
+      "step": 36083
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3042500424596094e-07,
+      "loss": 0.4563,
+      "step": 36084
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3028776085968975e-07,
+      "loss": 0.7413,
+      "step": 36085
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3015058924706226e-07,
+      "loss": 0.8395,
+      "step": 36086
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3001348940907542e-07,
+      "loss": 0.9968,
+      "step": 36087
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2987646134672737e-07,
+      "loss": 0.9847,
+      "step": 36088
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2973950506101284e-07,
+      "loss": 0.8242,
+      "step": 36089
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2960262055292884e-07,
+      "loss": 0.7749,
+      "step": 36090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2946580782347118e-07,
+      "loss": 0.8331,
+      "step": 36091
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2932906687363355e-07,
+      "loss": 0.6279,
+      "step": 36092
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2919239770441184e-07,
+      "loss": 0.734,
+      "step": 36093
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2905580031679853e-07,
+      "loss": 0.8046,
+      "step": 36094
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2891927471178845e-07,
+      "loss": 0.679,
+      "step": 36095
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2878282089037187e-07,
+      "loss": 0.9783,
+      "step": 36096
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2864643885354356e-07,
+      "loss": 0.8329,
+      "step": 36097
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2851012860229606e-07,
+      "loss": 0.7502,
+      "step": 36098
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2837389013761748e-07,
+      "loss": 0.8675,
+      "step": 36099
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2823772346050033e-07,
+      "loss": 0.8821,
+      "step": 36100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2810162857193498e-07,
+      "loss": 0.7352,
+      "step": 36101
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2796560547291058e-07,
+      "loss": 0.9239,
+      "step": 36102
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.278296541644164e-07,
+      "loss": 0.7875,
+      "step": 36103
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.276937746474416e-07,
+      "loss": 0.9757,
+      "step": 36104
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.275579669229743e-07,
+      "loss": 0.8141,
+      "step": 36105
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2742223099200147e-07,
+      "loss": 0.7432,
+      "step": 36106
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2728656685551123e-07,
+      "loss": 0.7684,
+      "step": 36107
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2715097451448832e-07,
+      "loss": 0.6877,
+      "step": 36108
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2701545396991978e-07,
+      "loss": 0.7482,
+      "step": 36109
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2688000522279254e-07,
+      "loss": 0.9398,
+      "step": 36110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2674462827408918e-07,
+      "loss": 0.5772,
+      "step": 36111
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2660932312479556e-07,
+      "loss": 0.9545,
+      "step": 36112
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2647408977589536e-07,
+      "loss": 0.544,
+      "step": 36113
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2633892822837113e-07,
+      "loss": 0.7926,
+      "step": 36114
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.262038384832076e-07,
+      "loss": 0.6978,
+      "step": 36115
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2606882054138624e-07,
+      "loss": 0.7783,
+      "step": 36116
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2593387440388848e-07,
+      "loss": 0.8426,
+      "step": 36117
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2579900007169687e-07,
+      "loss": 0.8282,
+      "step": 36118
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2566419754579064e-07,
+      "loss": 0.8496,
+      "step": 36119
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2552946682715116e-07,
+      "loss": 0.5998,
+      "step": 36120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2539480791675774e-07,
+      "loss": 0.9828,
+      "step": 36121
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2526022081559064e-07,
+      "loss": 0.6233,
+      "step": 36122
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.251257055246269e-07,
+      "loss": 0.6713,
+      "step": 36123
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.249912620448457e-07,
+      "loss": 0.7995,
+      "step": 36124
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2485689037722516e-07,
+      "loss": 0.7219,
+      "step": 36125
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2472259052274116e-07,
+      "loss": 0.7064,
+      "step": 36126
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2458836248237072e-07,
+      "loss": 0.5941,
+      "step": 36127
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2445420625709193e-07,
+      "loss": 0.8096,
+      "step": 36128
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2432012184787845e-07,
+      "loss": 0.6752,
+      "step": 36129
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2418610925570506e-07,
+      "loss": 0.9278,
+      "step": 36130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2405216848154766e-07,
+      "loss": 1.0133,
+      "step": 36131
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2391829952637768e-07,
+      "loss": 0.4924,
+      "step": 36132
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2378450239117214e-07,
+      "loss": 0.9329,
+      "step": 36133
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2365077707690242e-07,
+      "loss": 0.4974,
+      "step": 36134
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2351712358454115e-07,
+      "loss": 0.8174,
+      "step": 36135
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2338354191505974e-07,
+      "loss": 0.8898,
+      "step": 36136
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2325003206942965e-07,
+      "loss": 0.8731,
+      "step": 36137
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.231165940486234e-07,
+      "loss": 0.9852,
+      "step": 36138
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2298322785360917e-07,
+      "loss": 0.6707,
+      "step": 36139
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2284993348535722e-07,
+      "loss": 0.9336,
+      "step": 36140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2271671094483794e-07,
+      "loss": 0.4802,
+      "step": 36141
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2258356023301943e-07,
+      "loss": 0.9004,
+      "step": 36142
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.224504813508709e-07,
+      "loss": 0.692,
+      "step": 36143
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2231747429935714e-07,
+      "loss": 0.6909,
+      "step": 36144
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.221845390794496e-07,
+      "loss": 0.7281,
+      "step": 36145
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2205167569211308e-07,
+      "loss": 0.8228,
+      "step": 36146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2191888413831344e-07,
+      "loss": 1.0667,
+      "step": 36147
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2178616441901657e-07,
+      "loss": 0.7371,
+      "step": 36148
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2165351653518732e-07,
+      "loss": 0.8672,
+      "step": 36149
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.215209404877904e-07,
+      "loss": 0.9887,
+      "step": 36150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2138843627779062e-07,
+      "loss": 0.615,
+      "step": 36151
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2125600390615055e-07,
+      "loss": 0.4298,
+      "step": 36152
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2112364337383497e-07,
+      "loss": 0.685,
+      "step": 36153
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2099135468180423e-07,
+      "loss": 0.8982,
+      "step": 36154
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.20859137831022e-07,
+      "loss": 0.6833,
+      "step": 36155
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.207269928224486e-07,
+      "loss": 0.7056,
+      "step": 36156
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.205949196570466e-07,
+      "loss": 0.9718,
+      "step": 36157
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2046291833577416e-07,
+      "loss": 0.7335,
+      "step": 36158
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2033098885959381e-07,
+      "loss": 0.9623,
+      "step": 36159
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.2019913122946258e-07,
+      "loss": 1.0392,
+      "step": 36160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.200673454463408e-07,
+      "loss": 0.8444,
+      "step": 36161
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1993563151118548e-07,
+      "loss": 0.8103,
+      "step": 36162
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1980398942495698e-07,
+      "loss": 1.0027,
+      "step": 36163
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1967241918861004e-07,
+      "loss": 0.9931,
+      "step": 36164
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1954092080310288e-07,
+      "loss": 0.8748,
+      "step": 36165
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.194094942693913e-07,
+      "loss": 0.9642,
+      "step": 36166
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1927813958843127e-07,
+      "loss": 0.7335,
+      "step": 36167
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1914685676117754e-07,
+      "loss": 0.8759,
+      "step": 36168
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1901564578858493e-07,
+      "loss": 0.7341,
+      "step": 36169
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1888450667160822e-07,
+      "loss": 0.6606,
+      "step": 36170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1875343941119999e-07,
+      "loss": 0.6789,
+      "step": 36171
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1862244400831502e-07,
+      "loss": 0.7183,
+      "step": 36172
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1849152046390366e-07,
+      "loss": 0.7482,
+      "step": 36173
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1836066877892072e-07,
+      "loss": 0.7634,
+      "step": 36174
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1822988895431431e-07,
+      "loss": 0.6206,
+      "step": 36175
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1809918099103922e-07,
+      "loss": 0.8535,
+      "step": 36176
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1796854489004473e-07,
+      "loss": 0.8872,
+      "step": 36177
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1783798065227892e-07,
+      "loss": 0.8477,
+      "step": 36178
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1770748827869327e-07,
+      "loss": 0.6124,
+      "step": 36179
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1757706777023592e-07,
+      "loss": 0.9308,
+      "step": 36180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.174467191278561e-07,
+      "loss": 0.8031,
+      "step": 36181
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1731644235250084e-07,
+      "loss": 0.7323,
+      "step": 36182
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1718623744511715e-07,
+      "loss": 0.6047,
+      "step": 36183
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1705610440665317e-07,
+      "loss": 1.0204,
+      "step": 36184
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1692604323805479e-07,
+      "loss": 0.7671,
+      "step": 36185
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1679605394026794e-07,
+      "loss": 0.8887,
+      "step": 36186
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1666613651423742e-07,
+      "loss": 1.022,
+      "step": 36187
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1653629096090913e-07,
+      "loss": 0.7795,
+      "step": 36188
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1640651728122565e-07,
+      "loss": 0.5869,
+      "step": 36189
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1627681547613178e-07,
+      "loss": 0.6373,
+      "step": 36190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1614718554657012e-07,
+      "loss": 0.7558,
+      "step": 36191
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1601762749348322e-07,
+      "loss": 0.8356,
+      "step": 36192
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1588814131781368e-07,
+      "loss": 0.9442,
+      "step": 36193
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1575872702050406e-07,
+      "loss": 0.8032,
+      "step": 36194
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1562938460249473e-07,
+      "loss": 0.6158,
+      "step": 36195
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1550011406472606e-07,
+      "loss": 0.8674,
+      "step": 36196
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1537091540813838e-07,
+      "loss": 0.6395,
+      "step": 36197
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1524178863366985e-07,
+      "loss": 0.648,
+      "step": 36198
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1511273374226084e-07,
+      "loss": 0.6371,
+      "step": 36199
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1498375073485058e-07,
+      "loss": 0.8282,
+      "step": 36200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1485483961237609e-07,
+      "loss": 0.8263,
+      "step": 36201
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1472600037577553e-07,
+      "loss": 0.9101,
+      "step": 36202
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1459723302598369e-07,
+      "loss": 0.9607,
+      "step": 36203
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1446853756393982e-07,
+      "loss": 0.9738,
+      "step": 36204
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1433991399057764e-07,
+      "loss": 0.8392,
+      "step": 36205
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1421136230683305e-07,
+      "loss": 0.5955,
+      "step": 36206
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1408288251364308e-07,
+      "loss": 0.5606,
+      "step": 36207
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.139544746119392e-07,
+      "loss": 0.9549,
+      "step": 36208
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1382613860265623e-07,
+      "loss": 1.0597,
+      "step": 36209
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1369787448672675e-07,
+      "loss": 0.4513,
+      "step": 36210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1356968226508447e-07,
+      "loss": 0.7372,
+      "step": 36211
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1344156193866086e-07,
+      "loss": 1.0589,
+      "step": 36212
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.133135135083896e-07,
+      "loss": 0.6363,
+      "step": 36213
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1318553697519886e-07,
+      "loss": 0.8785,
+      "step": 36214
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.130576323400212e-07,
+      "loss": 0.9517,
+      "step": 36215
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1292979960378703e-07,
+      "loss": 0.8163,
+      "step": 36216
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1280203876742446e-07,
+      "loss": 0.8198,
+      "step": 36217
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1267434983186387e-07,
+      "loss": 0.7952,
+      "step": 36218
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.125467327980334e-07,
+      "loss": 0.762,
+      "step": 36219
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1241918766686122e-07,
+      "loss": 0.634,
+      "step": 36220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1229171443927434e-07,
+      "loss": 0.5632,
+      "step": 36221
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1216431311620091e-07,
+      "loss": 0.8724,
+      "step": 36222
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1203698369856463e-07,
+      "loss": 0.8665,
+      "step": 36223
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1190972618729479e-07,
+      "loss": 0.8044,
+      "step": 36224
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1178254058331616e-07,
+      "loss": 0.8549,
+      "step": 36225
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1165542688755138e-07,
+      "loss": 0.9546,
+      "step": 36226
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1152838510092857e-07,
+      "loss": 0.529,
+      "step": 36227
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.114014152243681e-07,
+      "loss": 0.8021,
+      "step": 36228
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1127451725879368e-07,
+      "loss": 0.7581,
+      "step": 36229
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1114769120513014e-07,
+      "loss": 0.8736,
+      "step": 36230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1102093706429895e-07,
+      "loss": 0.961,
+      "step": 36231
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1089425483722049e-07,
+      "loss": 0.8476,
+      "step": 36232
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1076764452481847e-07,
+      "loss": 0.6421,
+      "step": 36233
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1064110612801216e-07,
+      "loss": 0.6735,
+      "step": 36234
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1051463964772191e-07,
+      "loss": 0.8281,
+      "step": 36235
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.103882450848659e-07,
+      "loss": 0.7053,
+      "step": 36236
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.102619224403667e-07,
+      "loss": 0.6528,
+      "step": 36237
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1013567171514028e-07,
+      "loss": 0.9445,
+      "step": 36238
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.1000949291010588e-07,
+      "loss": 0.7872,
+      "step": 36239
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0988338602618053e-07,
+      "loss": 0.8155,
+      "step": 36240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.097573510642802e-07,
+      "loss": 0.8588,
+      "step": 36241
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0963138802532414e-07,
+      "loss": 0.619,
+      "step": 36242
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0950549691022605e-07,
+      "loss": 0.8905,
+      "step": 36243
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0937967771990299e-07,
+      "loss": 0.9239,
+      "step": 36244
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0925393045526977e-07,
+      "loss": 0.7051,
+      "step": 36245
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0912825511723902e-07,
+      "loss": 0.7918,
+      "step": 36246
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0900265170672664e-07,
+      "loss": 0.8349,
+      "step": 36247
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0887712022464525e-07,
+      "loss": 0.8085,
+      "step": 36248
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0875166067190746e-07,
+      "loss": 0.7857,
+      "step": 36249
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0862627304942696e-07,
+      "loss": 0.7616,
+      "step": 36250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0850095735811416e-07,
+      "loss": 0.901,
+      "step": 36251
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0837571359888055e-07,
+      "loss": 0.6424,
+      "step": 36252
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.082505417726376e-07,
+      "loss": 0.8546,
+      "step": 36253
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.081254418802935e-07,
+      "loss": 0.499,
+      "step": 36254
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0800041392276194e-07,
+      "loss": 1.1006,
+      "step": 36255
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0787545790094889e-07,
+      "loss": 0.9665,
+      "step": 36256
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0775057381576471e-07,
+      "loss": 0.6138,
+      "step": 36257
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0762576166811645e-07,
+      "loss": 0.7692,
+      "step": 36258
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0750102145891118e-07,
+      "loss": 0.5528,
+      "step": 36259
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0737635318905703e-07,
+      "loss": 0.9494,
+      "step": 36260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.072517568594622e-07,
+      "loss": 0.808,
+      "step": 36261
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.071272324710304e-07,
+      "loss": 0.8319,
+      "step": 36262
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0700278002466758e-07,
+      "loss": 0.5975,
+      "step": 36263
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0687839952127966e-07,
+      "loss": 0.8866,
+      "step": 36264
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0675409096177036e-07,
+      "loss": 0.806,
+      "step": 36265
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0662985434704342e-07,
+      "loss": 1.0456,
+      "step": 36266
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0650568967800368e-07,
+      "loss": 0.712,
+      "step": 36267
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0638159695555261e-07,
+      "loss": 0.8265,
+      "step": 36268
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0625757618059396e-07,
+      "loss": 0.6908,
+      "step": 36269
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.061336273540281e-07,
+      "loss": 0.7172,
+      "step": 36270
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0600975047675766e-07,
+      "loss": 0.597,
+      "step": 36271
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.058859455496819e-07,
+      "loss": 0.7957,
+      "step": 36272
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0576221257370345e-07,
+      "loss": 0.8651,
+      "step": 36273
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0563855154972047e-07,
+      "loss": 0.747,
+      "step": 36274
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0551496247863335e-07,
+      "loss": 0.6629,
+      "step": 36275
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0539144536133917e-07,
+      "loss": 0.7781,
+      "step": 36276
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0526800019873828e-07,
+      "loss": 0.7522,
+      "step": 36277
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0514462699172667e-07,
+      "loss": 0.5874,
+      "step": 36278
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0502132574120249e-07,
+      "loss": 0.8181,
+      "step": 36279
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0489809644806171e-07,
+      "loss": 0.7651,
+      "step": 36280
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0477493911320025e-07,
+      "loss": 0.8239,
+      "step": 36281
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0465185373751519e-07,
+      "loss": 0.9353,
+      "step": 36282
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0452884032190024e-07,
+      "loss": 0.7088,
+      "step": 36283
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0440589886725028e-07,
+      "loss": 0.8233,
+      "step": 36284
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0428302937445899e-07,
+      "loss": 0.9726,
+      "step": 36285
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0416023184442126e-07,
+      "loss": 0.8042,
+      "step": 36286
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0403750627802966e-07,
+      "loss": 0.936,
+      "step": 36287
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0391485267617685e-07,
+      "loss": 0.6561,
+      "step": 36288
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0379227103975209e-07,
+      "loss": 0.6653,
+      "step": 36289
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0366976136964913e-07,
+      "loss": 0.9265,
+      "step": 36290
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0354732366675946e-07,
+      "loss": 1.046,
+      "step": 36291
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0342495793197238e-07,
+      "loss": 0.7986,
+      "step": 36292
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0330266416617829e-07,
+      "loss": 1.0152,
+      "step": 36293
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0318044237026648e-07,
+      "loss": 0.6398,
+      "step": 36294
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0305829254512512e-07,
+      "loss": 0.7291,
+      "step": 36295
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0293621469164238e-07,
+      "loss": 0.5557,
+      "step": 36296
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0281420881070758e-07,
+      "loss": 0.8299,
+      "step": 36297
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0269227490320666e-07,
+      "loss": 0.7465,
+      "step": 36298
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0257041297002557e-07,
+      "loss": 0.9771,
+      "step": 36299
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.024486230120525e-07,
+      "loss": 0.5904,
+      "step": 36300
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0232690503017228e-07,
+      "loss": 0.8742,
+      "step": 36301
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0220525902526868e-07,
+      "loss": 0.8153,
+      "step": 36302
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0208368499822873e-07,
+      "loss": 0.999,
+      "step": 36303
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0196218294993399e-07,
+      "loss": 0.2905,
+      "step": 36304
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0184075288127038e-07,
+      "loss": 0.8976,
+      "step": 36305
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0171939479312054e-07,
+      "loss": 0.9691,
+      "step": 36306
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.015981086863671e-07,
+      "loss": 1.0311,
+      "step": 36307
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0147689456188937e-07,
+      "loss": 0.7759,
+      "step": 36308
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0135575242057105e-07,
+      "loss": 1.1009,
+      "step": 36309
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.012346822632937e-07,
+      "loss": 0.9385,
+      "step": 36310
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0111368409093769e-07,
+      "loss": 0.771,
+      "step": 36311
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0099275790438124e-07,
+      "loss": 0.787,
+      "step": 36312
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0087190370450472e-07,
+      "loss": 0.7955,
+      "step": 36313
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0075112149218746e-07,
+      "loss": 0.6172,
+      "step": 36314
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0063041126830542e-07,
+      "loss": 0.6059,
+      "step": 36315
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.005097730337401e-07,
+      "loss": 0.6687,
+      "step": 36316
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0038920678936748e-07,
+      "loss": 0.7784,
+      "step": 36317
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0026871253606351e-07,
+      "loss": 0.817,
+      "step": 36318
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0014829027470419e-07,
+      "loss": 0.7734,
+      "step": 36319
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0002794000616544e-07,
+      "loss": 0.8958,
+      "step": 36320
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.990766173132438e-08,
+      "loss": 0.5575,
+      "step": 36321
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.978745545105362e-08,
+      "loss": 0.9073,
+      "step": 36322
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.966732116622691e-08,
+      "loss": 0.8042,
+      "step": 36323
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.95472588777202e-08,
+      "loss": 0.9816,
+      "step": 36324
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.942726858640394e-08,
+      "loss": 1.0016,
+      "step": 36325
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.930735029315297e-08,
+      "loss": 0.8519,
+      "step": 36326
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.918750399883881e-08,
+      "loss": 0.9091,
+      "step": 36327
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.906772970433187e-08,
+      "loss": 0.9423,
+      "step": 36328
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.894802741050369e-08,
+      "loss": 1.0709,
+      "step": 36329
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.882839711822468e-08,
+      "loss": 1.1763,
+      "step": 36330
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.870883882836635e-08,
+      "loss": 0.9916,
+      "step": 36331
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.85893525417958e-08,
+      "loss": 0.7394,
+      "step": 36332
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.846993825938234e-08,
+      "loss": 0.7755,
+      "step": 36333
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.835059598199636e-08,
+      "loss": 1.0364,
+      "step": 36334
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.823132571050386e-08,
+      "loss": 0.8636,
+      "step": 36335
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.811212744577303e-08,
+      "loss": 0.8022,
+      "step": 36336
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.799300118867095e-08,
+      "loss": 0.7977,
+      "step": 36337
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.787394694006135e-08,
+      "loss": 0.7556,
+      "step": 36338
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.775496470081359e-08,
+      "loss": 0.5863,
+      "step": 36339
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.763605447179137e-08,
+      "loss": 0.9847,
+      "step": 36340
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.751721625385846e-08,
+      "loss": 0.7815,
+      "step": 36341
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.739845004788084e-08,
+      "loss": 0.8361,
+      "step": 36342
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.727975585472115e-08,
+      "loss": 1.0118,
+      "step": 36343
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.716113367524316e-08,
+      "loss": 0.9986,
+      "step": 36344
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.704258351030838e-08,
+      "loss": 0.6932,
+      "step": 36345
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.692410536077945e-08,
+      "loss": 0.6082,
+      "step": 36346
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.680569922751904e-08,
+      "loss": 0.7533,
+      "step": 36347
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.668736511138643e-08,
+      "loss": 0.5857,
+      "step": 36348
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.656910301324429e-08,
+      "loss": 0.772,
+      "step": 36349
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.645091293395082e-08,
+      "loss": 0.5422,
+      "step": 36350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.63327948743642e-08,
+      "loss": 0.9341,
+      "step": 36351
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.62147488353471e-08,
+      "loss": 0.9369,
+      "step": 36352
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.60967748177566e-08,
+      "loss": 0.8526,
+      "step": 36353
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.597887282244866e-08,
+      "loss": 0.6412,
+      "step": 36354
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.586104285028375e-08,
+      "loss": 0.8131,
+      "step": 36355
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.57432849021156e-08,
+      "loss": 0.938,
+      "step": 36356
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.562559897880353e-08,
+      "loss": 1.0857,
+      "step": 36357
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.55079850812013e-08,
+      "loss": 0.8889,
+      "step": 36358
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.539044321016489e-08,
+      "loss": 0.7373,
+      "step": 36359
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.527297336654917e-08,
+      "loss": 0.864,
+      "step": 36360
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.515557555120791e-08,
+      "loss": 0.9038,
+      "step": 36361
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.503824976499598e-08,
+      "loss": 0.87,
+      "step": 36362
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.492099600876492e-08,
+      "loss": 0.9948,
+      "step": 36363
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.480381428336849e-08,
+      "loss": 1.1027,
+      "step": 36364
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.468670458965823e-08,
+      "loss": 1.1265,
+      "step": 36365
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.45696669284879e-08,
+      "loss": 0.6754,
+      "step": 36366
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.445270130070572e-08,
+      "loss": 0.8945,
+      "step": 36367
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.433580770716432e-08,
+      "loss": 0.8198,
+      "step": 36368
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.421898614871194e-08,
+      "loss": 0.8417,
+      "step": 36369
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.410223662619899e-08,
+      "loss": 0.8665,
+      "step": 36370
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.39855591404748e-08,
+      "loss": 0.8174,
+      "step": 36371
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.386895369238758e-08,
+      "loss": 0.6886,
+      "step": 36372
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.375242028278664e-08,
+      "loss": 0.6184,
+      "step": 36373
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.363595891251686e-08,
+      "loss": 0.8756,
+      "step": 36374
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.351956958242648e-08,
+      "loss": 0.8037,
+      "step": 36375
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.340325229336256e-08,
+      "loss": 0.9295,
+      "step": 36376
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.328700704617e-08,
+      "loss": 0.5586,
+      "step": 36377
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.317083384169478e-08,
+      "loss": 1.0818,
+      "step": 36378
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.305473268078068e-08,
+      "loss": 0.7123,
+      "step": 36379
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.293870356427259e-08,
+      "loss": 0.9229,
+      "step": 36380
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.282274649301426e-08,
+      "loss": 0.6578,
+      "step": 36381
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.270686146785057e-08,
+      "loss": 0.9874,
+      "step": 36382
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.259104848962197e-08,
+      "loss": 0.8139,
+      "step": 36383
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.247530755917112e-08,
+      "loss": 0.79,
+      "step": 36384
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.235963867733954e-08,
+      "loss": 0.7826,
+      "step": 36385
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.224404184496994e-08,
+      "loss": 0.8685,
+      "step": 36386
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.21285170629016e-08,
+      "loss": 0.9802,
+      "step": 36387
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.201306433197499e-08,
+      "loss": 0.785,
+      "step": 36388
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.189768365302943e-08,
+      "loss": 0.7427,
+      "step": 36389
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.178237502690423e-08,
+      "loss": 0.6982,
+      "step": 36390
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.166713845443876e-08,
+      "loss": 0.7181,
+      "step": 36391
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.155197393646898e-08,
+      "loss": 0.983,
+      "step": 36392
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.143688147383423e-08,
+      "loss": 0.8998,
+      "step": 36393
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.13218610673694e-08,
+      "loss": 0.7361,
+      "step": 36394
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.120691271791381e-08,
+      "loss": 0.8658,
+      "step": 36395
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.109203642630127e-08,
+      "loss": 0.6986,
+      "step": 36396
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.097723219336885e-08,
+      "loss": 0.647,
+      "step": 36397
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.086250001994922e-08,
+      "loss": 0.9506,
+      "step": 36398
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.07478399068773e-08,
+      "loss": 0.9143,
+      "step": 36399
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.063325185498906e-08,
+      "loss": 0.7856,
+      "step": 36400
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.051873586511495e-08,
+      "loss": 0.9866,
+      "step": 36401
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.040429193808987e-08,
+      "loss": 0.7838,
+      "step": 36402
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.028992007474535e-08,
+      "loss": 0.6599,
+      "step": 36403
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.017562027591187e-08,
+      "loss": 0.6195,
+      "step": 36404
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.006139254242319e-08,
+      "loss": 0.838,
+      "step": 36405
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.994723687510643e-08,
+      "loss": 0.5687,
+      "step": 36406
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.983315327479647e-08,
+      "loss": 1.0073,
+      "step": 36407
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.971914174231933e-08,
+      "loss": 0.6177,
+      "step": 36408
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.960520227850433e-08,
+      "loss": 0.89,
+      "step": 36409
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.949133488418193e-08,
+      "loss": 0.7881,
+      "step": 36410
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.937753956017814e-08,
+      "loss": 0.8458,
+      "step": 36411
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.926381630732117e-08,
+      "loss": 0.989,
+      "step": 36412
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.915016512643926e-08,
+      "loss": 0.9242,
+      "step": 36413
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.90365860183573e-08,
+      "loss": 0.6893,
+      "step": 36414
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.89230789839024e-08,
+      "loss": 0.809,
+      "step": 36415
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.880964402389835e-08,
+      "loss": 0.8814,
+      "step": 36416
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.869628113917228e-08,
+      "loss": 0.9628,
+      "step": 36417
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.858299033054685e-08,
+      "loss": 1.0311,
+      "step": 36418
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.846977159884696e-08,
+      "loss": 0.8805,
+      "step": 36419
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.835662494489638e-08,
+      "loss": 0.9206,
+      "step": 36420
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.82435503695156e-08,
+      "loss": 0.8223,
+      "step": 36421
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.813054787352948e-08,
+      "loss": 0.8995,
+      "step": 36422
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.801761745775738e-08,
+      "loss": 0.8127,
+      "step": 36423
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.79047591230231e-08,
+      "loss": 0.6768,
+      "step": 36424
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.779197287014485e-08,
+      "loss": 0.6896,
+      "step": 36425
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.767925869994532e-08,
+      "loss": 0.8332,
+      "step": 36426
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.756661661324384e-08,
+      "loss": 0.5955,
+      "step": 36427
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.745404661085755e-08,
+      "loss": 0.7227,
+      "step": 36428
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.734154869360578e-08,
+      "loss": 0.9106,
+      "step": 36429
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.722912286230678e-08,
+      "loss": 0.8141,
+      "step": 36430
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.711676911777877e-08,
+      "loss": 1.0256,
+      "step": 36431
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.70044874608389e-08,
+      "loss": 0.702,
+      "step": 36432
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.689227789230315e-08,
+      "loss": 1.0647,
+      "step": 36433
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.678014041298643e-08,
+      "loss": 0.8083,
+      "step": 36434
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.6668075023707e-08,
+      "loss": 0.6435,
+      "step": 36435
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.655608172527752e-08,
+      "loss": 0.4927,
+      "step": 36436
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.64441605185129e-08,
+      "loss": 0.7573,
+      "step": 36437
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.633231140422694e-08,
+      "loss": 0.6133,
+      "step": 36438
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.622053438323342e-08,
+      "loss": 0.9604,
+      "step": 36439
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.610882945634613e-08,
+      "loss": 0.7265,
+      "step": 36440
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.599719662437556e-08,
+      "loss": 0.631,
+      "step": 36441
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.588563588813325e-08,
+      "loss": 0.7117,
+      "step": 36442
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.577414724843303e-08,
+      "loss": 0.5743,
+      "step": 36443
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.566273070608311e-08,
+      "loss": 0.6749,
+      "step": 36444
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.555138626189619e-08,
+      "loss": 0.5461,
+      "step": 36445
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.544011391667939e-08,
+      "loss": 0.8421,
+      "step": 36446
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.53289136712443e-08,
+      "loss": 0.7623,
+      "step": 36447
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.521778552639693e-08,
+      "loss": 0.8004,
+      "step": 36448
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.510672948294663e-08,
+      "loss": 0.791,
+      "step": 36449
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.499574554170276e-08,
+      "loss": 0.6775,
+      "step": 36450
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.488483370347023e-08,
+      "loss": 0.9001,
+      "step": 36451
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.477399396905617e-08,
+      "loss": 0.6224,
+      "step": 36452
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.46632263392666e-08,
+      "loss": 0.6142,
+      "step": 36453
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.455253081490644e-08,
+      "loss": 0.5746,
+      "step": 36454
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.444190739678171e-08,
+      "loss": 0.9237,
+      "step": 36455
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.433135608569731e-08,
+      "loss": 0.8494,
+      "step": 36456
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.422087688245595e-08,
+      "loss": 0.7957,
+      "step": 36457
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.411046978786142e-08,
+      "loss": 1.0419,
+      "step": 36458
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.40001348027164e-08,
+      "loss": 0.8033,
+      "step": 36459
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.388987192782472e-08,
+      "loss": 0.5224,
+      "step": 36460
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.377968116398572e-08,
+      "loss": 0.8416,
+      "step": 36461
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.36695625120021e-08,
+      "loss": 1.0423,
+      "step": 36462
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.355951597267542e-08,
+      "loss": 0.7345,
+      "step": 36463
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.344954154680507e-08,
+      "loss": 0.7158,
+      "step": 36464
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.333963923519039e-08,
+      "loss": 0.7505,
+      "step": 36465
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.322980903863076e-08,
+      "loss": 0.8425,
+      "step": 36466
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.312005095792552e-08,
+      "loss": 0.6394,
+      "step": 36467
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.301036499387293e-08,
+      "loss": 0.5162,
+      "step": 36468
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.290075114727014e-08,
+      "loss": 0.7454,
+      "step": 36469
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.279120941891427e-08,
+      "loss": 0.8382,
+      "step": 36470
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.268173980960138e-08,
+      "loss": 0.5922,
+      "step": 36471
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.257234232012967e-08,
+      "loss": 0.5998,
+      "step": 36472
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.246301695129077e-08,
+      "loss": 0.6989,
+      "step": 36473
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.235376370388403e-08,
+      "loss": 1.1078,
+      "step": 36474
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.224458257870105e-08,
+      "loss": 0.7462,
+      "step": 36475
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.213547357653784e-08,
+      "loss": 1.0373,
+      "step": 36476
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.20264366981871e-08,
+      "loss": 0.5443,
+      "step": 36477
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.191747194444045e-08,
+      "loss": 0.7563,
+      "step": 36478
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.180857931609165e-08,
+      "loss": 0.701,
+      "step": 36479
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.169975881393122e-08,
+      "loss": 0.9049,
+      "step": 36480
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.159101043875294e-08,
+      "loss": 0.7236,
+      "step": 36481
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.148233419134621e-08,
+      "loss": 0.6518,
+      "step": 36482
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.137373007250038e-08,
+      "loss": 0.8105,
+      "step": 36483
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.126519808300703e-08,
+      "loss": 1.0011,
+      "step": 36484
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.115673822365444e-08,
+      "loss": 0.7351,
+      "step": 36485
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.104835049523086e-08,
+      "loss": 0.7828,
+      "step": 36486
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.094003489852675e-08,
+      "loss": 0.953,
+      "step": 36487
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.083179143432706e-08,
+      "loss": 0.9186,
+      "step": 36488
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.072362010342005e-08,
+      "loss": 0.9141,
+      "step": 36489
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.061552090659285e-08,
+      "loss": 0.9879,
+      "step": 36490
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.050749384463041e-08,
+      "loss": 0.9005,
+      "step": 36491
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.039953891831875e-08,
+      "loss": 0.8022,
+      "step": 36492
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.029165612844392e-08,
+      "loss": 0.3102,
+      "step": 36493
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.018384547578973e-08,
+      "loss": 0.7741,
+      "step": 36494
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 8.007610696114e-08,
+      "loss": 0.6135,
+      "step": 36495
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.996844058527852e-08,
+      "loss": 0.8063,
+      "step": 36496
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.986084634898806e-08,
+      "loss": 0.8439,
+      "step": 36497
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.975332425305126e-08,
+      "loss": 0.7402,
+      "step": 36498
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.964587429824977e-08,
+      "loss": 0.6404,
+      "step": 36499
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.953849648536516e-08,
+      "loss": 0.5704,
+      "step": 36500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.943119081517902e-08,
+      "loss": 0.7796,
+      "step": 36501
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.932395728846965e-08,
+      "loss": 0.8175,
+      "step": 36502
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.921679590601861e-08,
+      "loss": 0.7821,
+      "step": 36503
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.910970666860418e-08,
+      "loss": 0.961,
+      "step": 36504
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.900268957700686e-08,
+      "loss": 0.7113,
+      "step": 36505
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.889574463200267e-08,
+      "loss": 0.6188,
+      "step": 36506
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.878887183437101e-08,
+      "loss": 0.5245,
+      "step": 36507
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.868207118488791e-08,
+      "loss": 0.6736,
+      "step": 36508
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.857534268433054e-08,
+      "loss": 0.7026,
+      "step": 36509
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.846868633347492e-08,
+      "loss": 0.864,
+      "step": 36510
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.836210213309714e-08,
+      "loss": 0.4684,
+      "step": 36511
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.82555900839721e-08,
+      "loss": 1.037,
+      "step": 36512
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.814915018687364e-08,
+      "loss": 1.1282,
+      "step": 36513
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.804278244257668e-08,
+      "loss": 0.8064,
+      "step": 36514
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.793648685185395e-08,
+      "loss": 0.9025,
+      "step": 36515
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.783026341547927e-08,
+      "loss": 0.7879,
+      "step": 36516
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.772411213422536e-08,
+      "loss": 0.8452,
+      "step": 36517
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.761803300886273e-08,
+      "loss": 0.7756,
+      "step": 36518
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.751202604016405e-08,
+      "loss": 0.756,
+      "step": 36519
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.740609122889987e-08,
+      "loss": 0.745,
+      "step": 36520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.730022857583952e-08,
+      "loss": 0.7922,
+      "step": 36521
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.719443808175464e-08,
+      "loss": 0.6832,
+      "step": 36522
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.70887197474135e-08,
+      "loss": 0.9278,
+      "step": 36523
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.698307357358548e-08,
+      "loss": 0.8021,
+      "step": 36524
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.687749956103774e-08,
+      "loss": 0.9416,
+      "step": 36525
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.677199771053967e-08,
+      "loss": 0.712,
+      "step": 36526
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.666656802285622e-08,
+      "loss": 1.0695,
+      "step": 36527
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.656121049875676e-08,
+      "loss": 0.8243,
+      "step": 36528
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.645592513900513e-08,
+      "loss": 0.4874,
+      "step": 36529
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.635071194436961e-08,
+      "loss": 0.7997,
+      "step": 36530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.62455709156118e-08,
+      "loss": 0.6116,
+      "step": 36531
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.614050205349999e-08,
+      "loss": 0.4158,
+      "step": 36532
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.603550535879467e-08,
+      "loss": 0.8104,
+      "step": 36533
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.59305808322619e-08,
+      "loss": 0.79,
+      "step": 36534
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.582572847466441e-08,
+      "loss": 0.8169,
+      "step": 36535
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.57209482867638e-08,
+      "loss": 0.8283,
+      "step": 36536
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.56162402693228e-08,
+      "loss": 0.8113,
+      "step": 36537
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.551160442310302e-08,
+      "loss": 0.9942,
+      "step": 36538
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.540704074886385e-08,
+      "loss": 0.6927,
+      "step": 36539
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.530254924736691e-08,
+      "loss": 0.9505,
+      "step": 36540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.51981299193727e-08,
+      "loss": 0.803,
+      "step": 36541
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.50937827656395e-08,
+      "loss": 0.8955,
+      "step": 36542
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.49895077869267e-08,
+      "loss": 0.9602,
+      "step": 36543
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.48853049839915e-08,
+      "loss": 0.8262,
+      "step": 36544
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.478117435759324e-08,
+      "loss": 0.9501,
+      "step": 36545
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.467711590848914e-08,
+      "loss": 0.9431,
+      "step": 36546
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.457312963743412e-08,
+      "loss": 0.9271,
+      "step": 36547
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.446921554518539e-08,
+      "loss": 0.9731,
+      "step": 36548
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.436537363249896e-08,
+      "loss": 0.8121,
+      "step": 36549
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.426160390013092e-08,
+      "loss": 0.8928,
+      "step": 36550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.415790634883291e-08,
+      "loss": 0.9999,
+      "step": 36551
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.405428097936206e-08,
+      "loss": 0.8186,
+      "step": 36552
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.395072779247004e-08,
+      "loss": 0.8875,
+      "step": 36553
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.384724678891064e-08,
+      "loss": 0.7816,
+      "step": 36554
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.374383796943663e-08,
+      "loss": 0.5862,
+      "step": 36555
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.364050133479961e-08,
+      "loss": 0.7858,
+      "step": 36556
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.353723688575232e-08,
+      "loss": 0.9035,
+      "step": 36557
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.343404462304194e-08,
+      "loss": 0.8925,
+      "step": 36558
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.33309245474234e-08,
+      "loss": 0.9799,
+      "step": 36559
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.322787665964393e-08,
+      "loss": 0.6367,
+      "step": 36560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.312490096045399e-08,
+      "loss": 0.5497,
+      "step": 36561
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.3021997450603e-08,
+      "loss": 0.5257,
+      "step": 36562
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.291916613083705e-08,
+      "loss": 0.9349,
+      "step": 36563
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.281640700190662e-08,
+      "loss": 1.0512,
+      "step": 36564
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.271372006455557e-08,
+      "loss": 0.8164,
+      "step": 36565
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.261110531953442e-08,
+      "loss": 1.0567,
+      "step": 36566
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.25085627675881e-08,
+      "loss": 0.9398,
+      "step": 36567
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.240609240946161e-08,
+      "loss": 0.941,
+      "step": 36568
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.23036942458999e-08,
+      "loss": 0.7943,
+      "step": 36569
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.22013682776479e-08,
+      "loss": 0.9805,
+      "step": 36570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.20991145054506e-08,
+      "loss": 0.8096,
+      "step": 36571
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.199693293005072e-08,
+      "loss": 0.8455,
+      "step": 36572
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.189482355219101e-08,
+      "loss": 0.8506,
+      "step": 36573
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.179278637261533e-08,
+      "loss": 0.6568,
+      "step": 36574
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.16908213920653e-08,
+      "loss": 0.8264,
+      "step": 36575
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.158892861128253e-08,
+      "loss": 0.9342,
+      "step": 36576
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.148710803100645e-08,
+      "loss": 0.6475,
+      "step": 36577
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.138535965197868e-08,
+      "loss": 0.9839,
+      "step": 36578
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.128368347493974e-08,
+      "loss": 0.7891,
+      "step": 36579
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.118207950062905e-08,
+      "loss": 0.7995,
+      "step": 36580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.108054772978379e-08,
+      "loss": 0.4367,
+      "step": 36581
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.097908816314337e-08,
+      "loss": 0.9908,
+      "step": 36582
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.087770080144496e-08,
+      "loss": 0.8892,
+      "step": 36583
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.077638564542799e-08,
+      "loss": 1.0195,
+      "step": 36584
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.067514269582743e-08,
+      "loss": 0.8631,
+      "step": 36585
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.057397195337934e-08,
+      "loss": 1.0954,
+      "step": 36586
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.047287341882092e-08,
+      "loss": 0.7081,
+      "step": 36587
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.037184709288491e-08,
+      "loss": 0.7755,
+      "step": 36588
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.027089297630852e-08,
+      "loss": 0.6768,
+      "step": 36589
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.017001106982446e-08,
+      "loss": 0.7355,
+      "step": 36590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 7.006920137416773e-08,
+      "loss": 0.8012,
+      "step": 36591
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.996846389006883e-08,
+      "loss": 1.1932,
+      "step": 36592
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.986779861826276e-08,
+      "loss": 0.8774,
+      "step": 36593
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.976720555948114e-08,
+      "loss": 0.6994,
+      "step": 36594
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.966668471445448e-08,
+      "loss": 0.8959,
+      "step": 36595
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.956623608391444e-08,
+      "loss": 0.8724,
+      "step": 36596
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.946585966859265e-08,
+      "loss": 0.9217,
+      "step": 36597
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.936555546921742e-08,
+      "loss": 0.6439,
+      "step": 36598
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.926532348651816e-08,
+      "loss": 0.7638,
+      "step": 36599
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.916516372122429e-08,
+      "loss": 1.0783,
+      "step": 36600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.906507617406411e-08,
+      "loss": 0.6724,
+      "step": 36601
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.896506084576482e-08,
+      "loss": 0.7733,
+      "step": 36602
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.886511773705585e-08,
+      "loss": 0.9109,
+      "step": 36603
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.876524684866215e-08,
+      "loss": 0.9071,
+      "step": 36604
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.866544818130982e-08,
+      "loss": 0.8235,
+      "step": 36605
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.856572173572496e-08,
+      "loss": 0.7271,
+      "step": 36606
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.846606751263473e-08,
+      "loss": 0.7404,
+      "step": 36607
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.836648551276082e-08,
+      "loss": 0.65,
+      "step": 36608
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.826697573682816e-08,
+      "loss": 0.832,
+      "step": 36609
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.816753818556177e-08,
+      "loss": 0.5727,
+      "step": 36610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.806817285968436e-08,
+      "loss": 1.0243,
+      "step": 36611
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.796887975991762e-08,
+      "loss": 1.0107,
+      "step": 36612
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.786965888698315e-08,
+      "loss": 0.806,
+      "step": 36613
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.777051024160486e-08,
+      "loss": 0.7078,
+      "step": 36614
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.767143382450214e-08,
+      "loss": 0.8783,
+      "step": 36615
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.757242963639555e-08,
+      "loss": 0.9351,
+      "step": 36616
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.747349767800559e-08,
+      "loss": 0.8131,
+      "step": 36617
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.737463795005062e-08,
+      "loss": 0.9652,
+      "step": 36618
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.727585045325003e-08,
+      "loss": 0.8689,
+      "step": 36619
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.717713518832325e-08,
+      "loss": 0.8248,
+      "step": 36620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.70784921559875e-08,
+      "loss": 1.0596,
+      "step": 36621
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.697992135695996e-08,
+      "loss": 0.7834,
+      "step": 36622
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.688142279195676e-08,
+      "loss": 0.7972,
+      "step": 36623
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.678299646169506e-08,
+      "loss": 0.6427,
+      "step": 36624
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.668464236689099e-08,
+      "loss": 0.8174,
+      "step": 36625
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.65863605082584e-08,
+      "loss": 1.0039,
+      "step": 36626
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.648815088651339e-08,
+      "loss": 0.9434,
+      "step": 36627
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.639001350236985e-08,
+      "loss": 0.5619,
+      "step": 36628
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.629194835654162e-08,
+      "loss": 0.9162,
+      "step": 36629
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.619395544974039e-08,
+      "loss": 0.9045,
+      "step": 36630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.609603478268e-08,
+      "loss": 0.928,
+      "step": 36631
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.599818635607324e-08,
+      "loss": 0.6748,
+      "step": 36632
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.590041017062954e-08,
+      "loss": 0.7312,
+      "step": 36633
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.580270622706275e-08,
+      "loss": 0.9505,
+      "step": 36634
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.57050745260801e-08,
+      "loss": 0.6259,
+      "step": 36635
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.560751506839435e-08,
+      "loss": 0.8886,
+      "step": 36636
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.551002785471383e-08,
+      "loss": 0.9988,
+      "step": 36637
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.541261288574796e-08,
+      "loss": 0.7813,
+      "step": 36638
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.531527016220508e-08,
+      "loss": 0.7889,
+      "step": 36639
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.521799968479126e-08,
+      "loss": 1.0876,
+      "step": 36640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.512080145421706e-08,
+      "loss": 0.6644,
+      "step": 36641
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.502367547118638e-08,
+      "loss": 0.8469,
+      "step": 36642
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.492662173640752e-08,
+      "loss": 0.7606,
+      "step": 36643
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.482964025058436e-08,
+      "loss": 0.7679,
+      "step": 36644
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.473273101442412e-08,
+      "loss": 0.939,
+      "step": 36645
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.46358940286318e-08,
+      "loss": 0.4843,
+      "step": 36646
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.453912929391015e-08,
+      "loss": 0.5618,
+      "step": 36647
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.444243681096197e-08,
+      "loss": 0.8767,
+      "step": 36648
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.434581658049221e-08,
+      "loss": 0.6194,
+      "step": 36649
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.424926860320369e-08,
+      "loss": 1.0316,
+      "step": 36650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.415279287979691e-08,
+      "loss": 0.8166,
+      "step": 36651
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.405638941097469e-08,
+      "loss": 0.7882,
+      "step": 36652
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.396005819743867e-08,
+      "loss": 0.6601,
+      "step": 36653
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.386379923988718e-08,
+      "loss": 1.0654,
+      "step": 36654
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.376761253902186e-08,
+      "loss": 0.8034,
+      "step": 36655
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.367149809554219e-08,
+      "loss": 0.7556,
+      "step": 36656
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.357545591014647e-08,
+      "loss": 0.6673,
+      "step": 36657
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.347948598353415e-08,
+      "loss": 0.7659,
+      "step": 36658
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.338358831640134e-08,
+      "loss": 0.8229,
+      "step": 36659
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.32877629094475e-08,
+      "loss": 0.9933,
+      "step": 36660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.319200976336759e-08,
+      "loss": 1.3525,
+      "step": 36661
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.309632887885886e-08,
+      "loss": 0.8355,
+      "step": 36662
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.300072025661852e-08,
+      "loss": 0.7986,
+      "step": 36663
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.290518389733935e-08,
+      "loss": 1.049,
+      "step": 36664
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.280971980171746e-08,
+      "loss": 0.862,
+      "step": 36665
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.271432797044674e-08,
+      "loss": 0.5648,
+      "step": 36666
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.261900840421997e-08,
+      "loss": 0.81,
+      "step": 36667
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.252376110373215e-08,
+      "loss": 0.7397,
+      "step": 36668
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.242858606967495e-08,
+      "loss": 0.5709,
+      "step": 36669
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.233348330274114e-08,
+      "loss": 0.6997,
+      "step": 36670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.223845280362128e-08,
+      "loss": 0.732,
+      "step": 36671
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.214349457300706e-08,
+      "loss": 0.8066,
+      "step": 36672
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.204860861158902e-08,
+      "loss": 0.9529,
+      "step": 36673
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.19537949200566e-08,
+      "loss": 0.4687,
+      "step": 36674
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.185905349910038e-08,
+      "loss": 0.9913,
+      "step": 36675
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.176438434940867e-08,
+      "loss": 0.7099,
+      "step": 36676
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.166978747166985e-08,
+      "loss": 1.0622,
+      "step": 36677
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.157526286657223e-08,
+      "loss": 0.9097,
+      "step": 36678
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.148081053480193e-08,
+      "loss": 0.8516,
+      "step": 36679
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.138643047704729e-08,
+      "loss": 0.6187,
+      "step": 36680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.129212269399443e-08,
+      "loss": 0.7015,
+      "step": 36681
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.119788718632946e-08,
+      "loss": 0.6859,
+      "step": 36682
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.11037239547363e-08,
+      "loss": 0.8226,
+      "step": 36683
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.100963299990104e-08,
+      "loss": 0.8487,
+      "step": 36684
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.091561432250647e-08,
+      "loss": 0.6573,
+      "step": 36685
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.082166792323874e-08,
+      "loss": 0.7824,
+      "step": 36686
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.072779380277838e-08,
+      "loss": 0.7411,
+      "step": 36687
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.06339919618093e-08,
+      "loss": 0.9083,
+      "step": 36688
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.05402624010143e-08,
+      "loss": 0.9409,
+      "step": 36689
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.044660512107392e-08,
+      "loss": 0.8968,
+      "step": 36690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.035302012266986e-08,
+      "loss": 0.9763,
+      "step": 36691
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.025950740648157e-08,
+      "loss": 0.6706,
+      "step": 36692
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.016606697319072e-08,
+      "loss": 0.7695,
+      "step": 36693
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.007269882347566e-08,
+      "loss": 1.1098,
+      "step": 36694
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.997940295801585e-08,
+      "loss": 0.7237,
+      "step": 36695
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.988617937748964e-08,
+      "loss": 1.1273,
+      "step": 36696
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.979302808257537e-08,
+      "loss": 1.0285,
+      "step": 36697
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.969994907394916e-08,
+      "loss": 0.7809,
+      "step": 36698
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.960694235228937e-08,
+      "loss": 0.8246,
+      "step": 36699
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9514007918271e-08,
+      "loss": 0.8882,
+      "step": 36700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9421145772571296e-08,
+      "loss": 0.8602,
+      "step": 36701
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.932835591586528e-08,
+      "loss": 0.8914,
+      "step": 36702
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.923563834882684e-08,
+      "loss": 0.8708,
+      "step": 36703
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.9142993072131006e-08,
+      "loss": 0.658,
+      "step": 36704
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.905042008645057e-08,
+      "loss": 0.6041,
+      "step": 36705
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.895791939245943e-08,
+      "loss": 0.8428,
+      "step": 36706
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8865490990831496e-08,
+      "loss": 0.7515,
+      "step": 36707
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8773134882236235e-08,
+      "loss": 0.513,
+      "step": 36708
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.868085106734755e-08,
+      "loss": 0.9072,
+      "step": 36709
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.858863954683491e-08,
+      "loss": 1.0055,
+      "step": 36710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8496500321369995e-08,
+      "loss": 1.1328,
+      "step": 36711
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8404433391622276e-08,
+      "loss": 0.8835,
+      "step": 36712
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.831243875826231e-08,
+      "loss": 0.7249,
+      "step": 36713
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8220516421957364e-08,
+      "loss": 0.5947,
+      "step": 36714
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.8128666383377994e-08,
+      "loss": 0.6851,
+      "step": 36715
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.803688864319035e-08,
+      "loss": 0.7843,
+      "step": 36716
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7945183202061664e-08,
+      "loss": 0.8733,
+      "step": 36717
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.785355006066029e-08,
+      "loss": 0.6184,
+      "step": 36718
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.776198921965237e-08,
+      "loss": 0.839,
+      "step": 36719
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7670500679702925e-08,
+      "loss": 0.7064,
+      "step": 36720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.757908444147808e-08,
+      "loss": 1.0367,
+      "step": 36721
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7487740505641766e-08,
+      "loss": 0.3843,
+      "step": 36722
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.739646887285788e-08,
+      "loss": 0.5807,
+      "step": 36723
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7305269543792564e-08,
+      "loss": 0.8243,
+      "step": 36724
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.7214142519106396e-08,
+      "loss": 0.9727,
+      "step": 36725
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.712308779946329e-08,
+      "loss": 0.6458,
+      "step": 36726
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.703210538552606e-08,
+      "loss": 0.946,
+      "step": 36727
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6941195277954165e-08,
+      "loss": 0.7655,
+      "step": 36728
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.685035747741152e-08,
+      "loss": 1.0207,
+      "step": 36729
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6759591984556494e-08,
+      "loss": 0.8727,
+      "step": 36730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.666889880004967e-08,
+      "loss": 0.9014,
+      "step": 36731
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.657827792455051e-08,
+      "loss": 0.6456,
+      "step": 36732
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.6487729358719604e-08,
+      "loss": 0.6275,
+      "step": 36733
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.639725310321309e-08,
+      "loss": 0.9037,
+      "step": 36734
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.630684915868934e-08,
+      "loss": 0.8973,
+      "step": 36735
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.621651752580781e-08,
+      "loss": 1.0692,
+      "step": 36736
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.612625820522355e-08,
+      "loss": 0.5787,
+      "step": 36737
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.603607119759269e-08,
+      "loss": 1.0197,
+      "step": 36738
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5945956503572485e-08,
+      "loss": 0.8944,
+      "step": 36739
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5855914123817966e-08,
+      "loss": 0.5195,
+      "step": 36740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.576594405898195e-08,
+      "loss": 0.8329,
+      "step": 36741
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5676046309721677e-08,
+      "loss": 0.8486,
+      "step": 36742
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.558622087668886e-08,
+      "loss": 1.174,
+      "step": 36743
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.5496467760537406e-08,
+      "loss": 0.7627,
+      "step": 36744
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.540678696192015e-08,
+      "loss": 0.909,
+      "step": 36745
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.531717848148877e-08,
+      "loss": 0.6664,
+      "step": 36746
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.522764231989608e-08,
+      "loss": 0.6531,
+      "step": 36747
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.513817847779157e-08,
+      "loss": 0.8001,
+      "step": 36748
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.504878695582694e-08,
+      "loss": 0.5509,
+      "step": 36749
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4959467754651665e-08,
+      "loss": 0.6755,
+      "step": 36750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4870220874915225e-08,
+      "loss": 0.7056,
+      "step": 36751
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4781046317267103e-08,
+      "loss": 0.8197,
+      "step": 36752
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4691944082354565e-08,
+      "loss": 0.8042,
+      "step": 36753
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.460291417082819e-08,
+      "loss": 1.0664,
+      "step": 36754
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.451395658333303e-08,
+      "loss": 0.6279,
+      "step": 36755
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4425071320517445e-08,
+      "loss": 0.7491,
+      "step": 36756
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4336258383027585e-08,
+      "loss": 0.3187,
+      "step": 36757
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4247517771507386e-08,
+      "loss": 0.8408,
+      "step": 36758
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4158849486604107e-08,
+      "loss": 0.7825,
+      "step": 36759
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.4070253528962776e-08,
+      "loss": 0.8631,
+      "step": 36760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.398172989922734e-08,
+      "loss": 0.8034,
+      "step": 36761
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.389327859804061e-08,
+      "loss": 0.7893,
+      "step": 36762
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3804899626046513e-08,
+      "loss": 0.6049,
+      "step": 36763
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.371659298388676e-08,
+      "loss": 0.7336,
+      "step": 36764
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3628358672205285e-08,
+      "loss": 0.8136,
+      "step": 36765
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.354019669164267e-08,
+      "loss": 0.6714,
+      "step": 36766
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.345210704284176e-08,
+      "loss": 0.9687,
+      "step": 36767
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.336408972643981e-08,
+      "loss": 1.0497,
+      "step": 36768
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.327614474307963e-08,
+      "loss": 0.6186,
+      "step": 36769
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3188272093399604e-08,
+      "loss": 0.8074,
+      "step": 36770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.310047177803923e-08,
+      "loss": 0.8935,
+      "step": 36771
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.3012743797635766e-08,
+      "loss": 0.572,
+      "step": 36772
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.292508815282982e-08,
+      "loss": 0.9041,
+      "step": 36773
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2837504844256427e-08,
+      "loss": 0.8086,
+      "step": 36774
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2749993872552866e-08,
+      "loss": 0.8283,
+      "step": 36775
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2662555238356395e-08,
+      "loss": 0.8815,
+      "step": 36776
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2575188942303176e-08,
+      "loss": 1.0257,
+      "step": 36777
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2487894985027154e-08,
+      "loss": 1.1426,
+      "step": 36778
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.240067336716337e-08,
+      "loss": 0.6783,
+      "step": 36779
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.231352408934687e-08,
+      "loss": 0.7772,
+      "step": 36780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.222644715221159e-08,
+      "loss": 0.5844,
+      "step": 36781
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.213944255639036e-08,
+      "loss": 0.7823,
+      "step": 36782
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.2052510302514904e-08,
+      "loss": 0.8721,
+      "step": 36783
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.196565039121804e-08,
+      "loss": 0.7693,
+      "step": 36784
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.187886282313259e-08,
+      "loss": 0.8976,
+      "step": 36785
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1792147598889175e-08,
+      "loss": 0.7288,
+      "step": 36786
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.170550471911728e-08,
+      "loss": 0.7662,
+      "step": 36787
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1618934184447525e-08,
+      "loss": 0.9522,
+      "step": 36788
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.153243599550939e-08,
+      "loss": 0.89,
+      "step": 36789
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.144601015293238e-08,
+      "loss": 0.5824,
+      "step": 36790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.135965665734488e-08,
+      "loss": 0.4714,
+      "step": 36791
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1273375509375276e-08,
+      "loss": 0.7899,
+      "step": 36792
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.1187166709649736e-08,
+      "loss": 0.8579,
+      "step": 36793
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.110103025879665e-08,
+      "loss": 0.6564,
+      "step": 36794
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.101496615744106e-08,
+      "loss": 0.7658,
+      "step": 36795
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.092897440620914e-08,
+      "loss": 0.7433,
+      "step": 36796
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.084305500572706e-08,
+      "loss": 0.922,
+      "step": 36797
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.075720795661876e-08,
+      "loss": 0.8932,
+      "step": 36798
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0671433259509296e-08,
+      "loss": 0.589,
+      "step": 36799
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0585730915022615e-08,
+      "loss": 0.886,
+      "step": 36800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.050010092378044e-08,
+      "loss": 0.6638,
+      "step": 36801
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0414543286405604e-08,
+      "loss": 0.7376,
+      "step": 36802
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0329058003522054e-08,
+      "loss": 0.4619,
+      "step": 36803
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.0243645075750414e-08,
+      "loss": 0.6197,
+      "step": 36804
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.01583045037124e-08,
+      "loss": 0.8965,
+      "step": 36805
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 5.007303628802751e-08,
+      "loss": 0.7385,
+      "step": 36806
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.998784042931637e-08,
+      "loss": 0.8783,
+      "step": 36807
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.990271692819848e-08,
+      "loss": 0.641,
+      "step": 36808
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.981766578529334e-08,
+      "loss": 0.8762,
+      "step": 36809
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.973268700121936e-08,
+      "loss": 0.6762,
+      "step": 36810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.964778057659381e-08,
+      "loss": 0.9981,
+      "step": 36811
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.956294651203397e-08,
+      "loss": 0.9268,
+      "step": 36812
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9478184808158245e-08,
+      "loss": 0.7411,
+      "step": 36813
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.93934954655817e-08,
+      "loss": 0.9032,
+      "step": 36814
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.9308878484920495e-08,
+      "loss": 0.922,
+      "step": 36815
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.922433386679082e-08,
+      "loss": 1.0237,
+      "step": 36816
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.913986161180662e-08,
+      "loss": 0.8394,
+      "step": 36817
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.905546172058184e-08,
+      "loss": 0.885,
+      "step": 36818
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.897113419373156e-08,
+      "loss": 0.7302,
+      "step": 36819
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.888687903186751e-08,
+      "loss": 0.7478,
+      "step": 36820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.880269623560252e-08,
+      "loss": 1.0905,
+      "step": 36821
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.871858580555056e-08,
+      "loss": 0.8779,
+      "step": 36822
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8634547742322234e-08,
+      "loss": 0.7107,
+      "step": 36823
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.855058204652707e-08,
+      "loss": 0.8573,
+      "step": 36824
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.846668871877902e-08,
+      "loss": 0.5645,
+      "step": 36825
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.838286775968537e-08,
+      "loss": 0.6227,
+      "step": 36826
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8299119169856744e-08,
+      "loss": 0.8325,
+      "step": 36827
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.8215442949901546e-08,
+      "loss": 0.6623,
+      "step": 36828
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.813183910042929e-08,
+      "loss": 1.09,
+      "step": 36829
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.804830762204726e-08,
+      "loss": 0.7632,
+      "step": 36830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.796484851536276e-08,
+      "loss": 0.7481,
+      "step": 36831
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7881461780983073e-08,
+      "loss": 0.5831,
+      "step": 36832
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.779814741951549e-08,
+      "loss": 0.6931,
+      "step": 36833
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.771490543156398e-08,
+      "loss": 0.4407,
+      "step": 36834
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7631735817735835e-08,
+      "loss": 0.5242,
+      "step": 36835
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.75486385786339e-08,
+      "loss": 0.9672,
+      "step": 36836
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.746561371486436e-08,
+      "loss": 0.8018,
+      "step": 36837
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7382661227028946e-08,
+      "loss": 0.8024,
+      "step": 36838
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7299781115732745e-08,
+      "loss": 1.1647,
+      "step": 36839
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.721697338157749e-08,
+      "loss": 1.013,
+      "step": 36840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.713423802516604e-08,
+      "loss": 0.7158,
+      "step": 36841
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.7051575047099006e-08,
+      "loss": 0.7836,
+      "step": 36842
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6968984447978147e-08,
+      "loss": 0.8478,
+      "step": 36843
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6886466228405204e-08,
+      "loss": 0.6568,
+      "step": 36844
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.680402038897747e-08,
+      "loss": 0.7326,
+      "step": 36845
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6721646930297794e-08,
+      "loss": 0.8222,
+      "step": 36846
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.663934585296348e-08,
+      "loss": 1.0249,
+      "step": 36847
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.655711715757183e-08,
+      "loss": 0.7575,
+      "step": 36848
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.647496084472347e-08,
+      "loss": 0.7935,
+      "step": 36849
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6392876915013486e-08,
+      "loss": 0.9892,
+      "step": 36850
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.631086536904028e-08,
+      "loss": 0.4509,
+      "step": 36851
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.6228926207400047e-08,
+      "loss": 0.6769,
+      "step": 36852
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.614705943068787e-08,
+      "loss": 0.751,
+      "step": 36853
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.606526503949993e-08,
+      "loss": 0.8128,
+      "step": 36854
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.59835430344302e-08,
+      "loss": 0.8183,
+      "step": 36855
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.590189341607487e-08,
+      "loss": 0.7322,
+      "step": 36856
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5820316185024584e-08,
+      "loss": 0.5436,
+      "step": 36857
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.573881134187552e-08,
+      "loss": 0.7398,
+      "step": 36858
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.565737888721944e-08,
+      "loss": 0.6473,
+      "step": 36859
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.557601882164808e-08,
+      "loss": 0.6303,
+      "step": 36860
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.549473114575209e-08,
+      "loss": 0.9347,
+      "step": 36861
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5413515860125436e-08,
+      "loss": 0.8252,
+      "step": 36862
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.533237296535653e-08,
+      "loss": 0.824,
+      "step": 36863
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.5251302462036015e-08,
+      "loss": 0.9416,
+      "step": 36864
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.517030435075342e-08,
+      "loss": 0.7804,
+      "step": 36865
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.508937863209717e-08,
+      "loss": 0.5104,
+      "step": 36866
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.500852530665789e-08,
+      "loss": 0.7481,
+      "step": 36867
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4927744375020676e-08,
+      "loss": 0.7448,
+      "step": 36868
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4847035837775055e-08,
+      "loss": 0.785,
+      "step": 36869
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.476639969550722e-08,
+      "loss": 0.8415,
+      "step": 36870
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.468583594880449e-08,
+      "loss": 0.8269,
+      "step": 36871
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.460534459825083e-08,
+      "loss": 0.8701,
+      "step": 36872
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.452492564443245e-08,
+      "loss": 0.8764,
+      "step": 36873
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.444457908793443e-08,
+      "loss": 0.8295,
+      "step": 36874
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.436430492934185e-08,
+      "loss": 0.9316,
+      "step": 36875
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.428410316923648e-08,
+      "loss": 0.5932,
+      "step": 36876
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.4203973808203406e-08,
+      "loss": 0.9066,
+      "step": 36877
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.412391684682438e-08,
+      "loss": 0.8244,
+      "step": 36878
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.404393228568227e-08,
+      "loss": 0.9475,
+      "step": 36879
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.396402012535772e-08,
+      "loss": 0.7968,
+      "step": 36880
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.38841803664325e-08,
+      "loss": 0.8164,
+      "step": 36881
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3804413009487235e-08,
+      "loss": 0.654,
+      "step": 36882
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.372471805510148e-08,
+      "loss": 0.7863,
+      "step": 36883
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.364509550385587e-08,
+      "loss": 0.8031,
+      "step": 36884
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3565545356327734e-08,
+      "loss": 0.7938,
+      "step": 36885
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.34860676130977e-08,
+      "loss": 0.6783,
+      "step": 36886
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3406662274741996e-08,
+      "loss": 0.691,
+      "step": 36887
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.332732934183903e-08,
+      "loss": 0.8937,
+      "step": 36888
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.3248068814965014e-08,
+      "loss": 0.8625,
+      "step": 36889
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.316888069469727e-08,
+      "loss": 0.9136,
+      "step": 36890
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.308976498161088e-08,
+      "loss": 0.9653,
+      "step": 36891
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.301072167628095e-08,
+      "loss": 0.779,
+      "step": 36892
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.293175077928258e-08,
+      "loss": 0.6902,
+      "step": 36893
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.285285229119085e-08,
+      "loss": 0.9107,
+      "step": 36894
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2774026212578646e-08,
+      "loss": 0.5723,
+      "step": 36895
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.269527254401995e-08,
+      "loss": 0.6855,
+      "step": 36896
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2616591286086525e-08,
+      "loss": 0.8382,
+      "step": 36897
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.253798243935125e-08,
+      "loss": 0.7604,
+      "step": 36898
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2459446004384785e-08,
+      "loss": 0.9147,
+      "step": 36899
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2380981981759994e-08,
+      "loss": 0.9581,
+      "step": 36900
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2302590372046424e-08,
+      "loss": 1.0774,
+      "step": 36901
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2224271175813625e-08,
+      "loss": 0.8891,
+      "step": 36902
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.2146024393632245e-08,
+      "loss": 0.9693,
+      "step": 36903
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.206785002606961e-08,
+      "loss": 0.933,
+      "step": 36904
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.198974807369638e-08,
+      "loss": 0.8901,
+      "step": 36905
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.191171853707876e-08,
+      "loss": 0.8394,
+      "step": 36906
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.18337614167863e-08,
+      "loss": 0.7857,
+      "step": 36907
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1755876713382994e-08,
+      "loss": 0.8457,
+      "step": 36908
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.167806442743838e-08,
+      "loss": 0.9537,
+      "step": 36909
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.160032455951535e-08,
+      "loss": 0.779,
+      "step": 36910
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.152265711018122e-08,
+      "loss": 1.0737,
+      "step": 36911
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.144506207999999e-08,
+      "loss": 0.5684,
+      "step": 36912
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.136753946953565e-08,
+      "loss": 0.9314,
+      "step": 36913
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.12900892793533e-08,
+      "loss": 0.6,
+      "step": 36914
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.1212711510015826e-08,
+      "loss": 0.6046,
+      "step": 36915
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.113540616208389e-08,
+      "loss": 0.9448,
+      "step": 36916
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.10581732361226e-08,
+      "loss": 0.6044,
+      "step": 36917
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0981012732691505e-08,
+      "loss": 0.9147,
+      "step": 36918
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.090392465235238e-08,
+      "loss": 0.7114,
+      "step": 36919
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.0826908995665884e-08,
+      "loss": 1.057,
+      "step": 36920
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.074996576319157e-08,
+      "loss": 0.8394,
+      "step": 36921
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.06730949554901e-08,
+      "loss": 0.5499,
+      "step": 36922
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.059629657311992e-08,
+      "loss": 0.7204,
+      "step": 36923
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.051957061663836e-08,
+      "loss": 0.7442,
+      "step": 36924
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.044291708660608e-08,
+      "loss": 0.8561,
+      "step": 36925
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.036633598357709e-08,
+      "loss": 1.0131,
+      "step": 36926
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.028982730811204e-08,
+      "loss": 0.8782,
+      "step": 36927
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.021339106076494e-08,
+      "loss": 0.6338,
+      "step": 36928
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.01370272420909e-08,
+      "loss": 0.674,
+      "step": 36929
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 4.006073585264725e-08,
+      "loss": 0.999,
+      "step": 36930
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.998451689298688e-08,
+      "loss": 0.6374,
+      "step": 36931
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9908370363667125e-08,
+      "loss": 1.0452,
+      "step": 36932
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.983229626523866e-08,
+      "loss": 0.7112,
+      "step": 36933
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9756294598255476e-08,
+      "loss": 0.7384,
+      "step": 36934
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.968036536327047e-08,
+      "loss": 0.7521,
+      "step": 36935
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.960450856083653e-08,
+      "loss": 0.6467,
+      "step": 36936
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.952872419150433e-08,
+      "loss": 0.7185,
+      "step": 36937
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.945301225582565e-08,
+      "loss": 0.6107,
+      "step": 36938
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.937737275435116e-08,
+      "loss": 0.8594,
+      "step": 36939
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.930180568762931e-08,
+      "loss": 0.6431,
+      "step": 36940
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.9226311056211886e-08,
+      "loss": 0.9869,
+      "step": 36941
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.915088886064622e-08,
+      "loss": 1.1302,
+      "step": 36942
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.907553910148076e-08,
+      "loss": 0.808,
+      "step": 36943
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.900026177926508e-08,
+      "loss": 0.8386,
+      "step": 36944
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.89250568945454e-08,
+      "loss": 1.1289,
+      "step": 36945
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.884992444786906e-08,
+      "loss": 0.927,
+      "step": 36946
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.877486443978229e-08,
+      "loss": 1.0123,
+      "step": 36947
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.869987687083132e-08,
+      "loss": 0.7766,
+      "step": 36948
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.862496174156016e-08,
+      "loss": 0.7905,
+      "step": 36949
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.855011905251504e-08,
+      "loss": 0.832,
+      "step": 36950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.847534880423887e-08,
+      "loss": 0.6883,
+      "step": 36951
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.840065099727675e-08,
+      "loss": 0.8918,
+      "step": 36952
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.832602563217158e-08,
+      "loss": 0.8535,
+      "step": 36953
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.825147270946516e-08,
+      "loss": 0.6594,
+      "step": 36954
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8176992229700396e-08,
+      "loss": 1.025,
+      "step": 36955
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.8102584193419054e-08,
+      "loss": 0.7662,
+      "step": 36956
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.802824860116183e-08,
+      "loss": 0.8454,
+      "step": 36957
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.79539854534694e-08,
+      "loss": 0.6806,
+      "step": 36958
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.787979475088244e-08,
+      "loss": 0.8132,
+      "step": 36959
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7805676493938294e-08,
+      "loss": 0.7669,
+      "step": 36960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7731630683178756e-08,
+      "loss": 0.8085,
+      "step": 36961
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7657657319141175e-08,
+      "loss": 0.6611,
+      "step": 36962
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7583756402362894e-08,
+      "loss": 0.8696,
+      "step": 36963
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7509927933382374e-08,
+      "loss": 0.7322,
+      "step": 36964
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.743617191273474e-08,
+      "loss": 0.8726,
+      "step": 36965
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.736248834095846e-08,
+      "loss": 0.5672,
+      "step": 36966
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7288877218588645e-08,
+      "loss": 0.6856,
+      "step": 36967
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.721533854616044e-08,
+      "loss": 0.9991,
+      "step": 36968
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.714187232420896e-08,
+      "loss": 0.9273,
+      "step": 36969
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.7068478553267116e-08,
+      "loss": 0.7988,
+      "step": 36970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.699515723387004e-08,
+      "loss": 0.6829,
+      "step": 36971
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.692190836655063e-08,
+      "loss": 0.681,
+      "step": 36972
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.68487319518418e-08,
+      "loss": 0.9534,
+      "step": 36973
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.677562799027534e-08,
+      "loss": 0.7747,
+      "step": 36974
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6702596482381946e-08,
+      "loss": 0.6185,
+      "step": 36975
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.662963742869452e-08,
+      "loss": 0.8274,
+      "step": 36976
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.655675082974375e-08,
+      "loss": 0.8998,
+      "step": 36977
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6483936686056986e-08,
+      "loss": 0.9484,
+      "step": 36978
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6411194998166035e-08,
+      "loss": 1.0056,
+      "step": 36979
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.633852576659935e-08,
+      "loss": 0.856,
+      "step": 36980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6265928991885414e-08,
+      "loss": 0.8133,
+      "step": 36981
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.619340467455157e-08,
+      "loss": 1.0073,
+      "step": 36982
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.612095281512629e-08,
+      "loss": 0.5734,
+      "step": 36983
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.6048573414135816e-08,
+      "loss": 0.8446,
+      "step": 36984
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.597626647210639e-08,
+      "loss": 0.8986,
+      "step": 36985
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.590403198956427e-08,
+      "loss": 0.9083,
+      "step": 36986
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.583186996703458e-08,
+      "loss": 0.7539,
+      "step": 36987
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5759780405042467e-08,
+      "loss": 0.7837,
+      "step": 36988
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.568776330411084e-08,
+      "loss": 0.8212,
+      "step": 36989
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5615818664764825e-08,
+      "loss": 0.8371,
+      "step": 36990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.554394648752624e-08,
+      "loss": 0.6632,
+      "step": 36991
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.547214677292021e-08,
+      "loss": 0.9282,
+      "step": 36992
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5400419521466335e-08,
+      "loss": 0.8154,
+      "step": 36993
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.53287647336864e-08,
+      "loss": 0.8819,
+      "step": 36994
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5257182410103344e-08,
+      "loss": 0.7707,
+      "step": 36995
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.518567255123673e-08,
+      "loss": 0.7772,
+      "step": 36996
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5114235157605036e-08,
+      "loss": 0.678,
+      "step": 36997
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.504287022973119e-08,
+      "loss": 0.8783,
+      "step": 36998
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.497157776813032e-08,
+      "loss": 0.8128,
+      "step": 36999
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.490035777332423e-08,
+      "loss": 0.9205,
+      "step": 37000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.482921024582808e-08,
+      "loss": 0.775,
+      "step": 37001
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.475813518616144e-08,
+      "loss": 0.8225,
+      "step": 37002
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.468713259483947e-08,
+      "loss": 0.6045,
+      "step": 37003
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.461620247237951e-08,
+      "loss": 0.6344,
+      "step": 37004
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.454534481929783e-08,
+      "loss": 0.7747,
+      "step": 37005
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.447455963610846e-08,
+      "loss": 0.8932,
+      "step": 37006
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.440384692332765e-08,
+      "loss": 0.8625,
+      "step": 37007
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.433320668146722e-08,
+      "loss": 0.9817,
+      "step": 37008
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.426263891104342e-08,
+      "loss": 0.7084,
+      "step": 37009
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.419214361256806e-08,
+      "loss": 0.963,
+      "step": 37010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.412172078655407e-08,
+      "loss": 0.9917,
+      "step": 37011
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.405137043351325e-08,
+      "loss": 0.8094,
+      "step": 37012
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3981092553958536e-08,
+      "loss": 0.874,
+      "step": 37013
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3910887148399516e-08,
+      "loss": 0.726,
+      "step": 37014
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3840754217346893e-08,
+      "loss": 0.6064,
+      "step": 37015
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.377069376131137e-08,
+      "loss": 1.23,
+      "step": 37016
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.370070578080142e-08,
+      "loss": 0.7366,
+      "step": 37017
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.363079027632776e-08,
+      "loss": 0.853,
+      "step": 37018
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3560947248396645e-08,
+      "loss": 0.5439,
+      "step": 37019
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.349117669751767e-08,
+      "loss": 0.9276,
+      "step": 37020
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.34214786241982e-08,
+      "loss": 0.793,
+      "step": 37021
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.335185302894339e-08,
+      "loss": 0.9592,
+      "step": 37022
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.328229991226173e-08,
+      "loss": 0.7812,
+      "step": 37023
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.321281927465725e-08,
+      "loss": 0.8516,
+      "step": 37024
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3143411116636215e-08,
+      "loss": 0.714,
+      "step": 37025
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3074075438702666e-08,
+      "loss": 0.8755,
+      "step": 37026
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.3004812241361763e-08,
+      "loss": 1.0752,
+      "step": 37027
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.293562152511642e-08,
+      "loss": 0.8729,
+      "step": 37028
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.286650329046959e-08,
+      "loss": 1.0596,
+      "step": 37029
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.279745753792529e-08,
+      "loss": 0.8029,
+      "step": 37030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.272848426798425e-08,
+      "loss": 0.9192,
+      "step": 37031
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.265958348114828e-08,
+      "loss": 0.825,
+      "step": 37032
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.2590755177919206e-08,
+      "loss": 0.857,
+      "step": 37033
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.252199935879663e-08,
+      "loss": 0.8367,
+      "step": 37034
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.245331602428126e-08,
+      "loss": 0.578,
+      "step": 37035
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.238470517487158e-08,
+      "loss": 0.9863,
+      "step": 37036
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.231616681106831e-08,
+      "loss": 0.821,
+      "step": 37037
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.224770093336771e-08,
+      "loss": 0.8824,
+      "step": 37038
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.217930754226939e-08,
+      "loss": 1.0434,
+      "step": 37039
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.211098663826961e-08,
+      "loss": 1.1371,
+      "step": 37040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.204273822186577e-08,
+      "loss": 1.0408,
+      "step": 37041
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.197456229355411e-08,
+      "loss": 1.0363,
+      "step": 37042
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.190645885382981e-08,
+      "loss": 0.6836,
+      "step": 37043
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.183842790318914e-08,
+      "loss": 0.6153,
+      "step": 37044
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.177046944212725e-08,
+      "loss": 1.1189,
+      "step": 37045
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.170258347113597e-08,
+      "loss": 0.8072,
+      "step": 37046
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.163476999071158e-08,
+      "loss": 0.8201,
+      "step": 37047
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.1567029001344784e-08,
+      "loss": 1.0364,
+      "step": 37048
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.149936050353075e-08,
+      "loss": 0.7828,
+      "step": 37049
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.14317644977602e-08,
+      "loss": 0.8206,
+      "step": 37050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.136424098452495e-08,
+      "loss": 0.929,
+      "step": 37051
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.129678996431462e-08,
+      "loss": 0.5899,
+      "step": 37052
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.122941143762215e-08,
+      "loss": 0.8036,
+      "step": 37053
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.116210540493603e-08,
+      "loss": 0.7602,
+      "step": 37054
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.109487186674698e-08,
+      "loss": 0.817,
+      "step": 37055
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.102771082354239e-08,
+      "loss": 0.6773,
+      "step": 37056
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0960622275810756e-08,
+      "loss": 0.9419,
+      "step": 37057
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0893606224041696e-08,
+      "loss": 1.0204,
+      "step": 37058
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.082666266872036e-08,
+      "loss": 0.9397,
+      "step": 37059
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.075979161033526e-08,
+      "loss": 0.6204,
+      "step": 37060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.069299304937157e-08,
+      "loss": 0.8625,
+      "step": 37061
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.062626698631554e-08,
+      "loss": 0.5836,
+      "step": 37062
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.055961342165237e-08,
+      "loss": 0.5743,
+      "step": 37063
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0493032355866094e-08,
+      "loss": 0.9894,
+      "step": 37064
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0426523789442994e-08,
+      "loss": 0.729,
+      "step": 37065
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.036008772286492e-08,
+      "loss": 0.8672,
+      "step": 37066
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.02937241566148e-08,
+      "loss": 0.8918,
+      "step": 37067
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0227433091176703e-08,
+      "loss": 0.8214,
+      "step": 37068
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0161214527031354e-08,
+      "loss": 0.9351,
+      "step": 37069
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.009506846466059e-08,
+      "loss": 0.6631,
+      "step": 37070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.0028994904545136e-08,
+      "loss": 0.6426,
+      "step": 37071
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9962993847166834e-08,
+      "loss": 0.8816,
+      "step": 37072
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.989706529300418e-08,
+      "loss": 0.719,
+      "step": 37073
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9831209242537907e-08,
+      "loss": 0.7562,
+      "step": 37074
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9765425696245408e-08,
+      "loss": 0.7763,
+      "step": 37075
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9699714654607413e-08,
+      "loss": 0.7741,
+      "step": 37076
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9634076118099098e-08,
+      "loss": 0.7111,
+      "step": 37077
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9568510087198966e-08,
+      "loss": 0.7237,
+      "step": 37078
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9503016562384413e-08,
+      "loss": 0.7484,
+      "step": 37079
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9437595544130615e-08,
+      "loss": 0.8514,
+      "step": 37080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.937224703291386e-08,
+      "loss": 0.7215,
+      "step": 37081
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.930697102920821e-08,
+      "loss": 0.7238,
+      "step": 37082
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.9241767533491062e-08,
+      "loss": 0.6078,
+      "step": 37083
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.917663654623426e-08,
+      "loss": 0.8229,
+      "step": 37084
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.911157806791298e-08,
+      "loss": 0.4785,
+      "step": 37085
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.904659209899907e-08,
+      "loss": 1.052,
+      "step": 37086
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8981678639965484e-08,
+      "loss": 0.9632,
+      "step": 37087
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8916837691285172e-08,
+      "loss": 0.8213,
+      "step": 37088
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8852069253427763e-08,
+      "loss": 0.8643,
+      "step": 37089
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8787373326866207e-08,
+      "loss": 0.5746,
+      "step": 37090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8722749912069025e-08,
+      "loss": 0.6931,
+      "step": 37091
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.865819900950806e-08,
+      "loss": 0.6927,
+      "step": 37092
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.859372061965182e-08,
+      "loss": 0.9661,
+      "step": 37093
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8529314742969938e-08,
+      "loss": 0.5984,
+      "step": 37094
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8464981379929814e-08,
+      "loss": 0.7101,
+      "step": 37095
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.840072053099996e-08,
+      "loss": 0.632,
+      "step": 37096
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.833653219664667e-08,
+      "loss": 0.653,
+      "step": 37097
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8272416377337353e-08,
+      "loss": 0.9194,
+      "step": 37098
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8208373073539407e-08,
+      "loss": 0.6834,
+      "step": 37099
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.81444022857158e-08,
+      "loss": 0.7838,
+      "step": 37100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8080504014335042e-08,
+      "loss": 0.8518,
+      "step": 37101
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.8016678259860098e-08,
+      "loss": 0.6155,
+      "step": 37102
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.795292502275393e-08,
+      "loss": 0.5137,
+      "step": 37103
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7889244303482833e-08,
+      "loss": 1.0989,
+      "step": 37104
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7825636102507547e-08,
+      "loss": 0.8285,
+      "step": 37105
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.776210042029215e-08,
+      "loss": 0.6937,
+      "step": 37106
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.769863725729849e-08,
+      "loss": 0.6835,
+      "step": 37107
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.763524661398731e-08,
+      "loss": 0.8211,
+      "step": 37108
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7571928490820465e-08,
+      "loss": 0.6129,
+      "step": 37109
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7508682888257587e-08,
+      "loss": 0.7569,
+      "step": 37110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.744550980675831e-08,
+      "loss": 0.7907,
+      "step": 37111
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.738240924678448e-08,
+      "loss": 0.6763,
+      "step": 37112
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7319381208791296e-08,
+      "loss": 0.8322,
+      "step": 37113
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7256425693239497e-08,
+      "loss": 0.648,
+      "step": 37114
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7193542700586494e-08,
+      "loss": 0.9618,
+      "step": 37115
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.7130732231290814e-08,
+      "loss": 0.6213,
+      "step": 37116
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.706799428580653e-08,
+      "loss": 1.0424,
+      "step": 37117
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.700532886459106e-08,
+      "loss": 1.0558,
+      "step": 37118
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.694273596810071e-08,
+      "loss": 0.9461,
+      "step": 37119
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.688021559679066e-08,
+      "loss": 0.7327,
+      "step": 37120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6817767751114997e-08,
+      "loss": 1.0745,
+      "step": 37121
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6755392431527803e-08,
+      "loss": 0.7323,
+      "step": 37122
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.669308963848316e-08,
+      "loss": 0.8927,
+      "step": 37123
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6630859372434038e-08,
+      "loss": 0.9574,
+      "step": 37124
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6568701633832295e-08,
+      "loss": 0.7044,
+      "step": 37125
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6506616423130905e-08,
+      "loss": 0.9823,
+      "step": 37126
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6444603740780617e-08,
+      "loss": 0.6798,
+      "step": 37127
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6382663587233292e-08,
+      "loss": 1.0178,
+      "step": 37128
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6320795962937464e-08,
+      "loss": 0.7597,
+      "step": 37129
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6259000868346097e-08,
+      "loss": 0.6893,
+      "step": 37130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6197278303905504e-08,
+      "loss": 0.6318,
+      "step": 37131
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6135628270066437e-08,
+      "loss": 0.8938,
+      "step": 37132
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6074050767276317e-08,
+      "loss": 0.8029,
+      "step": 37133
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.6012545795983668e-08,
+      "loss": 0.6603,
+      "step": 37134
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5951113356634807e-08,
+      "loss": 0.8807,
+      "step": 37135
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5889753449677145e-08,
+      "loss": 0.8414,
+      "step": 37136
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5828466075557e-08,
+      "loss": 0.9176,
+      "step": 37137
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.576725123472068e-08,
+      "loss": 0.9157,
+      "step": 37138
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5706108927611162e-08,
+      "loss": 0.593,
+      "step": 37139
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5645039154675867e-08,
+      "loss": 0.83,
+      "step": 37140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.558404191635666e-08,
+      "loss": 0.9668,
+      "step": 37141
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5523117213097638e-08,
+      "loss": 0.8604,
+      "step": 37142
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.546226504534177e-08,
+      "loss": 0.9453,
+      "step": 37143
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5401485413533154e-08,
+      "loss": 0.8383,
+      "step": 37144
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5340778318111436e-08,
+      "loss": 0.9795,
+      "step": 37145
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5280143759520704e-08,
+      "loss": 0.7461,
+      "step": 37146
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5219581738199496e-08,
+      "loss": 0.9714,
+      "step": 37147
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.515909225458968e-08,
+      "loss": 0.8703,
+      "step": 37148
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.5098675309129794e-08,
+      "loss": 0.9161,
+      "step": 37149
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.50383309022606e-08,
+      "loss": 1.0573,
+      "step": 37150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.497805903441952e-08,
+      "loss": 0.8372,
+      "step": 37151
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.491785970604621e-08,
+      "loss": 0.7647,
+      "step": 37152
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4857732917578093e-08,
+      "loss": 0.6899,
+      "step": 37153
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4797678669452597e-08,
+      "loss": 0.8154,
+      "step": 37154
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4737696962106038e-08,
+      "loss": 0.7862,
+      "step": 37155
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.467778779597363e-08,
+      "loss": 0.8953,
+      "step": 37156
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4617951171493904e-08,
+      "loss": 0.6858,
+      "step": 37157
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4558187089098738e-08,
+      "loss": 0.7419,
+      "step": 37158
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4498495549224456e-08,
+      "loss": 0.7725,
+      "step": 37159
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4438876552305145e-08,
+      "loss": 0.7229,
+      "step": 37160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.43793300987738e-08,
+      "loss": 0.8027,
+      "step": 37161
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4319856189063408e-08,
+      "loss": 0.8558,
+      "step": 37162
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.426045482360695e-08,
+      "loss": 0.7318,
+      "step": 37163
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4201126002837415e-08,
+      "loss": 0.768,
+      "step": 37164
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4141869727183354e-08,
+      "loss": 0.9603,
+      "step": 37165
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.408268599707886e-08,
+      "loss": 0.8851,
+      "step": 37166
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.4023574812952478e-08,
+      "loss": 0.6929,
+      "step": 37167
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3964536175233866e-08,
+      "loss": 1.0473,
+      "step": 37168
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3905570084353792e-08,
+      "loss": 0.8373,
+      "step": 37169
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3846676540739687e-08,
+      "loss": 0.9326,
+      "step": 37170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3787855544820103e-08,
+      "loss": 1.0293,
+      "step": 37171
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3729107097023585e-08,
+      "loss": 0.8236,
+      "step": 37172
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3670431197776457e-08,
+      "loss": 1.1274,
+      "step": 37173
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3611827847506152e-08,
+      "loss": 1.0712,
+      "step": 37174
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.355329704663789e-08,
+      "loss": 0.5608,
+      "step": 37175
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3494838795599106e-08,
+      "loss": 0.5641,
+      "step": 37176
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3436453094812793e-08,
+      "loss": 0.6047,
+      "step": 37177
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.337813994470417e-08,
+      "loss": 0.6989,
+      "step": 37178
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.331989934569845e-08,
+      "loss": 0.9796,
+      "step": 37179
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3261731298217517e-08,
+      "loss": 0.9831,
+      "step": 37180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3203635802685475e-08,
+      "loss": 0.6609,
+      "step": 37181
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3145612859525323e-08,
+      "loss": 0.6962,
+      "step": 37182
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3087662469156723e-08,
+      "loss": 0.8069,
+      "step": 37183
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.3029784632003783e-08,
+      "loss": 0.5721,
+      "step": 37184
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2971979348485053e-08,
+      "loss": 0.7738,
+      "step": 37185
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2914246619023528e-08,
+      "loss": 0.4585,
+      "step": 37186
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2856586444036655e-08,
+      "loss": 0.9863,
+      "step": 37187
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2798998823944097e-08,
+      "loss": 0.6892,
+      "step": 37188
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2741483759165518e-08,
+      "loss": 0.5861,
+      "step": 37189
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2684041250119472e-08,
+      "loss": 0.7815,
+      "step": 37190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2626671297222292e-08,
+      "loss": 0.9351,
+      "step": 37191
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2569373900891423e-08,
+      "loss": 0.6605,
+      "step": 37192
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.251214906154542e-08,
+      "loss": 0.6675,
+      "step": 37193
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.245499677959839e-08,
+      "loss": 0.8809,
+      "step": 37194
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.239791705546668e-08,
+      "loss": 1.0583,
+      "step": 37195
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2340909889565498e-08,
+      "loss": 0.774,
+      "step": 37196
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2283975282308966e-08,
+      "loss": 0.5972,
+      "step": 37197
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.22271132341112e-08,
+      "loss": 0.7156,
+      "step": 37198
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.217032374538519e-08,
+      "loss": 0.726,
+      "step": 37199
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2113606816546172e-08,
+      "loss": 0.8321,
+      "step": 37200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2056962448003817e-08,
+      "loss": 0.8609,
+      "step": 37201
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.2000390640172232e-08,
+      "loss": 0.8344,
+      "step": 37202
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.194389139346109e-08,
+      "loss": 0.9941,
+      "step": 37203
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1887464708282292e-08,
+      "loss": 0.7542,
+      "step": 37204
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1831110585045502e-08,
+      "loss": 0.7944,
+      "step": 37205
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1774829024161502e-08,
+      "loss": 1.2719,
+      "step": 37206
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.171862002603886e-08,
+      "loss": 0.821,
+      "step": 37207
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1662483591086137e-08,
+      "loss": 0.7299,
+      "step": 37208
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.160641971971189e-08,
+      "loss": 1.1489,
+      "step": 37209
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1550428412323577e-08,
+      "loss": 0.9497,
+      "step": 37210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1494509669329757e-08,
+      "loss": 0.8295,
+      "step": 37211
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.143866349113566e-08,
+      "loss": 0.6516,
+      "step": 37212
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1382889878146517e-08,
+      "loss": 0.7881,
+      "step": 37213
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1327188830770894e-08,
+      "loss": 0.6265,
+      "step": 37214
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.127156034941069e-08,
+      "loss": 0.8023,
+      "step": 37215
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1216004434472247e-08,
+      "loss": 0.9477,
+      "step": 37216
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1160521086358575e-08,
+      "loss": 0.9015,
+      "step": 37217
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1105110305474906e-08,
+      "loss": 0.9565,
+      "step": 37218
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.1049772092222032e-08,
+      "loss": 0.8844,
+      "step": 37219
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.099450644700407e-08,
+      "loss": 0.711,
+      "step": 37220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0939313370221813e-08,
+      "loss": 0.787,
+      "step": 37221
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0884192862276055e-08,
+      "loss": 1.1412,
+      "step": 37222
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.08291449235698e-08,
+      "loss": 0.8076,
+      "step": 37223
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.077416955450162e-08,
+      "loss": 0.7467,
+      "step": 37224
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0719266755472313e-08,
+      "loss": 1.0181,
+      "step": 37225
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0664436526881548e-08,
+      "loss": 0.646,
+      "step": 37226
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.060967886912568e-08,
+      "loss": 0.6299,
+      "step": 37227
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0554993782604394e-08,
+      "loss": 0.8934,
+      "step": 37228
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.050038126771625e-08,
+      "loss": 0.9125,
+      "step": 37229
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0445841324856497e-08,
+      "loss": 0.8106,
+      "step": 37230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.03913739544237e-08,
+      "loss": 0.5,
+      "step": 37231
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0336979156813097e-08,
+      "loss": 0.861,
+      "step": 37232
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.028265693241993e-08,
+      "loss": 0.8712,
+      "step": 37233
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.022840728163944e-08,
+      "loss": 0.8999,
+      "step": 37234
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.017423020486575e-08,
+      "loss": 0.7543,
+      "step": 37235
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.0120125702494108e-08,
+      "loss": 1.0839,
+      "step": 37236
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.006609377491642e-08,
+      "loss": 0.5772,
+      "step": 37237
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2.001213442252681e-08,
+      "loss": 1.0312,
+      "step": 37238
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.995824764571719e-08,
+      "loss": 0.8896,
+      "step": 37239
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9904433444879468e-08,
+      "loss": 1.055,
+      "step": 37240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9850691820404443e-08,
+      "loss": 0.606,
+      "step": 37241
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9797022772684027e-08,
+      "loss": 0.717,
+      "step": 37242
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9743426302107903e-08,
+      "loss": 0.8853,
+      "step": 37243
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9689902409065765e-08,
+      "loss": 0.7465,
+      "step": 37244
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9636451093947296e-08,
+      "loss": 0.8554,
+      "step": 37245
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9583072357139965e-08,
+      "loss": 0.8396,
+      "step": 37246
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.952976619903346e-08,
+      "loss": 0.7234,
+      "step": 37247
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9476532620015252e-08,
+      "loss": 1.0523,
+      "step": 37248
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9423371620470587e-08,
+      "loss": 0.6052,
+      "step": 37249
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9370283200789152e-08,
+      "loss": 1.0049,
+      "step": 37250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9317267361353976e-08,
+      "loss": 0.6974,
+      "step": 37251
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9264324102552523e-08,
+      "loss": 0.5985,
+      "step": 37252
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9211453424768932e-08,
+      "loss": 0.8261,
+      "step": 37253
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9158655328388452e-08,
+      "loss": 0.8369,
+      "step": 37254
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.910592981379411e-08,
+      "loss": 0.8637,
+      "step": 37255
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.905327688137115e-08,
+      "loss": 0.4738,
+      "step": 37256
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.9000696531500384e-08,
+      "loss": 1.054,
+      "step": 37257
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8948188764563724e-08,
+      "loss": 1.062,
+      "step": 37258
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.889575358094531e-08,
+      "loss": 0.8879,
+      "step": 37259
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8843390981024835e-08,
+      "loss": 0.5491,
+      "step": 37260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8791100965184218e-08,
+      "loss": 0.8278,
+      "step": 37261
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8738883533803153e-08,
+      "loss": 1.2227,
+      "step": 37262
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8686738687260232e-08,
+      "loss": 1.0218,
+      "step": 37263
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8634666425936255e-08,
+      "loss": 0.9337,
+      "step": 37264
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.858266675020981e-08,
+      "loss": 0.8499,
+      "step": 37265
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8530739660458375e-08,
+      "loss": 0.7057,
+      "step": 37266
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8478885157060534e-08,
+      "loss": 0.9395,
+      "step": 37267
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.842710324039154e-08,
+      "loss": 0.9117,
+      "step": 37268
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8375393910829987e-08,
+      "loss": 0.8772,
+      "step": 37269
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.832375716875001e-08,
+      "loss": 0.7985,
+      "step": 37270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8272193014529095e-08,
+      "loss": 0.7661,
+      "step": 37271
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8220701448540268e-08,
+      "loss": 0.6954,
+      "step": 37272
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.81692824711599e-08,
+      "loss": 0.8998,
+      "step": 37273
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8117936082759913e-08,
+      "loss": 0.9964,
+      "step": 37274
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8066662283715562e-08,
+      "loss": 0.8658,
+      "step": 37275
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.8015461074398776e-08,
+      "loss": 0.732,
+      "step": 37276
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7964332455181478e-08,
+      "loss": 0.5346,
+      "step": 37277
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7913276426436697e-08,
+      "loss": 0.9803,
+      "step": 37278
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7862292988535256e-08,
+      "loss": 0.8105,
+      "step": 37279
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7811382141846856e-08,
+      "loss": 0.8034,
+      "step": 37280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.776054388674342e-08,
+      "loss": 0.9588,
+      "step": 37281
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7709778223593543e-08,
+      "loss": 0.7027,
+      "step": 37282
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7659085152766932e-08,
+      "loss": 0.7949,
+      "step": 37283
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7608464674631064e-08,
+      "loss": 0.9178,
+      "step": 37284
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.755791678955565e-08,
+      "loss": 0.797,
+      "step": 37285
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7507441497907064e-08,
+      "loss": 0.5926,
+      "step": 37286
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7457038800053894e-08,
+      "loss": 0.8831,
+      "step": 37287
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7406708696361408e-08,
+      "loss": 0.7816,
+      "step": 37288
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.735645118719487e-08,
+      "loss": 0.7956,
+      "step": 37289
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7306266272921756e-08,
+      "loss": 0.8941,
+      "step": 37290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.725615395390512e-08,
+      "loss": 0.8421,
+      "step": 37291
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.720611423051133e-08,
+      "loss": 0.8552,
+      "step": 37292
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7156147103102315e-08,
+      "loss": 0.8854,
+      "step": 37293
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7106252572043346e-08,
+      "loss": 0.753,
+      "step": 37294
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.705643063769524e-08,
+      "loss": 0.5892,
+      "step": 37295
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7006681300422157e-08,
+      "loss": 0.6874,
+      "step": 37296
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6957004560584912e-08,
+      "loss": 0.935,
+      "step": 37297
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.690740041854433e-08,
+      "loss": 1.0113,
+      "step": 37298
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6857868874662344e-08,
+      "loss": 0.9156,
+      "step": 37299
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6808409929298665e-08,
+      "loss": 0.7956,
+      "step": 37300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6759023582811894e-08,
+      "loss": 0.6448,
+      "step": 37301
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.670970983556286e-08,
+      "loss": 0.515,
+      "step": 37302
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6660468687910157e-08,
+      "loss": 0.9437,
+      "step": 37303
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6611300140210173e-08,
+      "loss": 0.9033,
+      "step": 37304
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6562204192821507e-08,
+      "loss": 0.788,
+      "step": 37305
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.651318084610165e-08,
+      "loss": 1.2107,
+      "step": 37306
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6464230100405874e-08,
+      "loss": 0.9722,
+      "step": 37307
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6415351956090565e-08,
+      "loss": 0.5914,
+      "step": 37308
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6366546413512096e-08,
+      "loss": 0.8783,
+      "step": 37309
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6317813473024634e-08,
+      "loss": 0.8764,
+      "step": 37310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6269153134983452e-08,
+      "loss": 0.7311,
+      "step": 37311
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6220565399740484e-08,
+      "loss": 0.8249,
+      "step": 37312
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6172050267651006e-08,
+      "loss": 0.9,
+      "step": 37313
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.612360773906696e-08,
+      "loss": 0.8786,
+      "step": 37314
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.6075237814340284e-08,
+      "loss": 0.8868,
+      "step": 37315
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.602694049382403e-08,
+      "loss": 1.112,
+      "step": 37316
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5978715777867915e-08,
+      "loss": 0.7699,
+      "step": 37317
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5930563666822772e-08,
+      "loss": 0.3369,
+      "step": 37318
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5882484161039434e-08,
+      "loss": 0.8016,
+      "step": 37319
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.583447726086762e-08,
+      "loss": 0.7398,
+      "step": 37320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.578654296665705e-08,
+      "loss": 0.8665,
+      "step": 37321
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.573868127875411e-08,
+      "loss": 0.9254,
+      "step": 37322
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5690892197509634e-08,
+      "loss": 0.7075,
+      "step": 37323
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.56431757232689e-08,
+      "loss": 0.8165,
+      "step": 37324
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.559553185637941e-08,
+      "loss": 0.6375,
+      "step": 37325
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.554796059718866e-08,
+      "loss": 0.8101,
+      "step": 37326
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5500461946040823e-08,
+      "loss": 1.0328,
+      "step": 37327
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5453035903283397e-08,
+      "loss": 1.1765,
+      "step": 37328
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.540568246926055e-08,
+      "loss": 0.6611,
+      "step": 37329
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5358401644315348e-08,
+      "loss": 0.9195,
+      "step": 37330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5311193428793058e-08,
+      "loss": 0.6893,
+      "step": 37331
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5264057823036747e-08,
+      "loss": 1.0234,
+      "step": 37332
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5216994827388365e-08,
+      "loss": 0.7215,
+      "step": 37333
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5170004442190967e-08,
+      "loss": 0.6214,
+      "step": 37334
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5123086667786502e-08,
+      "loss": 0.7148,
+      "step": 37335
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.507624150451581e-08,
+      "loss": 1.0384,
+      "step": 37336
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.5029468952718617e-08,
+      "loss": 0.8928,
+      "step": 37337
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4982769012736876e-08,
+      "loss": 0.7417,
+      "step": 37338
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.49361416849092e-08,
+      "loss": 0.7831,
+      "step": 37339
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.488958696957421e-08,
+      "loss": 0.675,
+      "step": 37340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4843104867070523e-08,
+      "loss": 0.5655,
+      "step": 37341
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4796695377737869e-08,
+      "loss": 0.7747,
+      "step": 37342
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4750358501911533e-08,
+      "loss": 0.8008,
+      "step": 37343
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4704094239930133e-08,
+      "loss": 0.6987,
+      "step": 37344
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4657902592128958e-08,
+      "loss": 0.7383,
+      "step": 37345
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4611783558844405e-08,
+      "loss": 0.5987,
+      "step": 37346
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4565737140411762e-08,
+      "loss": 0.7932,
+      "step": 37347
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4519763337166314e-08,
+      "loss": 0.8989,
+      "step": 37348
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4473862149441131e-08,
+      "loss": 0.7622,
+      "step": 37349
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4428033577571498e-08,
+      "loss": 1.0888,
+      "step": 37350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4382277621890483e-08,
+      "loss": 0.9844,
+      "step": 37351
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4336594282730043e-08,
+      "loss": 0.961,
+      "step": 37352
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4290983560422134e-08,
+      "loss": 0.804,
+      "step": 37353
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4245445455299822e-08,
+      "loss": 0.6427,
+      "step": 37354
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4199979967692845e-08,
+      "loss": 0.6657,
+      "step": 37355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.415458709793316e-08,
+      "loss": 0.6886,
+      "step": 37356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.41092668463505e-08,
+      "loss": 0.7631,
+      "step": 37357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4064019213273495e-08,
+      "loss": 0.5793,
+      "step": 37358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.4018844199031877e-08,
+      "loss": 0.8586,
+      "step": 37359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3973741803954278e-08,
+      "loss": 0.6028,
+      "step": 37360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3928712028369318e-08,
+      "loss": 0.7406,
+      "step": 37361
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3883754872602296e-08,
+      "loss": 0.7621,
+      "step": 37362
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3838870336981836e-08,
+      "loss": 0.8592,
+      "step": 37363
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3794058421835455e-08,
+      "loss": 0.6665,
+      "step": 37364
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3749319127486228e-08,
+      "loss": 0.6074,
+      "step": 37365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.370465245426167e-08,
+      "loss": 0.7823,
+      "step": 37366
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3660058402484856e-08,
+      "loss": 0.5616,
+      "step": 37367
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3615536972482191e-08,
+      "loss": 0.8094,
+      "step": 37368
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.357108816457564e-08,
+      "loss": 0.6726,
+      "step": 37369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.352671197908939e-08,
+      "loss": 0.7884,
+      "step": 37370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3482408416345406e-08,
+      "loss": 0.8637,
+      "step": 37371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3438177476666758e-08,
+      "loss": 0.9359,
+      "step": 37372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3394019160374305e-08,
+      "loss": 0.6829,
+      "step": 37373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3349933467790011e-08,
+      "loss": 0.8482,
+      "step": 37374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3305920399233618e-08,
+      "loss": 0.5544,
+      "step": 37375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3261979955025983e-08,
+      "loss": 0.8129,
+      "step": 37376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.321811213548574e-08,
+      "loss": 0.9111,
+      "step": 37377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.317431694093263e-08,
+      "loss": 0.8659,
+      "step": 37378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.313059437168418e-08,
+      "loss": 0.6035,
+      "step": 37379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3086944428060132e-08,
+      "loss": 0.7757,
+      "step": 37380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3043367110375792e-08,
+      "loss": 0.692,
+      "step": 37381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2999862418949793e-08,
+      "loss": 1.1472,
+      "step": 37382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2956430354098549e-08,
+      "loss": 0.5665,
+      "step": 37383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.291307091613625e-08,
+      "loss": 0.784,
+      "step": 37384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2869784105379312e-08,
+      "loss": 0.9024,
+      "step": 37385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2826569922141929e-08,
+      "loss": 0.7336,
+      "step": 37386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2783428366739404e-08,
+      "loss": 0.6908,
+      "step": 37387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2740359439484817e-08,
+      "loss": 0.6776,
+      "step": 37388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2697363140691254e-08,
+      "loss": 0.8083,
+      "step": 37389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2654439470671798e-08,
+      "loss": 0.7032,
+      "step": 37390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2611588429737309e-08,
+      "loss": 0.97,
+      "step": 37391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2568810018201983e-08,
+      "loss": 0.6386,
+      "step": 37392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.252610423637335e-08,
+      "loss": 0.9599,
+      "step": 37393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2483471084565602e-08,
+      "loss": 0.8658,
+      "step": 37394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2440910563086273e-08,
+      "loss": 0.674,
+      "step": 37395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2398422672246225e-08,
+      "loss": 0.7823,
+      "step": 37396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2356007412354099e-08,
+      "loss": 0.7005,
+      "step": 37397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2313664783717427e-08,
+      "loss": 1.1206,
+      "step": 37398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2271394786645963e-08,
+      "loss": 0.8603,
+      "step": 37399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2229197421445016e-08,
+      "loss": 0.9311,
+      "step": 37400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2187072688423229e-08,
+      "loss": 0.6311,
+      "step": 37401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2145020587885914e-08,
+      "loss": 0.8399,
+      "step": 37402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2103041120139491e-08,
+      "loss": 1.0002,
+      "step": 37403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.2061134285488163e-08,
+      "loss": 0.7437,
+      "step": 37404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.201930008423724e-08,
+      "loss": 0.8016,
+      "step": 37405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1977538516692033e-08,
+      "loss": 0.8372,
+      "step": 37406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1935849583154523e-08,
+      "loss": 0.9755,
+      "step": 37407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1894233283928913e-08,
+      "loss": 0.7113,
+      "step": 37408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.1852689619318291e-08,
+      "loss": 0.803,
+      "step": 37409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.181121858962353e-08,
+      "loss": 0.9152,
+      "step": 37410
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1769820195145498e-08,
+      "loss": 0.8351,
+      "step": 37411
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1728494436188399e-08,
+      "loss": 0.6808,
+      "step": 37412
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1687241313049768e-08,
+      "loss": 0.5554,
+      "step": 37413
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.164606082603159e-08,
+      "loss": 0.8144,
+      "step": 37414
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1604952975431404e-08,
+      "loss": 0.8932,
+      "step": 37415
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1563917761550081e-08,
+      "loss": 0.7327,
+      "step": 37416
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1522955184684048e-08,
+      "loss": 0.4917,
+      "step": 37417
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1482065245133068e-08,
+      "loss": 0.7348,
+      "step": 37418
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1441247943193568e-08,
+      "loss": 0.7632,
+      "step": 37419
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1400503279163088e-08,
+      "loss": 0.721,
+      "step": 37420
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.135983125333695e-08,
+      "loss": 0.5727,
+      "step": 37421
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1319231866010472e-08,
+      "loss": 0.7729,
+      "step": 37422
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1278705117481192e-08,
+      "loss": 0.6633,
+      "step": 37423
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.12382510080411e-08,
+      "loss": 0.7571,
+      "step": 37424
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1197869537986627e-08,
+      "loss": 0.8617,
+      "step": 37425
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1157560707609761e-08,
+      "loss": 0.8075,
+      "step": 37426
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1117324517204709e-08,
+      "loss": 0.8695,
+      "step": 37427
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.107716096706346e-08,
+      "loss": 1.067,
+      "step": 37428
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.1037070057479115e-08,
+      "loss": 0.7646,
+      "step": 37429
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0997051788741442e-08,
+      "loss": 1.0727,
+      "step": 37430
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0957106161143537e-08,
+      "loss": 0.7564,
+      "step": 37431
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0917233174974062e-08,
+      "loss": 0.7528,
+      "step": 37432
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0877432830523894e-08,
+      "loss": 0.5886,
+      "step": 37433
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0837705128081688e-08,
+      "loss": 0.933,
+      "step": 37434
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0798050067938325e-08,
+      "loss": 0.7276,
+      "step": 37435
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0758467650379134e-08,
+      "loss": 0.6939,
+      "step": 37436
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0718957875694991e-08,
+      "loss": 0.9549,
+      "step": 37437
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0679520744171223e-08,
+      "loss": 0.8259,
+      "step": 37438
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.064015625609538e-08,
+      "loss": 0.7652,
+      "step": 37439
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0600864411753897e-08,
+      "loss": 1.0174,
+      "step": 37440
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0561645211432103e-08,
+      "loss": 1.0078,
+      "step": 37441
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0522498655415326e-08,
+      "loss": 0.7871,
+      "step": 37442
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0483424743987781e-08,
+      "loss": 0.9261,
+      "step": 37443
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0444423477434796e-08,
+      "loss": 0.9185,
+      "step": 37444
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0405494856038366e-08,
+      "loss": 0.938,
+      "step": 37445
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0366638880083824e-08,
+      "loss": 0.8302,
+      "step": 37446
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.032785554985094e-08,
+      "loss": 0.5786,
+      "step": 37447
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0289144865623934e-08,
+      "loss": 0.9542,
+      "step": 37448
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0250506827683692e-08,
+      "loss": 0.5312,
+      "step": 37449
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0211941436311102e-08,
+      "loss": 0.5257,
+      "step": 37450
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0173448691787046e-08,
+      "loss": 0.7486,
+      "step": 37451
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0135028594390195e-08,
+      "loss": 0.6694,
+      "step": 37452
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0096681144401433e-08,
+      "loss": 0.6272,
+      "step": 37453
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0058406342098315e-08,
+      "loss": 0.7767,
+      "step": 37454
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0020204187759507e-08,
+      "loss": 0.7345,
+      "step": 37455
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.982074681664788e-09,
+      "loss": 0.7389,
+      "step": 37456
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.944017824088381e-09,
+      "loss": 0.9966,
+      "step": 37457
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.906033615308953e-09,
+      "loss": 0.961,
+      "step": 37458
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.868122055601837e-09,
+      "loss": 0.9154,
+      "step": 37459
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.83028314524348e-09,
+      "loss": 0.6369,
+      "step": 37460
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.792516884508107e-09,
+      "loss": 0.7629,
+      "step": 37461
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.754823273671055e-09,
+      "loss": 0.8881,
+      "step": 37462
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.717202313006547e-09,
+      "loss": 0.9422,
+      "step": 37463
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.679654002788808e-09,
+      "loss": 0.5894,
+      "step": 37464
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.642178343288732e-09,
+      "loss": 0.7121,
+      "step": 37465
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.604775334780548e-09,
+      "loss": 0.7671,
+      "step": 37466
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.567444977536256e-09,
+      "loss": 0.917,
+      "step": 37467
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.530187271826751e-09,
+      "loss": 0.9933,
+      "step": 37468
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.493002217922931e-09,
+      "loss": 0.7933,
+      "step": 37469
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.455889816095687e-09,
+      "loss": 0.7772,
+      "step": 37470
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.418850066613693e-09,
+      "loss": 0.9762,
+      "step": 37471
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.381882969747846e-09,
+      "loss": 0.6521,
+      "step": 37472
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.344988525766818e-09,
+      "loss": 0.9374,
+      "step": 37473
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.308166734938173e-09,
+      "loss": 0.7529,
+      "step": 37474
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.271417597530586e-09,
+      "loss": 0.6661,
+      "step": 37475
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.234741113809398e-09,
+      "loss": 0.8408,
+      "step": 37476
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.198137284043285e-09,
+      "loss": 0.9275,
+      "step": 37477
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.16160610849759e-09,
+      "loss": 0.6921,
+      "step": 37478
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.125147587438766e-09,
+      "loss": 1.1087,
+      "step": 37479
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.088761721131045e-09,
+      "loss": 0.8255,
+      "step": 37480
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.052448509838663e-09,
+      "loss": 0.9048,
+      "step": 37481
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.016207953826961e-09,
+      "loss": 0.6697,
+      "step": 37482
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.980040053359063e-09,
+      "loss": 0.8347,
+      "step": 37483
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.943944808698091e-09,
+      "loss": 0.7121,
+      "step": 37484
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.907922220104947e-09,
+      "loss": 0.8318,
+      "step": 37485
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.871972287843866e-09,
+      "loss": 1.1622,
+      "step": 37486
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.836095012174639e-09,
+      "loss": 0.8384,
+      "step": 37487
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.80029039335928e-09,
+      "loss": 0.7532,
+      "step": 37488
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.764558431657578e-09,
+      "loss": 1.1084,
+      "step": 37489
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.728899127329327e-09,
+      "loss": 0.7056,
+      "step": 37490
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.69331248063432e-09,
+      "loss": 0.726,
+      "step": 37491
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.657798491831238e-09,
+      "loss": 0.8385,
+      "step": 37492
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.622357161177652e-09,
+      "loss": 0.9302,
+      "step": 37493
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.586988488932246e-09,
+      "loss": 0.811,
+      "step": 37494
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.551692475351481e-09,
+      "loss": 0.5827,
+      "step": 37495
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.516469120692928e-09,
+      "loss": 0.6668,
+      "step": 37496
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.481318425210828e-09,
+      "loss": 0.8489,
+      "step": 37497
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.446240389163862e-09,
+      "loss": 0.8129,
+      "step": 37498
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.411235012804053e-09,
+      "loss": 0.7645,
+      "step": 37499
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.376302296387862e-09,
+      "loss": 0.6798,
+      "step": 37500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.341442240168419e-09,
+      "loss": 0.7677,
+      "step": 37501
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.306654844399964e-09,
+      "loss": 0.8875,
+      "step": 37502
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.27194010933452e-09,
+      "loss": 0.8272,
+      "step": 37503
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.237298035225217e-09,
+      "loss": 0.523,
+      "step": 37504
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.202728622324075e-09,
+      "loss": 0.8254,
+      "step": 37505
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.168231870882004e-09,
+      "loss": 0.5659,
+      "step": 37506
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.133807781149916e-09,
+      "loss": 0.9344,
+      "step": 37507
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.099456353378721e-09,
+      "loss": 0.9307,
+      "step": 37508
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.065177587818219e-09,
+      "loss": 0.7953,
+      "step": 37509
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 8.030971484717099e-09,
+      "loss": 0.7073,
+      "step": 37510
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.996838044324051e-09,
+      "loss": 1.0286,
+      "step": 37511
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.962777266887767e-09,
+      "loss": 0.6952,
+      "step": 37512
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.928789152655825e-09,
+      "loss": 0.8983,
+      "step": 37513
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.894873701874694e-09,
+      "loss": 0.8301,
+      "step": 37514
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.861030914791956e-09,
+      "loss": 0.6413,
+      "step": 37515
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.827260791654078e-09,
+      "loss": 0.5764,
+      "step": 37516
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.793563332705312e-09,
+      "loss": 0.6774,
+      "step": 37517
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.759938538191015e-09,
+      "loss": 0.7698,
+      "step": 37518
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.726386408356546e-09,
+      "loss": 0.7451,
+      "step": 37519
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.692906943446154e-09,
+      "loss": 0.7648,
+      "step": 37520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.65950014370187e-09,
+      "loss": 0.5956,
+      "step": 37521
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.626166009366831e-09,
+      "loss": 0.7462,
+      "step": 37522
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.592904540684177e-09,
+      "loss": 0.5974,
+      "step": 37523
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.559715737895935e-09,
+      "loss": 1.0252,
+      "step": 37524
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.526599601243022e-09,
+      "loss": 0.67,
+      "step": 37525
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.493556130966361e-09,
+      "loss": 0.9562,
+      "step": 37526
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.460585327305758e-09,
+      "loss": 0.801,
+      "step": 37527
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.427687190502131e-09,
+      "loss": 0.9295,
+      "step": 37528
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.394861720793067e-09,
+      "loss": 0.8857,
+      "step": 37529
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.362108918418376e-09,
+      "loss": 0.6937,
+      "step": 37530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.329428783616754e-09,
+      "loss": 0.9391,
+      "step": 37531
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.29682131662468e-09,
+      "loss": 0.643,
+      "step": 37532
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.2642865176797416e-09,
+      "loss": 0.9477,
+      "step": 37533
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.231824387019526e-09,
+      "loss": 0.6884,
+      "step": 37534
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.19943492487829e-09,
+      "loss": 0.7988,
+      "step": 37535
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.167118131492512e-09,
+      "loss": 1.0288,
+      "step": 37536
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.134874007097559e-09,
+      "loss": 1.0233,
+      "step": 37537
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.102702551926577e-09,
+      "loss": 0.8213,
+      "step": 37538
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.070603766214934e-09,
+      "loss": 0.9246,
+      "step": 37539
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.038577650195777e-09,
+      "loss": 0.3596,
+      "step": 37540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 7.006624204102252e-09,
+      "loss": 1.0523,
+      "step": 37541
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.974743428165287e-09,
+      "loss": 0.9013,
+      "step": 37542
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.942935322618028e-09,
+      "loss": 0.713,
+      "step": 37543
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.9111998876914e-09,
+      "loss": 0.8246,
+      "step": 37544
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.8795371236163315e-09,
+      "loss": 0.4988,
+      "step": 37545
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.847947030622637e-09,
+      "loss": 0.5089,
+      "step": 37546
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.816429608941244e-09,
+      "loss": 0.8126,
+      "step": 37547
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.784984858799748e-09,
+      "loss": 0.524,
+      "step": 37548
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.753612780426855e-09,
+      "loss": 0.7191,
+      "step": 37549
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.72231337405238e-09,
+      "loss": 0.8461,
+      "step": 37550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.691086639902811e-09,
+      "loss": 0.776,
+      "step": 37551
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.659932578205741e-09,
+      "loss": 0.833,
+      "step": 37552
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.628851189186547e-09,
+      "loss": 0.5714,
+      "step": 37553
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.597842473072824e-09,
+      "loss": 0.7512,
+      "step": 37554
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.566906430089947e-09,
+      "loss": 0.9443,
+      "step": 37555
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.536043060461072e-09,
+      "loss": 0.8597,
+      "step": 37556
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.505252364411574e-09,
+      "loss": 1.039,
+      "step": 37557
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.474534342166827e-09,
+      "loss": 0.7785,
+      "step": 37558
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.443888993948877e-09,
+      "loss": 0.8047,
+      "step": 37559
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.413316319979768e-09,
+      "loss": 0.689,
+      "step": 37560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.382816320483765e-09,
+      "loss": 0.7568,
+      "step": 37561
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.3523889956806915e-09,
+      "loss": 0.75,
+      "step": 37562
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.3220343457937035e-09,
+      "loss": 0.5593,
+      "step": 37563
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.291752371042625e-09,
+      "loss": 1.0441,
+      "step": 37564
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.26154307164728e-09,
+      "loss": 1.2193,
+      "step": 37565
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.231406447827493e-09,
+      "loss": 0.9255,
+      "step": 37566
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.201342499803087e-09,
+      "loss": 0.7644,
+      "step": 37567
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.171351227792777e-09,
+      "loss": 0.525,
+      "step": 37568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.141432632013056e-09,
+      "loss": 0.8382,
+      "step": 37569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.111586712683748e-09,
+      "loss": 0.5867,
+      "step": 37570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.081813470020237e-09,
+      "loss": 0.5579,
+      "step": 37571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.052112904239016e-09,
+      "loss": 0.8854,
+      "step": 37572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.022485015557688e-09,
+      "loss": 0.9219,
+      "step": 37573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.992929804190528e-09,
+      "loss": 0.8931,
+      "step": 37574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.9634472703518075e-09,
+      "loss": 0.6235,
+      "step": 37575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.93403741425802e-09,
+      "loss": 0.721,
+      "step": 37576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.904700236120109e-09,
+      "loss": 0.6955,
+      "step": 37577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.875435736154567e-09,
+      "loss": 0.7034,
+      "step": 37578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.846243914571226e-09,
+      "loss": 0.9718,
+      "step": 37579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.817124771585469e-09,
+      "loss": 0.5949,
+      "step": 37580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.78807830740602e-09,
+      "loss": 1.0382,
+      "step": 37581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.75910452224604e-09,
+      "loss": 0.9182,
+      "step": 37582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.730203416315361e-09,
+      "loss": 0.7073,
+      "step": 37583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.701374989824926e-09,
+      "loss": 0.9409,
+      "step": 37584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.672619242983457e-09,
+      "loss": 0.6921,
+      "step": 37585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.643936176000786e-09,
+      "loss": 0.9668,
+      "step": 37586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.615325789085635e-09,
+      "loss": 1.1977,
+      "step": 37587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.586788082445615e-09,
+      "loss": 0.8413,
+      "step": 37588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.558323056287229e-09,
+      "loss": 0.7898,
+      "step": 37589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.529930710820308e-09,
+      "loss": 0.7145,
+      "step": 37590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.501611046248023e-09,
+      "loss": 0.8572,
+      "step": 37591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.473364062779097e-09,
+      "loss": 1.0314,
+      "step": 37592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.445189760616698e-09,
+      "loss": 0.7383,
+      "step": 37593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.4170881399673304e-09,
+      "loss": 0.7596,
+      "step": 37594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.389059201034164e-09,
+      "loss": 0.8968,
+      "step": 37595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.361102944022589e-09,
+      "loss": 0.5745,
+      "step": 37596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.333219369134668e-09,
+      "loss": 0.6585,
+      "step": 37597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.305408476573571e-09,
+      "loss": 0.9352,
+      "step": 37598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.277670266541357e-09,
+      "loss": 0.9555,
+      "step": 37599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.250004739238979e-09,
+      "loss": 1.0945,
+      "step": 37600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.222411894869606e-09,
+      "loss": 0.7573,
+      "step": 37601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.194891733631968e-09,
+      "loss": 0.8794,
+      "step": 37602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.167444255727017e-09,
+      "loss": 0.777,
+      "step": 37603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.140069461354591e-09,
+      "loss": 0.6421,
+      "step": 37604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.112767350713421e-09,
+      "loss": 0.6148,
+      "step": 37605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.085537924002238e-09,
+      "loss": 0.7017,
+      "step": 37606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.058381181418659e-09,
+      "loss": 0.8324,
+      "step": 37607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.031297123160306e-09,
+      "loss": 1.0694,
+      "step": 37608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 5.004285749424797e-09,
+      "loss": 0.7899,
+      "step": 37609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.977347060407534e-09,
+      "loss": 0.7016,
+      "step": 37610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.950481056306134e-09,
+      "loss": 0.9854,
+      "step": 37611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.923687737312665e-09,
+      "loss": 0.7932,
+      "step": 37612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.89696710362586e-09,
+      "loss": 0.7558,
+      "step": 37613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.870319155437786e-09,
+      "loss": 0.8024,
+      "step": 37614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.843743892942731e-09,
+      "loss": 0.9814,
+      "step": 37615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.817241316332766e-09,
+      "loss": 0.9071,
+      "step": 37616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.790811425803288e-09,
+      "loss": 0.7531,
+      "step": 37617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.764454221544146e-09,
+      "loss": 0.6272,
+      "step": 37618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.738169703748519e-09,
+      "loss": 0.9187,
+      "step": 37619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.711957872606254e-09,
+      "loss": 1.0708,
+      "step": 37620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.685818728308311e-09,
+      "loss": 0.8153,
+      "step": 37621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.6597522710456474e-09,
+      "loss": 0.7217,
+      "step": 37622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.633758501007002e-09,
+      "loss": 0.5703,
+      "step": 37623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.607837418381111e-09,
+      "loss": 0.6608,
+      "step": 37624
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.581989023357824e-09,
+      "loss": 0.7608,
+      "step": 37625
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.556213316124769e-09,
+      "loss": 1.0602,
+      "step": 37626
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.530510296867352e-09,
+      "loss": 0.5804,
+      "step": 37627
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.5048799657754214e-09,
+      "loss": 0.8347,
+      "step": 37628
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.479322323033275e-09,
+      "loss": 0.8547,
+      "step": 37629
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.45383736882743e-09,
+      "loss": 0.6744,
+      "step": 37630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.428425103343293e-09,
+      "loss": 0.7268,
+      "step": 37631
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.4030855267651606e-09,
+      "loss": 0.9031,
+      "step": 37632
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.377818639278442e-09,
+      "loss": 0.8825,
+      "step": 37633
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.352624441066322e-09,
+      "loss": 0.8066,
+      "step": 37634
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.327502932311989e-09,
+      "loss": 0.5803,
+      "step": 37635
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.302454113197518e-09,
+      "loss": 1.0822,
+      "step": 37636
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.277477983906098e-09,
+      "loss": 0.578,
+      "step": 37637
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.252574544618693e-09,
+      "loss": 0.8806,
+      "step": 37638
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.22774379551627e-09,
+      "loss": 1.0188,
+      "step": 37639
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.202985736779797e-09,
+      "loss": 0.692,
+      "step": 37640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.178300368589127e-09,
+      "loss": 0.6838,
+      "step": 37641
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.15368769112412e-09,
+      "loss": 0.814,
+      "step": 37642
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.129147704563518e-09,
+      "loss": 0.7935,
+      "step": 37643
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.10468040908607e-09,
+      "loss": 0.8971,
+      "step": 37644
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0802858048683004e-09,
+      "loss": 0.5588,
+      "step": 37645
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.055963892090065e-09,
+      "loss": 0.7018,
+      "step": 37646
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.031714670925668e-09,
+      "loss": 0.7744,
+      "step": 37647
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 4.0075381415538575e-09,
+      "loss": 0.8822,
+      "step": 37648
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.9834343041489364e-09,
+      "loss": 1.345,
+      "step": 37649
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.959403158885211e-09,
+      "loss": 0.8951,
+      "step": 37650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.935444705940317e-09,
+      "loss": 0.5996,
+      "step": 37651
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.911558945486338e-09,
+      "loss": 0.8289,
+      "step": 37652
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.887745877697579e-09,
+      "loss": 0.791,
+      "step": 37653
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.864005502747236e-09,
+      "loss": 0.568,
+      "step": 37654
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8403378208073934e-09,
+      "loss": 0.7061,
+      "step": 37655
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.8167428320512456e-09,
+      "loss": 0.7772,
+      "step": 37656
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.793220536649767e-09,
+      "loss": 0.8718,
+      "step": 37657
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.769770934773931e-09,
+      "loss": 0.6356,
+      "step": 37658
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.746394026594713e-09,
+      "loss": 0.5771,
+      "step": 37659
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.7230898122808666e-09,
+      "loss": 0.6715,
+      "step": 37660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6998582920033666e-09,
+      "loss": 1.0195,
+      "step": 37661
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6766994659309665e-09,
+      "loss": 0.8075,
+      "step": 37662
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.6536133342313096e-09,
+      "loss": 0.7616,
+      "step": 37663
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.63059989707204e-09,
+      "loss": 0.4648,
+      "step": 37664
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.607659154621912e-09,
+      "loss": 0.576,
+      "step": 37665
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5847911070463483e-09,
+      "loss": 0.8799,
+      "step": 37666
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5619957545129926e-09,
+      "loss": 0.5368,
+      "step": 37667
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.539273097186158e-09,
+      "loss": 0.5989,
+      "step": 37668
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5166231352323776e-09,
+      "loss": 0.7956,
+      "step": 37669
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.494045868814855e-09,
+      "loss": 0.8709,
+      "step": 37670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4715412981001227e-09,
+      "loss": 0.8206,
+      "step": 37671
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.449109423249164e-09,
+      "loss": 0.9225,
+      "step": 37672
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4267502444274013e-09,
+      "loss": 0.5211,
+      "step": 37673
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.4044637617958176e-09,
+      "loss": 0.8626,
+      "step": 37674
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.382249975516505e-09,
+      "loss": 0.6385,
+      "step": 37675
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3601088857526663e-09,
+      "loss": 1.0633,
+      "step": 37676
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.338040492664174e-09,
+      "loss": 0.727,
+      "step": 37677
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.3160447964109e-09,
+      "loss": 0.5934,
+      "step": 37678
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2941217971538266e-09,
+      "loss": 0.8392,
+      "step": 37679
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2722714950517154e-09,
+      "loss": 0.8605,
+      "step": 37680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.250493890264439e-09,
+      "loss": 1.0927,
+      "step": 37681
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2287889829485386e-09,
+      "loss": 0.9318,
+      "step": 37682
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.2071567732638866e-09,
+      "loss": 0.7441,
+      "step": 37683
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.185597261367024e-09,
+      "loss": 0.8736,
+      "step": 37684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.164110447414492e-09,
+      "loss": 0.509,
+      "step": 37685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.142696331561723e-09,
+      "loss": 0.8316,
+      "step": 37686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1213549139663678e-09,
+      "loss": 0.7427,
+      "step": 37687
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.1000861947827475e-09,
+      "loss": 1.1194,
+      "step": 37688
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0788901741651832e-09,
+      "loss": 0.9345,
+      "step": 37689
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0577668522679958e-09,
+      "loss": 0.7225,
+      "step": 37690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0367162292443964e-09,
+      "loss": 0.4323,
+      "step": 37691
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.0157383052487055e-09,
+      "loss": 0.9165,
+      "step": 37692
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9948330804319138e-09,
+      "loss": 0.8125,
+      "step": 37693
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9740005549472316e-09,
+      "loss": 0.954,
+      "step": 37694
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.95324072894565e-09,
+      "loss": 0.7628,
+      "step": 37695
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9325536025792688e-09,
+      "loss": 0.6383,
+      "step": 37696
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.9119391759957484e-09,
+      "loss": 1.0331,
+      "step": 37697
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.891397449348299e-09,
+      "loss": 0.8034,
+      "step": 37698
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.87092842278347e-09,
+      "loss": 0.8004,
+      "step": 37699
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.850532096452252e-09,
+      "loss": 0.5852,
+      "step": 37700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.830208470502305e-09,
+      "loss": 1.0124,
+      "step": 37701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.8099575450801776e-09,
+      "loss": 0.8781,
+      "step": 37702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7897793203357503e-09,
+      "loss": 0.7148,
+      "step": 37703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7696737964133523e-09,
+      "loss": 0.7021,
+      "step": 37704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.749640973459533e-09,
+      "loss": 0.6798,
+      "step": 37705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7296808516208417e-09,
+      "loss": 0.8339,
+      "step": 37706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.7097934310427176e-09,
+      "loss": 0.5732,
+      "step": 37707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6899787118683794e-09,
+      "loss": 0.8566,
+      "step": 37708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6702366942432666e-09,
+      "loss": 0.5865,
+      "step": 37709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6505673783094875e-09,
+      "loss": 0.982,
+      "step": 37710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6309707642113714e-09,
+      "loss": 0.9008,
+      "step": 37711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.6114468520921366e-09,
+      "loss": 0.7392,
+      "step": 37712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5919956420916715e-09,
+      "loss": 0.6027,
+      "step": 37713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5726171343520846e-09,
+      "loss": 0.9139,
+      "step": 37714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5533113290154843e-09,
+      "loss": 0.9624,
+      "step": 37715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.5340782262206488e-09,
+      "loss": 0.7765,
+      "step": 37716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.514917826108576e-09,
+      "loss": 0.8741,
+      "step": 37717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.495830128818044e-09,
+      "loss": 0.8845,
+      "step": 37718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.476815134487831e-09,
+      "loss": 0.8773,
+      "step": 37719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4578728432567144e-09,
+      "loss": 0.8211,
+      "step": 37720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4390032552623623e-09,
+      "loss": 0.7975,
+      "step": 37721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.4202063706413313e-09,
+      "loss": 0.8582,
+      "step": 37722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.40148218953018e-09,
+      "loss": 1.0613,
+      "step": 37723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.382830712066575e-09,
+      "loss": 0.6528,
+      "step": 37724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.364251938384854e-09,
+      "loss": 0.8391,
+      "step": 37725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.345745868621574e-09,
+      "loss": 0.9774,
+      "step": 37726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3273125029088516e-09,
+      "loss": 0.8408,
+      "step": 37727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.3089518413832444e-09,
+      "loss": 0.9876,
+      "step": 37728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.290663884175759e-09,
+      "loss": 1.2108,
+      "step": 37729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.272448631421842e-09,
+      "loss": 0.8292,
+      "step": 37730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2543060832525e-09,
+      "loss": 0.8922,
+      "step": 37731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.2362362398009596e-09,
+      "loss": 0.8766,
+      "step": 37732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.218239101196007e-09,
+      "loss": 0.8126,
+      "step": 37733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.200314667571979e-09,
+      "loss": 0.865,
+      "step": 37734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.182462939056551e-09,
+      "loss": 0.6983,
+      "step": 37735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1646839157807296e-09,
+      "loss": 0.9511,
+      "step": 37736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.146977597873301e-09,
+      "loss": 0.7967,
+      "step": 37737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1293439854641607e-09,
+      "loss": 0.7556,
+      "step": 37738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.1117830786798743e-09,
+      "loss": 0.8355,
+      "step": 37739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0942948776481175e-09,
+      "loss": 0.9408,
+      "step": 37740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0768793824987866e-09,
+      "loss": 0.511,
+      "step": 37741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0595365933551157e-09,
+      "loss": 0.7463,
+      "step": 37742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0422665103458916e-09,
+      "loss": 0.6611,
+      "step": 37743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0250691335943483e-09,
+      "loss": 0.8873,
+      "step": 37744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2.0079444632270516e-09,
+      "loss": 0.9354,
+      "step": 37745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9908924993694567e-09,
+      "loss": 0.621,
+      "step": 37746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.973913242142578e-09,
+      "loss": 0.855,
+      "step": 37747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9570066916729802e-09,
+      "loss": 0.7905,
+      "step": 37748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.9401728480816784e-09,
+      "loss": 0.8161,
+      "step": 37749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.923411711491907e-09,
+      "loss": 0.7276,
+      "step": 37750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.906723282025791e-09,
+      "loss": 0.71,
+      "step": 37751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8901075598032337e-09,
+      "loss": 0.793,
+      "step": 37752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.87356454494636e-09,
+      "loss": 0.8724,
+      "step": 37753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8570942375761847e-09,
+      "loss": 0.8749,
+      "step": 37754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8406966378103909e-09,
+      "loss": 0.7826,
+      "step": 37755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8243717457699929e-09,
+      "loss": 0.7612,
+      "step": 37756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.8081195615726743e-09,
+      "loss": 0.7543,
+      "step": 37757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.791940085336119e-09,
+      "loss": 0.684,
+      "step": 37758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7758333171802312e-09,
+      "loss": 0.5522,
+      "step": 37759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7597992572204737e-09,
+      "loss": 0.7983,
+      "step": 37760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7438379055734201e-09,
+      "loss": 0.9142,
+      "step": 37761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7279492623556437e-09,
+      "loss": 0.7759,
+      "step": 37762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.712133327682608e-09,
+      "loss": 0.6906,
+      "step": 37763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6963901016686657e-09,
+      "loss": 0.5365,
+      "step": 37764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.68071958442928e-09,
+      "loss": 0.9134,
+      "step": 37765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6651217760776938e-09,
+      "loss": 0.7098,
+      "step": 37766
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.64959667672826e-09,
+      "loss": 0.7222,
+      "step": 37767
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6341442864931111e-09,
+      "loss": 0.561,
+      "step": 37768
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6187646054843798e-09,
+      "loss": 0.5812,
+      "step": 37769
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.6034576338141982e-09,
+      "loss": 0.7054,
+      "step": 37770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5882233715935891e-09,
+      "loss": 0.8899,
+      "step": 37771
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5730618189346846e-09,
+      "loss": 0.716,
+      "step": 37772
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.557972975946287e-09,
+      "loss": 0.7219,
+      "step": 37773
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.542956842739418e-09,
+      "loss": 0.9156,
+      "step": 37774
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5280134194217699e-09,
+      "loss": 1.0105,
+      "step": 37775
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.5131427061021443e-09,
+      "loss": 0.941,
+      "step": 37776
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.498344702890453e-09,
+      "loss": 0.5912,
+      "step": 37777
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4836194098921674e-09,
+      "loss": 0.7024,
+      "step": 37778
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4689668272149793e-09,
+      "loss": 0.78,
+      "step": 37779
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4543869549665802e-09,
+      "loss": 0.892,
+      "step": 37780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4398797932524412e-09,
+      "loss": 0.7085,
+      "step": 37781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4254453421769233e-09,
+      "loss": 0.6025,
+      "step": 37782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.4110836018466079e-09,
+      "loss": 0.7709,
+      "step": 37783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3967945723647458e-09,
+      "loss": 0.5535,
+      "step": 37784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3825782538368083e-09,
+      "loss": 0.7724,
+      "step": 37785
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3684346463638253e-09,
+      "loss": 1.0404,
+      "step": 37786
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3543637500501584e-09,
+      "loss": 0.8903,
+      "step": 37787
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3403655649979475e-09,
+      "loss": 0.4184,
+      "step": 37788
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3264400913093334e-09,
+      "loss": 0.6435,
+      "step": 37789
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3125873290842362e-09,
+      "loss": 0.6495,
+      "step": 37790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2988072784247963e-09,
+      "loss": 0.8135,
+      "step": 37791
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2850999394309338e-09,
+      "loss": 1.1249,
+      "step": 37792
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.271465312202569e-09,
+      "loss": 0.6985,
+      "step": 37793
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2579033968385112e-09,
+      "loss": 0.7938,
+      "step": 37794
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2444141934364605e-09,
+      "loss": 0.8341,
+      "step": 37795
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2309977020952269e-09,
+      "loss": 0.8458,
+      "step": 37796
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2176539229136198e-09,
+      "loss": 0.6662,
+      "step": 37797
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2043828559860083e-09,
+      "loss": 0.7,
+      "step": 37798
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1911845014123125e-09,
+      "loss": 0.8609,
+      "step": 37799
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.178058859285791e-09,
+      "loss": 0.6404,
+      "step": 37800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1650059297019234e-09,
+      "loss": 0.6962,
+      "step": 37801
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1520257127572987e-09,
+      "loss": 0.7433,
+      "step": 37802
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1391182085451758e-09,
+      "loss": 0.5671,
+      "step": 37803
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1262834171610337e-09,
+      "loss": 0.8105,
+      "step": 37804
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.113521338695911e-09,
+      "loss": 0.7713,
+      "step": 37805
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1008319732430661e-09,
+      "loss": 0.6815,
+      "step": 37806
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.088215320895758e-09,
+      "loss": 0.8327,
+      "step": 37807
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0756713817450248e-09,
+      "loss": 0.5112,
+      "step": 37808
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.063200155883015e-09,
+      "loss": 0.9283,
+      "step": 37809
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.050801643399657e-09,
+      "loss": 0.7309,
+      "step": 37810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0384758443848786e-09,
+      "loss": 0.7248,
+      "step": 37811
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0262227589286078e-09,
+      "loss": 0.8087,
+      "step": 37812
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0140423871207727e-09,
+      "loss": 0.6469,
+      "step": 37813
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0019347290479708e-09,
+      "loss": 0.9519,
+      "step": 37814
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.898997848001302e-10,
+      "loss": 0.8506,
+      "step": 37815
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.779375544649583e-10,
+      "loss": 0.688,
+      "step": 37816
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.660480381279424e-10,
+      "loss": 0.8934,
+      "step": 37817
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.5423123587568e-10,
+      "loss": 0.92,
+      "step": 37818
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.424871477947684e-10,
+      "loss": 0.6756,
+      "step": 37819
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.308157739706947e-10,
+      "loss": 0.6829,
+      "step": 37820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.19217114487836e-10,
+      "loss": 0.7175,
+      "step": 37821
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.07691169430569e-10,
+      "loss": 1.0199,
+      "step": 37822
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.962379388832709e-10,
+      "loss": 0.6598,
+      "step": 37823
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.848574229292084e-10,
+      "loss": 0.7603,
+      "step": 37824
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.735496216505379e-10,
+      "loss": 0.3972,
+      "step": 37825
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.623145351305262e-10,
+      "loss": 0.7544,
+      "step": 37826
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.511521634491094e-10,
+      "loss": 0.8526,
+      "step": 37827
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.400625066884438e-10,
+      "loss": 0.6501,
+      "step": 37828
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.290455649306861e-10,
+      "loss": 1.22,
+      "step": 37829
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.181013382524416e-10,
+      "loss": 1.1102,
+      "step": 37830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.072298267369772e-10,
+      "loss": 0.8741,
+      "step": 37831
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.96431030460898e-10,
+      "loss": 1.0078,
+      "step": 37832
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.857049495041402e-10,
+      "loss": 0.8224,
+      "step": 37833
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.750515839433093e-10,
+      "loss": 0.7558,
+      "step": 37834
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.644709338572309e-10,
+      "loss": 0.7427,
+      "step": 37835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.539629993225106e-10,
+      "loss": 0.6297,
+      "step": 37836
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.435277804146435e-10,
+      "loss": 0.7499,
+      "step": 37837
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.331652772113452e-10,
+      "loss": 0.8923,
+      "step": 37838
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.228754897858903e-10,
+      "loss": 0.8053,
+      "step": 37839
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.126584182137741e-10,
+      "loss": 0.8977,
+      "step": 37840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.025140625704919e-10,
+      "loss": 0.8398,
+      "step": 37841
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.924424229282079e-10,
+      "loss": 0.7825,
+      "step": 37842
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.824434993613071e-10,
+      "loss": 0.71,
+      "step": 37843
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.725172919419542e-10,
+      "loss": 1.0263,
+      "step": 37844
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.626638007434239e-10,
+      "loss": 0.7884,
+      "step": 37845
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.5288302583566e-10,
+      "loss": 0.8999,
+      "step": 37846
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.43174967289717e-10,
+      "loss": 0.6784,
+      "step": 37847
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.335396251777592e-10,
+      "loss": 0.8395,
+      "step": 37848
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.239769995697309e-10,
+      "loss": 0.7707,
+      "step": 37849
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.144870905333555e-10,
+      "loss": 0.7872,
+      "step": 37850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.050698981396874e-10,
+      "loss": 0.6526,
+      "step": 37851
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.957254224553399e-10,
+      "loss": 0.8789,
+      "step": 37852
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.864536635502572e-10,
+      "loss": 1.1029,
+      "step": 37853
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.772546214910524e-10,
+      "loss": 0.7245,
+      "step": 37854
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.681282963432289e-10,
+      "loss": 0.9971,
+      "step": 37855
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.590746881756204e-10,
+      "loss": 0.7757,
+      "step": 37856
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.500937970526199e-10,
+      "loss": 0.674,
+      "step": 37857
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.411856230386204e-10,
+      "loss": 0.9074,
+      "step": 37858
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.323501662013453e-10,
+      "loss": 0.7802,
+      "step": 37859
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.23587426601857e-10,
+      "loss": 0.8659,
+      "step": 37860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.148974043056587e-10,
+      "loss": 1.225,
+      "step": 37861
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.06280099376033e-10,
+      "loss": 0.7324,
+      "step": 37862
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.977355118740423e-10,
+      "loss": 0.6604,
+      "step": 37863
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.892636418640795e-10,
+      "loss": 1.031,
+      "step": 37864
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.808644894060965e-10,
+      "loss": 0.8834,
+      "step": 37865
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.725380545611558e-10,
+      "loss": 0.8708,
+      "step": 37866
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.6428433739142966e-10,
+      "loss": 0.6679,
+      "step": 37867
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.561033379546498e-10,
+      "loss": 0.7456,
+      "step": 37868
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.479950563129887e-10,
+      "loss": 0.6849,
+      "step": 37869
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.399594925230677e-10,
+      "loss": 0.9307,
+      "step": 37870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.3199664664372865e-10,
+      "loss": 0.7816,
+      "step": 37871
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2410651873381334e-10,
+      "loss": 0.7771,
+      "step": 37872
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.1628910885105345e-10,
+      "loss": 1.127,
+      "step": 37873
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.085444170498498e-10,
+      "loss": 0.8559,
+      "step": 37874
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.008724433890443e-10,
+      "loss": 0.6292,
+      "step": 37875
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.9327318792414804e-10,
+      "loss": 0.8003,
+      "step": 37876
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.8574665070845176e-10,
+      "loss": 0.9483,
+      "step": 37877
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.782928317985768e-10,
+      "loss": 0.6332,
+      "step": 37878
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.709117312489241e-10,
+      "loss": 0.5861,
+      "step": 37879
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.6360334911167414e-10,
+      "loss": 0.8202,
+      "step": 37880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.563676854412279e-10,
+      "loss": 0.9956,
+      "step": 37881
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.492047402886556e-10,
+      "loss": 1.1981,
+      "step": 37882
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.421145137083581e-10,
+      "loss": 1.0143,
+      "step": 37883
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.3509700574918535e-10,
+      "loss": 0.6799,
+      "step": 37884
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2815221646553817e-10,
+      "loss": 1.0138,
+      "step": 37885
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.21280145904046e-10,
+      "loss": 0.8144,
+      "step": 37886
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.144807941179995e-10,
+      "loss": 0.7502,
+      "step": 37887
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0775416115513826e-10,
+      "loss": 0.9114,
+      "step": 37888
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0110024706431207e-10,
+      "loss": 0.6905,
+      "step": 37889
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.94519051895481e-10,
+      "loss": 0.7505,
+      "step": 37890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8801057569416423e-10,
+      "loss": 0.9326,
+      "step": 37891
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.815748185092115e-10,
+      "loss": 0.6306,
+      "step": 37892
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.7521178038725225e-10,
+      "loss": 0.7923,
+      "step": 37893
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.689214613738056e-10,
+      "loss": 0.6066,
+      "step": 37894
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.627038615155009e-10,
+      "loss": 0.7598,
+      "step": 37895
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.565589808578572e-10,
+      "loss": 0.693,
+      "step": 37896
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.5048681944417343e-10,
+      "loss": 0.9248,
+      "step": 37897
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4448737731996854e-10,
+      "loss": 0.8363,
+      "step": 37898
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3856065452854125e-10,
+      "loss": 0.5911,
+      "step": 37899
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3270665111208013e-10,
+      "loss": 0.5951,
+      "step": 37900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.269253671138838e-10,
+      "loss": 0.9382,
+      "step": 37901
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2121680257614075e-10,
+      "loss": 1.1382,
+      "step": 37902
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1558095753992924e-10,
+      "loss": 0.7606,
+      "step": 37903
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.1001783204632752e-10,
+      "loss": 1.0323,
+      "step": 37904
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.0452742613641386e-10,
+      "loss": 1.0538,
+      "step": 37905
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.9910973985015623e-10,
+      "loss": 0.7252,
+      "step": 37906
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.937647732253023e-10,
+      "loss": 0.5711,
+      "step": 37907
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8849252630182002e-10,
+      "loss": 0.639,
+      "step": 37908
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.8329299911856724e-10,
+      "loss": 0.7363,
+      "step": 37909
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7816619171329154e-10,
+      "loss": 0.7292,
+      "step": 37910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7311210412263023e-10,
+      "loss": 0.9591,
+      "step": 37911
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6813073638322076e-10,
+      "loss": 0.5067,
+      "step": 37912
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.632220885317004e-10,
+      "loss": 0.8562,
+      "step": 37913
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5838616060359634e-10,
+      "loss": 0.852,
+      "step": 37914
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5362295263443572e-10,
+      "loss": 0.6781,
+      "step": 37915
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.489324646575252e-10,
+      "loss": 1.0245,
+      "step": 37916
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4431469670950216e-10,
+      "loss": 0.8013,
+      "step": 37917
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3976964882145283e-10,
+      "loss": 0.8784,
+      "step": 37918
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3529732102779413e-10,
+      "loss": 0.8729,
+      "step": 37919
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3089771336072256e-10,
+      "loss": 0.8096,
+      "step": 37920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2657082585243453e-10,
+      "loss": 0.8655,
+      "step": 37921
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.2231665853401632e-10,
+      "loss": 0.8952,
+      "step": 37922
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.1813521143655416e-10,
+      "loss": 0.7292,
+      "step": 37923
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.140264845911343e-10,
+      "loss": 0.7418,
+      "step": 37924
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0999047802551232e-10,
+      "loss": 0.9394,
+      "step": 37925
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.060271917718847e-10,
+      "loss": 1.2136,
+      "step": 37926
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0213662585689677e-10,
+      "loss": 0.9784,
+      "step": 37927
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.831878031052455e-11,
+      "loss": 0.8408,
+      "step": 37928
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.457365515830319e-11,
+      "loss": 0.788,
+      "step": 37929
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.09012504302087e-11,
+      "loss": 0.8914,
+      "step": 37930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.730156615066598e-11,
+      "loss": 0.7295,
+      "step": 37931
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.37746023463204e-11,
+      "loss": 0.8385,
+      "step": 37932
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.03203590438173e-11,
+      "loss": 0.674,
+      "step": 37933
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.693883626758159e-11,
+      "loss": 0.9702,
+      "step": 37934
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.363003404314839e-11,
+      "loss": 0.9955,
+      "step": 37935
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.039395239272218e-11,
+      "loss": 0.6403,
+      "step": 37936
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.723059134072785e-11,
+      "loss": 0.7278,
+      "step": 37937
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.41399509115903e-11,
+      "loss": 1.0326,
+      "step": 37938
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.112203112529358e-11,
+      "loss": 1.0571,
+      "step": 37939
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.817683200515234e-11,
+      "loss": 0.6879,
+      "step": 37940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.5304353572260826e-11,
+      "loss": 0.7228,
+      "step": 37941
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.2504595848823505e-11,
+      "loss": 0.7752,
+      "step": 37942
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.977755885260394e-11,
+      "loss": 1.0238,
+      "step": 37943
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.712324260580659e-11,
+      "loss": 0.8927,
+      "step": 37944
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.454164712619502e-11,
+      "loss": 0.9606,
+      "step": 37945
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.2032772433753254e-11,
+      "loss": 0.8193,
+      "step": 37946
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.959661854513463e-11,
+      "loss": 0.7384,
+      "step": 37947
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.7233185480323174e-11,
+      "loss": 0.8846,
+      "step": 37948
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.4942473253751777e-11,
+      "loss": 0.5173,
+      "step": 37949
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.272448188429422e-11,
+      "loss": 0.8264,
+      "step": 37950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.0579211387493646e-11,
+      "loss": 0.6276,
+      "step": 37951
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.8506661778893164e-11,
+      "loss": 0.6641,
+      "step": 37952
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.6506833074035897e-11,
+      "loss": 0.6795,
+      "step": 37953
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.4579725285134305e-11,
+      "loss": 0.728,
+      "step": 37954
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.2725338429951948e-11,
+      "loss": 0.8197,
+      "step": 37955
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.094367251848084e-11,
+      "loss": 0.6237,
+      "step": 37956
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.92347275662641e-11,
+      "loss": 0.7478,
+      "step": 37957
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7598503583293735e-11,
+      "loss": 1.0491,
+      "step": 37958
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6035000584002647e-11,
+      "loss": 0.808,
+      "step": 37959
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4544218578382841e-11,
+      "loss": 1.0583,
+      "step": 37960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3126157576426324e-11,
+      "loss": 0.8357,
+      "step": 37961
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.178081759034555e-11,
+      "loss": 0.8503,
+      "step": 37962
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0508198629022304e-11,
+      "loss": 0.8393,
+      "step": 37963
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.308300702448591e-12,
+      "loss": 0.8657,
+      "step": 37964
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 8.181123817285752e-12,
+      "loss": 0.8006,
+      "step": 37965
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 7.1266679846360156e-12,
+      "loss": 0.8049,
+      "step": 37966
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.1449332089402735e-12,
+      "loss": 0.6256,
+      "step": 37967
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.235919500190534e-12,
+      "loss": 0.6962,
+      "step": 37968
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.399626863937911e-12,
+      "loss": 0.7897,
+      "step": 37969
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.636055305733521e-12,
+      "loss": 0.7612,
+      "step": 37970
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.9452048311284785e-12,
+      "loss": 0.7113,
+      "step": 37971
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2.3270754467841217e-12,
+      "loss": 0.6229,
+      "step": 37972
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7816671549208964e-12,
+      "loss": 0.5747,
+      "step": 37973
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3089799610899179e-12,
+      "loss": 0.9024,
+      "step": 37974
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.090138675116323e-13,
+      "loss": 0.8073,
+      "step": 37975
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 5.817688786269315e-13,
+      "loss": 0.8522,
+      "step": 37976
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.2724499554603886e-13,
+      "loss": 0.8521,
+      "step": 37977
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.4544222048940016e-13,
+      "loss": 0.7323,
+      "step": 37978
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.636055567746155e-14,
+      "loss": 0.8034,
+      "step": 37979
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0,
+      "loss": 1.0363,
+      "step": 37980
+    },
     {
       "epoch": 1.0,
-      "step": 20,
-      "total_flos": 528928358400.0,
-      "train_loss": 2.351900500059128,
-      "train_runtime": 55.52,
-      "train_samples_per_second": 2.864,
-      "train_steps_per_second": 0.36
+      "step": 37980,
+      "total_flos": 867413013528576.0,
+      "train_loss": 0.9314424522648741,
+      "train_runtime": 44117.7787,
+      "train_samples_per_second": 13.774,
+      "train_steps_per_second": 0.861
     }
   ],
   "logging_steps": 1.0,
-  "max_steps": 20,
+  "max_steps": 37980,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 50000,
-  "total_flos": 528928358400.0,
-  "train_batch_size": 1,
+  "total_flos": 867413013528576.0,
+  "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null
 }