{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9999288104221542,
  "eval_steps": 352,
  "global_step": 7023,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 0.0,
      "loss": 3.5899,
      "step": 1
    },
    {
      "epoch": 0.0,
      "eval_loss": 4.178114414215088,
      "eval_runtime": 473.0271,
      "eval_samples_per_second": 43.31,
      "eval_steps_per_second": 14.437,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 4.1967,
      "step": 2
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.000000000000001e-06,
      "loss": 4.2584,
      "step": 3
    },
    {
      "epoch": 0.0,
      "learning_rate": 6e-06,
      "loss": 4.2861,
      "step": 4
    },
    {
      "epoch": 0.0,
      "learning_rate": 8.000000000000001e-06,
      "loss": 3.449,
      "step": 5
    },
    {
      "epoch": 0.0,
      "learning_rate": 1e-05,
      "loss": 3.6379,
      "step": 6
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.2e-05,
      "loss": 3.5514,
      "step": 7
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.4000000000000001e-05,
      "loss": 3.3639,
      "step": 8
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.6000000000000003e-05,
      "loss": 3.2568,
      "step": 9
    },
    {
      "epoch": 0.0,
      "learning_rate": 1.8e-05,
      "loss": 3.3578,
      "step": 10
    },
    {
      "epoch": 0.0,
      "learning_rate": 2e-05,
      "loss": 3.3858,
      "step": 11
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.2000000000000003e-05,
      "loss": 3.4463,
      "step": 12
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.4e-05,
      "loss": 3.3023,
      "step": 13
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.6000000000000002e-05,
      "loss": 3.3967,
      "step": 14
    },
    {
      "epoch": 0.0,
      "learning_rate": 2.8000000000000003e-05,
      "loss": 3.3402,
      "step": 15
    },
    {
      "epoch": 0.0,
      "learning_rate": 3e-05,
      "loss": 3.2368,
      "step": 16
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.2000000000000005e-05,
      "loss": 3.3075,
      "step": 17
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.4000000000000007e-05,
      "loss": 3.3434,
      "step": 18
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.6e-05,
      "loss": 3.3037,
      "step": 19
    },
    {
      "epoch": 0.0,
      "learning_rate": 3.8e-05,
      "loss": 3.2338,
      "step": 20
    },
    {
      "epoch": 0.0,
      "learning_rate": 4e-05,
      "loss": 3.1854,
      "step": 21
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.2e-05,
      "loss": 3.2481,
      "step": 22
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.4000000000000006e-05,
      "loss": 3.2006,
      "step": 23
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.600000000000001e-05,
      "loss": 3.2349,
      "step": 24
    },
    {
      "epoch": 0.0,
      "learning_rate": 4.8e-05,
      "loss": 3.2476,
      "step": 25
    },
    {
      "epoch": 0.0,
      "learning_rate": 5e-05,
      "loss": 3.169,
      "step": 26
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.2000000000000004e-05,
      "loss": 3.2349,
      "step": 27
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.4000000000000005e-05,
      "loss": 3.2007,
      "step": 28
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.6000000000000006e-05,
      "loss": 3.2467,
      "step": 29
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.8e-05,
      "loss": 3.2974,
      "step": 30
    },
    {
      "epoch": 0.0,
      "learning_rate": 6e-05,
      "loss": 3.0882,
      "step": 31
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.2e-05,
      "loss": 3.1722,
      "step": 32
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.400000000000001e-05,
      "loss": 3.1964,
      "step": 33
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.6e-05,
      "loss": 3.2419,
      "step": 34
    },
    {
      "epoch": 0.0,
      "learning_rate": 6.800000000000001e-05,
      "loss": 3.1016,
      "step": 35
    },
    {
      "epoch": 0.01,
      "learning_rate": 7e-05,
      "loss": 3.1917,
      "step": 36
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.2e-05,
      "loss": 3.2126,
      "step": 37
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.4e-05,
      "loss": 3.2239,
      "step": 38
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.6e-05,
      "loss": 3.1186,
      "step": 39
    },
    {
      "epoch": 0.01,
      "learning_rate": 7.800000000000001e-05,
      "loss": 3.2016,
      "step": 40
    },
    {
      "epoch": 0.01,
      "learning_rate": 8e-05,
      "loss": 3.2209,
      "step": 41
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.2e-05,
      "loss": 3.2119,
      "step": 42
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.4e-05,
      "loss": 3.3246,
      "step": 43
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.6e-05,
      "loss": 3.1526,
      "step": 44
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.800000000000001e-05,
      "loss": 3.22,
      "step": 45
    },
    {
      "epoch": 0.01,
      "learning_rate": 9e-05,
      "loss": 3.1782,
      "step": 46
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.200000000000001e-05,
      "loss": 3.1501,
      "step": 47
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.4e-05,
      "loss": 3.2063,
      "step": 48
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.6e-05,
      "loss": 3.2535,
      "step": 49
    },
    {
      "epoch": 0.01,
      "learning_rate": 9.8e-05,
      "loss": 3.1148,
      "step": 50
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001,
      "loss": 3.232,
      "step": 51
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00010200000000000001,
      "loss": 3.1592,
      "step": 52
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00010400000000000001,
      "loss": 3.3169,
      "step": 53
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00010600000000000002,
      "loss": 3.2689,
      "step": 54
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00010800000000000001,
      "loss": 3.2544,
      "step": 55
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00011000000000000002,
      "loss": 3.2462,
      "step": 56
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00011200000000000001,
      "loss": 3.2163,
      "step": 57
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00011399999999999999,
      "loss": 3.1604,
      "step": 58
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000116,
      "loss": 3.2952,
      "step": 59
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000118,
      "loss": 3.2749,
      "step": 60
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00012,
      "loss": 3.2169,
      "step": 61
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000122,
      "loss": 3.2749,
      "step": 62
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000124,
      "loss": 3.26,
      "step": 63
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000126,
      "loss": 3.2608,
      "step": 64
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00012800000000000002,
      "loss": 3.1959,
      "step": 65
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00013000000000000002,
      "loss": 3.1977,
      "step": 66
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000132,
      "loss": 3.2184,
      "step": 67
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000134,
      "loss": 3.1538,
      "step": 68
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00013600000000000003,
      "loss": 3.3864,
      "step": 69
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000138,
      "loss": 3.2612,
      "step": 70
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00014,
      "loss": 3.3267,
      "step": 71
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000142,
      "loss": 3.3697,
      "step": 72
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000144,
      "loss": 3.2061,
      "step": 73
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000146,
      "loss": 3.1879,
      "step": 74
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000148,
      "loss": 3.2455,
      "step": 75
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00015000000000000001,
      "loss": 3.2665,
      "step": 76
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000152,
      "loss": 3.3222,
      "step": 77
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000154,
      "loss": 3.2912,
      "step": 78
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00015600000000000002,
      "loss": 3.3734,
      "step": 79
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00015800000000000002,
      "loss": 3.2328,
      "step": 80
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00016,
      "loss": 3.3124,
      "step": 81
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000162,
      "loss": 3.2042,
      "step": 82
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000164,
      "loss": 3.3126,
      "step": 83
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000166,
      "loss": 3.3218,
      "step": 84
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000168,
      "loss": 3.3561,
      "step": 85
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00017,
      "loss": 3.4065,
      "step": 86
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000172,
      "loss": 3.4102,
      "step": 87
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000174,
      "loss": 3.4352,
      "step": 88
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00017600000000000002,
      "loss": 3.2285,
      "step": 89
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00017800000000000002,
      "loss": 3.3196,
      "step": 90
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00018,
      "loss": 3.4187,
      "step": 91
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000182,
      "loss": 3.505,
      "step": 92
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00018400000000000003,
      "loss": 3.4067,
      "step": 93
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00018600000000000002,
      "loss": 3.4128,
      "step": 94
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000188,
      "loss": 3.4732,
      "step": 95
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019,
      "loss": 3.3481,
      "step": 96
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000192,
      "loss": 3.4073,
      "step": 97
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000194,
      "loss": 3.5063,
      "step": 98
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.000196,
      "loss": 3.3983,
      "step": 99
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019800000000000002,
      "loss": 3.4428,
      "step": 100
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0002,
      "loss": 3.3764,
      "step": 101
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001999711107901199,
      "loss": 3.3813,
      "step": 102
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.0001999422215802398,
      "loss": 3.4354,
      "step": 103
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019991333237035966,
      "loss": 3.2258,
      "step": 104
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00019988444316047956,
      "loss": 3.3943,
      "step": 105
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019985555395059945,
      "loss": 3.3837,
      "step": 106
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019982666474071937,
      "loss": 3.3826,
      "step": 107
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019979777553083924,
      "loss": 3.4589,
      "step": 108
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019976888632095913,
      "loss": 3.3665,
      "step": 109
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019973999711107903,
      "loss": 3.4585,
      "step": 110
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019971110790119892,
      "loss": 3.5142,
      "step": 111
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001996822186913188,
      "loss": 3.4871,
      "step": 112
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019965332948143868,
      "loss": 3.4486,
      "step": 113
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019962444027155858,
      "loss": 3.5011,
      "step": 114
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019959555106167847,
      "loss": 3.3932,
      "step": 115
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019956666185179836,
      "loss": 3.5907,
      "step": 116
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019953777264191826,
      "loss": 3.4969,
      "step": 117
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019950888343203815,
      "loss": 3.5815,
      "step": 118
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019947999422215804,
      "loss": 3.4717,
      "step": 119
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019945110501227794,
      "loss": 3.4238,
      "step": 120
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001994222158023978,
      "loss": 3.5099,
      "step": 121
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001993933265925177,
      "loss": 3.5688,
      "step": 122
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001993644373826376,
      "loss": 3.4204,
      "step": 123
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019933554817275749,
      "loss": 3.5261,
      "step": 124
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019930665896287738,
      "loss": 3.6011,
      "step": 125
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019927776975299725,
      "loss": 3.5179,
      "step": 126
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019924888054311714,
      "loss": 3.5101,
      "step": 127
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019921999133323706,
      "loss": 3.5243,
      "step": 128
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019919110212335696,
      "loss": 3.6438,
      "step": 129
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019916221291347682,
      "loss": 3.4501,
      "step": 130
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019913332370359672,
      "loss": 3.4317,
      "step": 131
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001991044344937166,
      "loss": 3.5163,
      "step": 132
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001990755452838365,
      "loss": 3.4548,
      "step": 133
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019904665607395637,
      "loss": 3.5115,
      "step": 134
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019901776686407626,
      "loss": 3.5406,
      "step": 135
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019898887765419616,
      "loss": 3.4877,
      "step": 136
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019895998844431605,
      "loss": 3.567,
      "step": 137
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019893109923443595,
      "loss": 3.5098,
      "step": 138
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019890221002455584,
      "loss": 3.4999,
      "step": 139
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019887332081467573,
      "loss": 3.515,
      "step": 140
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019884443160479563,
      "loss": 3.6411,
      "step": 141
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019881554239491552,
      "loss": 3.6045,
      "step": 142
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001987866531850354,
      "loss": 3.5639,
      "step": 143
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019875776397515528,
      "loss": 3.3698,
      "step": 144
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019872887476527518,
      "loss": 3.5527,
      "step": 145
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019869998555539507,
      "loss": 3.4809,
      "step": 146
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019867109634551496,
      "loss": 3.5931,
      "step": 147
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019864220713563483,
      "loss": 3.4159,
      "step": 148
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019861331792575475,
      "loss": 3.5496,
      "step": 149
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019858442871587464,
      "loss": 3.5335,
      "step": 150
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019855553950599454,
      "loss": 3.4063,
      "step": 151
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001985266502961144,
      "loss": 3.4679,
      "step": 152
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001984977610862343,
      "loss": 3.4876,
      "step": 153
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001984688718763542,
      "loss": 3.5482,
      "step": 154
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019843998266647409,
      "loss": 3.6088,
      "step": 155
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019841109345659395,
      "loss": 3.5708,
      "step": 156
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019838220424671385,
      "loss": 3.4716,
      "step": 157
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019835331503683374,
      "loss": 3.6032,
      "step": 158
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019832442582695366,
      "loss": 3.4731,
      "step": 159
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019829553661707353,
      "loss": 3.4766,
      "step": 160
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019826664740719342,
      "loss": 3.4977,
      "step": 161
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019823775819731332,
      "loss": 3.4669,
      "step": 162
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001982088689874332,
      "loss": 3.6169,
      "step": 163
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.0001981799797775531,
      "loss": 3.4544,
      "step": 164
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019815109056767297,
      "loss": 3.5422,
      "step": 165
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019812220135779286,
      "loss": 3.6311,
      "step": 166
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019809331214791276,
      "loss": 3.513,
      "step": 167
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019806442293803265,
      "loss": 3.5973,
      "step": 168
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019803553372815255,
      "loss": 3.4974,
      "step": 169
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019800664451827244,
      "loss": 3.6382,
      "step": 170
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019797775530839233,
      "loss": 3.5406,
      "step": 171
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019794886609851223,
      "loss": 3.596,
      "step": 172
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019791997688863212,
      "loss": 3.5675,
      "step": 173
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.000197891087678752,
      "loss": 3.4857,
      "step": 174
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00019786219846887188,
      "loss": 3.5948,
      "step": 175
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019783330925899178,
      "loss": 3.5377,
      "step": 176
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019780442004911167,
      "loss": 3.4115,
      "step": 177
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019777553083923154,
      "loss": 3.5925,
      "step": 178
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019774664162935143,
      "loss": 3.6072,
      "step": 179
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019771775241947135,
      "loss": 3.4459,
      "step": 180
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019768886320959124,
      "loss": 3.675,
      "step": 181
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001976599739997111,
      "loss": 3.5013,
      "step": 182
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.000197631084789831,
      "loss": 3.5503,
      "step": 183
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001976021955799509,
      "loss": 3.4639,
      "step": 184
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001975733063700708,
      "loss": 3.6222,
      "step": 185
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001975444171601907,
      "loss": 3.6222,
      "step": 186
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019751552795031055,
      "loss": 3.5586,
      "step": 187
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019748663874043045,
      "loss": 3.353,
      "step": 188
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019745774953055034,
      "loss": 3.5714,
      "step": 189
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019742886032067023,
      "loss": 3.5703,
      "step": 190
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019739997111079013,
      "loss": 3.4602,
      "step": 191
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019737108190091002,
      "loss": 3.5076,
      "step": 192
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019734219269102992,
      "loss": 3.6046,
      "step": 193
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001973133034811498,
      "loss": 3.4482,
      "step": 194
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001972844142712697,
      "loss": 3.4445,
      "step": 195
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019725552506138957,
      "loss": 3.5888,
      "step": 196
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019722663585150946,
      "loss": 3.2793,
      "step": 197
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019719774664162936,
      "loss": 3.4957,
      "step": 198
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019716885743174925,
      "loss": 3.6092,
      "step": 199
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019713996822186912,
      "loss": 3.5575,
      "step": 200
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019711107901198904,
      "loss": 3.4428,
      "step": 201
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019708218980210893,
      "loss": 3.4696,
      "step": 202
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019705330059222883,
      "loss": 3.6291,
      "step": 203
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001970244113823487,
      "loss": 3.5864,
      "step": 204
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001969955221724686,
      "loss": 3.6399,
      "step": 205
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019696663296258848,
      "loss": 3.547,
      "step": 206
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019693774375270838,
      "loss": 3.5143,
      "step": 207
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019690885454282827,
      "loss": 3.4557,
      "step": 208
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019687996533294814,
      "loss": 3.4584,
      "step": 209
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019685107612306803,
      "loss": 3.4184,
      "step": 210
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019682218691318795,
      "loss": 3.5372,
      "step": 211
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019679329770330784,
      "loss": 3.4767,
      "step": 212
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001967644084934277,
      "loss": 3.5187,
      "step": 213
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001967355192835476,
      "loss": 3.5174,
      "step": 214
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001967066300736675,
      "loss": 3.5481,
      "step": 215
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001966777408637874,
      "loss": 3.5149,
      "step": 216
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001966488516539073,
      "loss": 3.5833,
      "step": 217
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019661996244402715,
      "loss": 3.3972,
      "step": 218
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019659107323414705,
      "loss": 3.5472,
      "step": 219
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019656218402426694,
      "loss": 3.5609,
      "step": 220
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019653329481438683,
      "loss": 3.6299,
      "step": 221
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019650440560450673,
      "loss": 3.4379,
      "step": 222
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019647551639462662,
      "loss": 3.5986,
      "step": 223
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019644662718474652,
      "loss": 3.5838,
      "step": 224
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001964177379748664,
      "loss": 3.4134,
      "step": 225
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019638884876498628,
      "loss": 3.3992,
      "step": 226
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019635995955510617,
      "loss": 3.5541,
      "step": 227
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019633107034522606,
      "loss": 3.5869,
      "step": 228
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019630218113534596,
      "loss": 3.5119,
      "step": 229
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019627329192546585,
      "loss": 3.4892,
      "step": 230
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019624440271558572,
      "loss": 3.6,
      "step": 231
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019621551350570564,
      "loss": 3.4632,
      "step": 232
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019618662429582553,
      "loss": 3.5046,
      "step": 233
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019615773508594543,
      "loss": 3.5646,
      "step": 234
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001961288458760653,
      "loss": 3.5583,
      "step": 235
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001960999566661852,
      "loss": 3.5467,
      "step": 236
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019607106745630508,
      "loss": 3.3647,
      "step": 237
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019604217824642498,
      "loss": 3.6019,
      "step": 238
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019601328903654484,
      "loss": 3.3073,
      "step": 239
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019598439982666474,
      "loss": 3.5884,
      "step": 240
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019595551061678463,
      "loss": 3.509,
      "step": 241
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019592662140690452,
      "loss": 3.3664,
      "step": 242
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019589773219702442,
      "loss": 3.5424,
      "step": 243
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001958688429871443,
      "loss": 3.3911,
      "step": 244
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0001958399537772642,
      "loss": 3.4583,
      "step": 245
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001958110645673841,
      "loss": 3.5823,
      "step": 246
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000195782175357504,
      "loss": 3.5662,
      "step": 247
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019575328614762386,
      "loss": 3.472,
      "step": 248
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019572439693774375,
      "loss": 3.5061,
      "step": 249
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019569550772786365,
      "loss": 3.5083,
      "step": 250
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019566661851798354,
      "loss": 3.4507,
      "step": 251
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019563772930810344,
      "loss": 3.4526,
      "step": 252
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019560884009822333,
      "loss": 3.5103,
      "step": 253
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019557995088834322,
      "loss": 3.5585,
      "step": 254
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019555106167846312,
      "loss": 3.505,
      "step": 255
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000195522172468583,
      "loss": 3.4499,
      "step": 256
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019549328325870288,
      "loss": 3.3568,
      "step": 257
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019546439404882277,
      "loss": 3.7057,
      "step": 258
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019543550483894266,
      "loss": 3.517,
      "step": 259
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019540661562906256,
      "loss": 3.4745,
      "step": 260
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019537772641918243,
      "loss": 3.4541,
      "step": 261
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019534883720930232,
      "loss": 3.5701,
      "step": 262
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001953199479994222,
      "loss": 3.6835,
      "step": 263
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019529105878954213,
      "loss": 3.5116,
      "step": 264
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.000195262169579662,
      "loss": 3.3648,
      "step": 265
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001952332803697819,
      "loss": 3.3722,
      "step": 266
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001952043911599018,
      "loss": 3.3931,
      "step": 267
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019517550195002168,
      "loss": 3.5807,
      "step": 268
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019514661274014158,
      "loss": 3.5416,
      "step": 269
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019511772353026144,
      "loss": 3.582,
      "step": 270
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019508883432038134,
      "loss": 3.4652,
      "step": 271
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019505994511050123,
      "loss": 3.5732,
      "step": 272
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019503105590062112,
      "loss": 3.5297,
      "step": 273
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019500216669074102,
      "loss": 3.7003,
      "step": 274
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001949732774808609,
      "loss": 3.5983,
      "step": 275
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001949443882709808,
      "loss": 3.6118,
      "step": 276
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001949154990611007,
      "loss": 3.4791,
      "step": 277
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001948866098512206,
      "loss": 3.6391,
      "step": 278
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019485772064134046,
      "loss": 3.5959,
      "step": 279
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019482883143146035,
      "loss": 3.5101,
      "step": 280
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019479994222158025,
      "loss": 3.5986,
      "step": 281
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019477105301170014,
      "loss": 3.5414,
      "step": 282
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019474216380182,
      "loss": 3.6653,
      "step": 283
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001947132745919399,
      "loss": 3.5027,
      "step": 284
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019468438538205982,
      "loss": 3.4249,
      "step": 285
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019465549617217972,
      "loss": 3.4433,
      "step": 286
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019462660696229958,
      "loss": 3.6223,
      "step": 287
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019459771775241948,
      "loss": 3.5467,
      "step": 288
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019456882854253937,
      "loss": 3.61,
      "step": 289
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019453993933265926,
      "loss": 3.6137,
      "step": 290
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019451105012277916,
      "loss": 3.5088,
      "step": 291
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019448216091289903,
      "loss": 3.5359,
      "step": 292
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019445327170301892,
      "loss": 3.5458,
      "step": 293
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001944243824931388,
      "loss": 3.5681,
      "step": 294
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019439549328325873,
      "loss": 3.4299,
      "step": 295
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001943666040733786,
      "loss": 3.5176,
      "step": 296
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001943377148634985,
      "loss": 3.641,
      "step": 297
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001943088256536184,
      "loss": 3.4167,
      "step": 298
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019427993644373828,
      "loss": 3.4261,
      "step": 299
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019425104723385818,
      "loss": 3.4559,
      "step": 300
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019422215802397804,
      "loss": 3.5327,
      "step": 301
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019419326881409794,
      "loss": 3.4398,
      "step": 302
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019416437960421783,
      "loss": 3.5033,
      "step": 303
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019413549039433772,
      "loss": 3.6118,
      "step": 304
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001941066011844576,
      "loss": 3.4967,
      "step": 305
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001940777119745775,
      "loss": 3.4717,
      "step": 306
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001940488227646974,
      "loss": 3.355,
      "step": 307
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001940199335548173,
      "loss": 3.5887,
      "step": 308
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019399104434493717,
      "loss": 3.6378,
      "step": 309
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019396215513505706,
      "loss": 3.5147,
      "step": 310
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019393326592517695,
      "loss": 3.4631,
      "step": 311
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019390437671529685,
      "loss": 3.5992,
      "step": 312
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019387548750541674,
      "loss": 3.5586,
      "step": 313
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001938465982955366,
      "loss": 3.5399,
      "step": 314
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0001938177090856565,
      "loss": 3.4912,
      "step": 315
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.00019378881987577642,
      "loss": 3.6064,
      "step": 316
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019375993066589632,
      "loss": 3.7006,
      "step": 317
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019373104145601618,
      "loss": 3.5503,
      "step": 318
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019370215224613608,
      "loss": 3.5618,
      "step": 319
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019367326303625597,
      "loss": 3.651,
      "step": 320
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019364437382637587,
      "loss": 3.5181,
      "step": 321
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019361548461649576,
      "loss": 3.706,
      "step": 322
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019358659540661563,
      "loss": 3.601,
      "step": 323
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019355770619673552,
      "loss": 3.5891,
      "step": 324
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001935288169868554,
      "loss": 3.3504,
      "step": 325
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001934999277769753,
      "loss": 3.4464,
      "step": 326
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001934710385670952,
      "loss": 3.5102,
      "step": 327
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001934421493572151,
      "loss": 3.5796,
      "step": 328
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000193413260147335,
      "loss": 3.5751,
      "step": 329
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019338437093745488,
      "loss": 3.4911,
      "step": 330
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019335548172757475,
      "loss": 3.5161,
      "step": 331
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019332659251769464,
      "loss": 3.4412,
      "step": 332
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019329770330781454,
      "loss": 3.4896,
      "step": 333
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019326881409793443,
      "loss": 3.5323,
      "step": 334
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019323992488805432,
      "loss": 3.633,
      "step": 335
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001932110356781742,
      "loss": 3.5541,
      "step": 336
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001931821464682941,
      "loss": 3.5204,
      "step": 337
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000193153257258414,
      "loss": 3.7043,
      "step": 338
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001931243680485339,
      "loss": 3.5533,
      "step": 339
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019309547883865377,
      "loss": 3.5269,
      "step": 340
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019306658962877366,
      "loss": 3.4716,
      "step": 341
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019303770041889355,
      "loss": 3.4324,
      "step": 342
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019300881120901345,
      "loss": 3.5459,
      "step": 343
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019297992199913334,
      "loss": 3.543,
      "step": 344
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001929510327892532,
      "loss": 3.5318,
      "step": 345
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001929221435793731,
      "loss": 3.6708,
      "step": 346
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000192893254369493,
      "loss": 3.5583,
      "step": 347
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001928643651596129,
      "loss": 3.5955,
      "step": 348
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019283547594973278,
      "loss": 3.4893,
      "step": 349
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019280658673985268,
      "loss": 3.6199,
      "step": 350
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019277769752997257,
      "loss": 3.4667,
      "step": 351
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019274880832009247,
      "loss": 3.4689,
      "step": 352
    },
    {
      "epoch": 0.05,
      "eval_loss": 3.59875750541687,
      "eval_runtime": 472.4811,
      "eval_samples_per_second": 43.36,
      "eval_steps_per_second": 14.453,
      "step": 352
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019271991911021233,
      "loss": 3.5819,
      "step": 353
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019269102990033223,
      "loss": 3.5459,
      "step": 354
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019266214069045212,
      "loss": 3.5412,
      "step": 355
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.000192633251480572,
      "loss": 3.5603,
      "step": 356
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001926043622706919,
      "loss": 3.3498,
      "step": 357
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001925754730608118,
      "loss": 3.2999,
      "step": 358
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001925465838509317,
      "loss": 3.5822,
      "step": 359
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001925176946410516,
      "loss": 3.583,
      "step": 360
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019248880543117148,
      "loss": 3.6066,
      "step": 361
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019245991622129135,
      "loss": 3.6497,
      "step": 362
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019243102701141124,
      "loss": 3.551,
      "step": 363
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019240213780153114,
      "loss": 3.3119,
      "step": 364
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019237324859165103,
      "loss": 3.5062,
      "step": 365
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001923443593817709,
      "loss": 3.5501,
      "step": 366
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001923154701718908,
      "loss": 3.5859,
      "step": 367
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019228658096201069,
      "loss": 3.5105,
      "step": 368
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001922576917521306,
      "loss": 3.5117,
      "step": 369
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019222880254225047,
      "loss": 3.5607,
      "step": 370
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019219991333237037,
      "loss": 3.5971,
      "step": 371
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019217102412249026,
      "loss": 3.573,
      "step": 372
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019214213491261015,
      "loss": 3.5984,
      "step": 373
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019211324570273005,
      "loss": 3.4534,
      "step": 374
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019208435649284991,
      "loss": 3.5523,
      "step": 375
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001920554672829698,
      "loss": 3.5324,
      "step": 376
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001920265780730897,
      "loss": 3.5271,
      "step": 377
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001919976888632096,
      "loss": 3.5481,
      "step": 378
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0001919687996533295,
      "loss": 3.5117,
      "step": 379
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019193991044344938,
      "loss": 3.3971,
      "step": 380
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019191102123356928,
      "loss": 3.5115,
      "step": 381
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019188213202368917,
      "loss": 3.51,
      "step": 382
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019185324281380907,
      "loss": 3.6974,
      "step": 383
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019182435360392893,
      "loss": 3.5188,
      "step": 384
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019179546439404883,
      "loss": 3.5516,
      "step": 385
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.00019176657518416872,
      "loss": 3.6164,
      "step": 386
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019173768597428861,
      "loss": 3.6524,
      "step": 387
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019170879676440848,
      "loss": 3.6024,
      "step": 388
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019167990755452837,
      "loss": 3.4035,
      "step": 389
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001916510183446483,
      "loss": 3.3601,
      "step": 390
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001916221291347682,
      "loss": 3.5517,
      "step": 391
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019159323992488806,
      "loss": 3.4734,
      "step": 392
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019156435071500795,
      "loss": 3.4639,
      "step": 393
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019153546150512784,
      "loss": 3.4812,
      "step": 394
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019150657229524774,
      "loss": 3.7109,
      "step": 395
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019147768308536763,
      "loss": 3.4681,
      "step": 396
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001914487938754875,
      "loss": 3.6094,
      "step": 397
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001914199046656074,
      "loss": 3.5162,
      "step": 398
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019139101545572729,
      "loss": 3.4222,
      "step": 399
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001913621262458472,
      "loss": 3.577,
      "step": 400
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019133323703596707,
      "loss": 3.6015,
      "step": 401
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019130434782608697,
      "loss": 3.6354,
      "step": 402
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019127545861620686,
      "loss": 3.4482,
      "step": 403
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019124656940632675,
      "loss": 3.5481,
      "step": 404
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019121768019644665,
      "loss": 3.4145,
      "step": 405
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019118879098656652,
      "loss": 3.365,
      "step": 406
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001911599017766864,
      "loss": 3.5916,
      "step": 407
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001911310125668063,
      "loss": 3.6518,
      "step": 408
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001911021233569262,
      "loss": 3.6107,
      "step": 409
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001910732341470461,
      "loss": 3.549,
      "step": 410
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019104434493716598,
      "loss": 3.6271,
      "step": 411
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019101545572728588,
      "loss": 3.5452,
      "step": 412
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019098656651740577,
      "loss": 3.6264,
      "step": 413
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019095767730752564,
      "loss": 3.5031,
      "step": 414
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019092878809764553,
      "loss": 3.5594,
      "step": 415
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019089989888776543,
      "loss": 3.3923,
      "step": 416
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019087100967788532,
      "loss": 3.3375,
      "step": 417
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019084212046800521,
      "loss": 3.5472,
      "step": 418
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019081323125812508,
      "loss": 3.6317,
      "step": 419
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019078434204824497,
      "loss": 3.6466,
      "step": 420
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001907554528383649,
      "loss": 3.5929,
      "step": 421
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001907265636284848,
      "loss": 3.4864,
      "step": 422
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019069767441860466,
      "loss": 3.5866,
      "step": 423
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019066878520872455,
      "loss": 3.3665,
      "step": 424
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019063989599884444,
      "loss": 3.5244,
      "step": 425
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019061100678896434,
      "loss": 3.3427,
      "step": 426
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019058211757908423,
      "loss": 3.6314,
      "step": 427
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001905532283692041,
      "loss": 3.5961,
      "step": 428
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000190524339159324,
      "loss": 3.6028,
      "step": 429
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019049544994944389,
      "loss": 3.565,
      "step": 430
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001904665607395638,
      "loss": 3.6294,
      "step": 431
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019043767152968367,
      "loss": 3.5588,
      "step": 432
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019040878231980357,
      "loss": 3.5346,
      "step": 433
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019037989310992346,
      "loss": 3.5715,
      "step": 434
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019035100390004335,
      "loss": 3.6058,
      "step": 435
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019032211469016322,
      "loss": 3.6087,
      "step": 436
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019029322548028312,
      "loss": 3.5346,
      "step": 437
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.000190264336270403,
      "loss": 3.5482,
      "step": 438
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001902354470605229,
      "loss": 3.553,
      "step": 439
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001902065578506428,
      "loss": 3.5424,
      "step": 440
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019017766864076266,
      "loss": 3.3804,
      "step": 441
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019014877943088258,
      "loss": 3.5441,
      "step": 442
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019011989022100248,
      "loss": 3.5132,
      "step": 443
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019009100101112237,
      "loss": 3.5497,
      "step": 444
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019006211180124224,
      "loss": 3.5957,
      "step": 445
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019003322259136213,
      "loss": 3.5656,
      "step": 446
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00019000433338148203,
      "loss": 3.5914,
      "step": 447
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018997544417160192,
      "loss": 3.5543,
      "step": 448
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018994655496172181,
      "loss": 3.6238,
      "step": 449
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018991766575184168,
      "loss": 3.4488,
      "step": 450
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018988877654196157,
      "loss": 3.5148,
      "step": 451
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0001898598873320815,
      "loss": 3.6172,
      "step": 452
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018983099812220136,
      "loss": 3.5025,
      "step": 453
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018980210891232126,
      "loss": 3.4935,
      "step": 454
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018977321970244115,
      "loss": 3.4364,
      "step": 455
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.00018974433049256104,
      "loss": 3.5489,
      "step": 456
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018971544128268094,
      "loss": 3.5546,
      "step": 457
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001896865520728008,
      "loss": 3.4535,
      "step": 458
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001896576628629207,
      "loss": 3.545,
      "step": 459
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001896287736530406,
      "loss": 3.5972,
      "step": 460
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018959988444316049,
      "loss": 3.6178,
      "step": 461
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018957099523328038,
      "loss": 3.3621,
      "step": 462
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018954210602340027,
      "loss": 3.4738,
      "step": 463
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018951321681352017,
      "loss": 3.6158,
      "step": 464
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018948432760364006,
      "loss": 3.4559,
      "step": 465
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018945543839375995,
      "loss": 3.4635,
      "step": 466
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018942654918387982,
      "loss": 3.7056,
      "step": 467
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018939765997399972,
      "loss": 3.5761,
      "step": 468
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001893687707641196,
      "loss": 3.537,
      "step": 469
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001893398815542395,
      "loss": 3.6607,
      "step": 470
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001893109923443594,
      "loss": 3.4395,
      "step": 471
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018928210313447926,
      "loss": 3.4993,
      "step": 472
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018925321392459918,
      "loss": 3.4373,
      "step": 473
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018922432471471908,
      "loss": 3.5638,
      "step": 474
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018919543550483895,
      "loss": 3.5201,
      "step": 475
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018916654629495884,
      "loss": 3.3825,
      "step": 476
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018913765708507873,
      "loss": 3.4802,
      "step": 477
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018910876787519863,
      "loss": 3.4747,
      "step": 478
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018907987866531852,
      "loss": 3.4511,
      "step": 479
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001890509894554384,
      "loss": 3.5461,
      "step": 480
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018902210024555828,
      "loss": 3.455,
      "step": 481
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018899321103567817,
      "loss": 3.5123,
      "step": 482
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018896432182579807,
      "loss": 3.5906,
      "step": 483
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018893543261591796,
      "loss": 3.421,
      "step": 484
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018890654340603786,
      "loss": 3.5119,
      "step": 485
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018887765419615775,
      "loss": 3.6598,
      "step": 486
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018884876498627764,
      "loss": 3.5187,
      "step": 487
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018881987577639754,
      "loss": 3.5036,
      "step": 488
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001887909865665174,
      "loss": 3.6146,
      "step": 489
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001887620973566373,
      "loss": 3.6127,
      "step": 490
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001887332081467572,
      "loss": 3.494,
      "step": 491
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018870431893687709,
      "loss": 3.5811,
      "step": 492
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018867542972699695,
      "loss": 3.6014,
      "step": 493
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018864654051711687,
      "loss": 3.4939,
      "step": 494
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018861765130723677,
      "loss": 3.5964,
      "step": 495
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018858876209735666,
      "loss": 3.5323,
      "step": 496
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018855987288747653,
      "loss": 3.5668,
      "step": 497
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018853098367759642,
      "loss": 3.5236,
      "step": 498
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018850209446771632,
      "loss": 3.6417,
      "step": 499
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001884732052578362,
      "loss": 3.7174,
      "step": 500
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001884443160479561,
      "loss": 3.4265,
      "step": 501
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018841542683807597,
      "loss": 3.4595,
      "step": 502
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018838653762819586,
      "loss": 3.3742,
      "step": 503
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018835764841831576,
      "loss": 3.6812,
      "step": 504
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018832875920843568,
      "loss": 3.64,
      "step": 505
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018829986999855555,
      "loss": 3.4502,
      "step": 506
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018827098078867544,
      "loss": 3.4054,
      "step": 507
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018824209157879533,
      "loss": 3.4928,
      "step": 508
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018821320236891523,
      "loss": 3.4226,
      "step": 509
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018818431315903512,
      "loss": 3.5288,
      "step": 510
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000188155423949155,
      "loss": 3.5659,
      "step": 511
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018812653473927488,
      "loss": 3.5009,
      "step": 512
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018809764552939477,
      "loss": 3.4401,
      "step": 513
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018806875631951467,
      "loss": 3.5149,
      "step": 514
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018803986710963456,
      "loss": 3.5233,
      "step": 515
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018801097789975446,
      "loss": 3.4313,
      "step": 516
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018798208868987435,
      "loss": 3.5574,
      "step": 517
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018795319947999424,
      "loss": 3.5375,
      "step": 518
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001879243102701141,
      "loss": 3.5875,
      "step": 519
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.000187895421060234,
      "loss": 3.4169,
      "step": 520
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001878665318503539,
      "loss": 3.4563,
      "step": 521
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0001878376426404738,
      "loss": 3.3797,
      "step": 522
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018780875343059369,
      "loss": 3.4075,
      "step": 523
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018777986422071355,
      "loss": 3.5773,
      "step": 524
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018775097501083345,
      "loss": 3.5523,
      "step": 525
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.00018772208580095337,
      "loss": 3.5288,
      "step": 526
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018769319659107326,
      "loss": 3.6048,
      "step": 527
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018766430738119313,
      "loss": 3.56,
      "step": 528
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018763541817131302,
      "loss": 3.5448,
      "step": 529
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018760652896143292,
      "loss": 3.6165,
      "step": 530
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001875776397515528,
      "loss": 3.495,
      "step": 531
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001875487505416727,
      "loss": 3.6955,
      "step": 532
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018751986133179257,
      "loss": 3.4293,
      "step": 533
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018749097212191246,
      "loss": 3.4176,
      "step": 534
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018746208291203236,
      "loss": 3.5641,
      "step": 535
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018743319370215228,
      "loss": 3.6007,
      "step": 536
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018740430449227215,
      "loss": 3.4111,
      "step": 537
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018737541528239204,
      "loss": 3.456,
      "step": 538
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018734652607251193,
      "loss": 3.507,
      "step": 539
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018731763686263183,
      "loss": 3.6267,
      "step": 540
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001872887476527517,
      "loss": 3.5538,
      "step": 541
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001872598584428716,
      "loss": 3.4112,
      "step": 542
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018723096923299148,
      "loss": 3.5022,
      "step": 543
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018720208002311138,
      "loss": 3.6393,
      "step": 544
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018717319081323127,
      "loss": 3.6233,
      "step": 545
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018714430160335114,
      "loss": 3.651,
      "step": 546
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018711541239347106,
      "loss": 3.6567,
      "step": 547
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018708652318359095,
      "loss": 3.3348,
      "step": 548
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018705763397371084,
      "loss": 3.5277,
      "step": 549
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001870287447638307,
      "loss": 3.4866,
      "step": 550
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001869998555539506,
      "loss": 3.5642,
      "step": 551
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001869709663440705,
      "loss": 3.4392,
      "step": 552
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001869420771341904,
      "loss": 3.6856,
      "step": 553
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018691318792431029,
      "loss": 3.4941,
      "step": 554
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018688429871443015,
      "loss": 3.5972,
      "step": 555
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018685540950455005,
      "loss": 3.5312,
      "step": 556
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018682652029466997,
      "loss": 3.5929,
      "step": 557
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018679763108478986,
      "loss": 3.6768,
      "step": 558
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018676874187490973,
      "loss": 3.5386,
      "step": 559
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018673985266502962,
      "loss": 3.4734,
      "step": 560
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018671096345514952,
      "loss": 3.5429,
      "step": 561
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001866820742452694,
      "loss": 3.4172,
      "step": 562
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018665318503538928,
      "loss": 3.5298,
      "step": 563
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018662429582550917,
      "loss": 3.547,
      "step": 564
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018659540661562906,
      "loss": 3.4804,
      "step": 565
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018656651740574896,
      "loss": 3.5119,
      "step": 566
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018653762819586885,
      "loss": 3.5192,
      "step": 567
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018650873898598875,
      "loss": 3.5618,
      "step": 568
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018647984977610864,
      "loss": 3.5257,
      "step": 569
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018645096056622853,
      "loss": 3.5611,
      "step": 570
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018642207135634843,
      "loss": 3.3794,
      "step": 571
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001863931821464683,
      "loss": 3.4684,
      "step": 572
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001863642929365882,
      "loss": 3.5756,
      "step": 573
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018633540372670808,
      "loss": 3.6312,
      "step": 574
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018630651451682798,
      "loss": 3.447,
      "step": 575
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018627762530694787,
      "loss": 3.5265,
      "step": 576
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018624873609706774,
      "loss": 3.5244,
      "step": 577
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018621984688718766,
      "loss": 3.5743,
      "step": 578
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018619095767730755,
      "loss": 3.4801,
      "step": 579
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018616206846742742,
      "loss": 3.5325,
      "step": 580
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001861331792575473,
      "loss": 3.603,
      "step": 581
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001861042900476672,
      "loss": 3.4021,
      "step": 582
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0001860754008377871,
      "loss": 3.5223,
      "step": 583
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000186046511627907,
      "loss": 3.5119,
      "step": 584
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018601762241802686,
      "loss": 3.4831,
      "step": 585
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018598873320814675,
      "loss": 3.4985,
      "step": 586
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018595984399826665,
      "loss": 3.6217,
      "step": 587
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018593095478838654,
      "loss": 3.5672,
      "step": 588
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018590206557850643,
      "loss": 3.4158,
      "step": 589
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018587317636862633,
      "loss": 3.5823,
      "step": 590
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018584428715874622,
      "loss": 3.5354,
      "step": 591
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018581539794886612,
      "loss": 3.479,
      "step": 592
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.000185786508738986,
      "loss": 3.4195,
      "step": 593
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018575761952910588,
      "loss": 3.6027,
      "step": 594
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018572873031922577,
      "loss": 3.6142,
      "step": 595
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.00018569984110934566,
      "loss": 3.5468,
      "step": 596
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018567095189946556,
      "loss": 3.504,
      "step": 597
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018564206268958545,
      "loss": 3.5126,
      "step": 598
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018561317347970535,
      "loss": 3.4528,
      "step": 599
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018558428426982524,
      "loss": 3.3295,
      "step": 600
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018555539505994513,
      "loss": 3.5697,
      "step": 601
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000185526505850065,
      "loss": 3.5069,
      "step": 602
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001854976166401849,
      "loss": 3.5449,
      "step": 603
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001854687274303048,
      "loss": 3.4998,
      "step": 604
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018543983822042468,
      "loss": 3.2614,
      "step": 605
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018541094901054458,
      "loss": 3.5806,
      "step": 606
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018538205980066444,
      "loss": 3.568,
      "step": 607
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018535317059078434,
      "loss": 3.6575,
      "step": 608
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018532428138090426,
      "loss": 3.4329,
      "step": 609
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018529539217102415,
      "loss": 3.5881,
      "step": 610
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018526650296114402,
      "loss": 3.4916,
      "step": 611
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001852376137512639,
      "loss": 3.5186,
      "step": 612
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001852087245413838,
      "loss": 3.528,
      "step": 613
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001851798353315037,
      "loss": 3.4139,
      "step": 614
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001851509461216236,
      "loss": 3.5184,
      "step": 615
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018512205691174346,
      "loss": 3.5681,
      "step": 616
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018509316770186335,
      "loss": 3.5774,
      "step": 617
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018506427849198325,
      "loss": 3.5222,
      "step": 618
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018503538928210314,
      "loss": 3.561,
      "step": 619
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018500650007222303,
      "loss": 3.4017,
      "step": 620
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018497761086234293,
      "loss": 3.4531,
      "step": 621
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018494872165246282,
      "loss": 3.5615,
      "step": 622
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018491983244258272,
      "loss": 3.5878,
      "step": 623
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018489094323270258,
      "loss": 3.6159,
      "step": 624
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018486205402282248,
      "loss": 3.496,
      "step": 625
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018483316481294237,
      "loss": 3.4758,
      "step": 626
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018480427560306226,
      "loss": 3.4669,
      "step": 627
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018477538639318216,
      "loss": 3.6198,
      "step": 628
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018474649718330203,
      "loss": 3.5373,
      "step": 629
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018471760797342195,
      "loss": 3.5283,
      "step": 630
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018468871876354184,
      "loss": 3.5688,
      "step": 631
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018465982955366173,
      "loss": 3.439,
      "step": 632
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001846309403437816,
      "loss": 3.4053,
      "step": 633
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001846020511339015,
      "loss": 3.3194,
      "step": 634
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001845731619240214,
      "loss": 3.4613,
      "step": 635
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018454427271414128,
      "loss": 3.5199,
      "step": 636
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018451538350426118,
      "loss": 3.5138,
      "step": 637
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018448649429438104,
      "loss": 3.5518,
      "step": 638
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018445760508450094,
      "loss": 3.5416,
      "step": 639
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018442871587462083,
      "loss": 3.4615,
      "step": 640
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018439982666474075,
      "loss": 3.638,
      "step": 641
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018437093745486062,
      "loss": 3.5444,
      "step": 642
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001843420482449805,
      "loss": 3.638,
      "step": 643
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001843131590351004,
      "loss": 3.451,
      "step": 644
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001842842698252203,
      "loss": 3.5095,
      "step": 645
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018425538061534017,
      "loss": 3.6475,
      "step": 646
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018422649140546006,
      "loss": 3.5567,
      "step": 647
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018419760219557995,
      "loss": 3.4693,
      "step": 648
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018416871298569985,
      "loss": 3.6224,
      "step": 649
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018413982377581974,
      "loss": 3.4074,
      "step": 650
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018411093456593963,
      "loss": 3.4079,
      "step": 651
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018408204535605953,
      "loss": 3.6761,
      "step": 652
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018405315614617942,
      "loss": 3.4546,
      "step": 653
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018402426693629932,
      "loss": 3.4474,
      "step": 654
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018399537772641918,
      "loss": 3.5165,
      "step": 655
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018396648851653908,
      "loss": 3.5854,
      "step": 656
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018393759930665897,
      "loss": 3.5001,
      "step": 657
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018390871009677886,
      "loss": 3.5853,
      "step": 658
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018387982088689876,
      "loss": 3.5338,
      "step": 659
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018385093167701863,
      "loss": 3.5332,
      "step": 660
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018382204246713852,
      "loss": 3.5204,
      "step": 661
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018379315325725844,
      "loss": 3.4821,
      "step": 662
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018376426404737833,
      "loss": 3.4963,
      "step": 663
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001837353748374982,
      "loss": 3.5454,
      "step": 664
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0001837064856276181,
      "loss": 3.5903,
      "step": 665
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.000183677596417738,
      "loss": 3.5212,
      "step": 666
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.00018364870720785788,
      "loss": 3.5175,
      "step": 667
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018361981799797775,
      "loss": 3.5781,
      "step": 668
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018359092878809764,
      "loss": 3.4063,
      "step": 669
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018356203957821754,
      "loss": 3.3629,
      "step": 670
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018353315036833743,
      "loss": 3.512,
      "step": 671
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018350426115845732,
      "loss": 3.5793,
      "step": 672
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018347537194857722,
      "loss": 3.4232,
      "step": 673
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001834464827386971,
      "loss": 3.608,
      "step": 674
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.000183417593528817,
      "loss": 3.4862,
      "step": 675
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001833887043189369,
      "loss": 3.5062,
      "step": 676
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018335981510905677,
      "loss": 3.5194,
      "step": 677
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018333092589917666,
      "loss": 3.4268,
      "step": 678
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018330203668929655,
      "loss": 3.4015,
      "step": 679
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018327314747941645,
      "loss": 3.5977,
      "step": 680
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018324425826953634,
      "loss": 3.4977,
      "step": 681
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001832153690596562,
      "loss": 3.5338,
      "step": 682
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018318647984977613,
      "loss": 3.5607,
      "step": 683
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018315759063989602,
      "loss": 3.562,
      "step": 684
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018312870143001592,
      "loss": 3.5139,
      "step": 685
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018309981222013578,
      "loss": 3.5853,
      "step": 686
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018307092301025568,
      "loss": 3.5065,
      "step": 687
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018304203380037557,
      "loss": 3.5055,
      "step": 688
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018301314459049546,
      "loss": 3.494,
      "step": 689
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018298425538061533,
      "loss": 3.563,
      "step": 690
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018295536617073523,
      "loss": 3.5116,
      "step": 691
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018292647696085512,
      "loss": 3.5446,
      "step": 692
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018289758775097504,
      "loss": 3.4678,
      "step": 693
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001828686985410949,
      "loss": 3.5063,
      "step": 694
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001828398093312148,
      "loss": 3.49,
      "step": 695
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001828109201213347,
      "loss": 3.5029,
      "step": 696
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001827820309114546,
      "loss": 3.504,
      "step": 697
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018275314170157448,
      "loss": 3.5638,
      "step": 698
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018272425249169435,
      "loss": 3.5541,
      "step": 699
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018269536328181424,
      "loss": 3.5315,
      "step": 700
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018266647407193414,
      "loss": 3.5832,
      "step": 701
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018263758486205403,
      "loss": 3.5319,
      "step": 702
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018260869565217392,
      "loss": 3.5295,
      "step": 703
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018257980644229382,
      "loss": 3.519,
      "step": 704
    },
    {
      "epoch": 0.1,
      "eval_loss": 3.620858669281006,
      "eval_runtime": 471.8255,
      "eval_samples_per_second": 43.421,
      "eval_steps_per_second": 14.474,
      "step": 704
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001825509172324137,
      "loss": 3.4535,
      "step": 705
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001825220280225336,
      "loss": 3.33,
      "step": 706
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018249313881265347,
      "loss": 3.4858,
      "step": 707
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018246424960277337,
      "loss": 3.4598,
      "step": 708
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018243536039289326,
      "loss": 3.5521,
      "step": 709
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018240647118301315,
      "loss": 3.5254,
      "step": 710
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018237758197313305,
      "loss": 3.5396,
      "step": 711
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018234869276325291,
      "loss": 3.5268,
      "step": 712
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001823198035533728,
      "loss": 3.6661,
      "step": 713
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018229091434349273,
      "loss": 3.5742,
      "step": 714
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018226202513361262,
      "loss": 3.6033,
      "step": 715
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001822331359237325,
      "loss": 3.5636,
      "step": 716
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018220424671385238,
      "loss": 3.5835,
      "step": 717
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018217535750397228,
      "loss": 3.5446,
      "step": 718
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018214646829409217,
      "loss": 3.5002,
      "step": 719
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018211757908421206,
      "loss": 3.3511,
      "step": 720
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018208868987433193,
      "loss": 3.4461,
      "step": 721
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018205980066445183,
      "loss": 3.4288,
      "step": 722
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018203091145457172,
      "loss": 3.4757,
      "step": 723
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001820020222446916,
      "loss": 3.4678,
      "step": 724
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001819731330348115,
      "loss": 3.3662,
      "step": 725
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001819442438249314,
      "loss": 3.6651,
      "step": 726
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001819153546150513,
      "loss": 3.5741,
      "step": 727
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001818864654051712,
      "loss": 3.636,
      "step": 728
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018185757619529106,
      "loss": 3.5109,
      "step": 729
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018182868698541095,
      "loss": 3.4351,
      "step": 730
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018179979777553084,
      "loss": 3.495,
      "step": 731
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018177090856565074,
      "loss": 3.5606,
      "step": 732
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018174201935577063,
      "loss": 3.511,
      "step": 733
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001817131301458905,
      "loss": 3.4717,
      "step": 734
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.00018168424093601042,
      "loss": 3.5888,
      "step": 735
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001816553517261303,
      "loss": 3.3341,
      "step": 736
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0001816264625162502,
      "loss": 3.4462,
      "step": 737
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018159757330637007,
      "loss": 3.5978,
      "step": 738
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018156868409648997,
      "loss": 3.5568,
      "step": 739
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018153979488660986,
      "loss": 3.4909,
      "step": 740
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018151090567672975,
      "loss": 3.4277,
      "step": 741
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018148201646684965,
      "loss": 3.5154,
      "step": 742
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018145312725696951,
      "loss": 3.5742,
      "step": 743
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001814242380470894,
      "loss": 3.4974,
      "step": 744
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001813953488372093,
      "loss": 3.4369,
      "step": 745
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018136645962732922,
      "loss": 3.5458,
      "step": 746
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001813375704174491,
      "loss": 3.5238,
      "step": 747
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018130868120756898,
      "loss": 3.4968,
      "step": 748
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018127979199768888,
      "loss": 3.5419,
      "step": 749
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018125090278780877,
      "loss": 3.3761,
      "step": 750
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018122201357792864,
      "loss": 3.6181,
      "step": 751
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018119312436804853,
      "loss": 3.5131,
      "step": 752
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018116423515816843,
      "loss": 3.4547,
      "step": 753
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018113534594828832,
      "loss": 3.5306,
      "step": 754
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001811064567384082,
      "loss": 3.5011,
      "step": 755
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001810775675285281,
      "loss": 3.5871,
      "step": 756
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000181048678318648,
      "loss": 3.4811,
      "step": 757
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001810197891087679,
      "loss": 3.544,
      "step": 758
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001809908998988878,
      "loss": 3.5499,
      "step": 759
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018096201068900766,
      "loss": 3.6015,
      "step": 760
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018093312147912755,
      "loss": 3.4873,
      "step": 761
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018090423226924744,
      "loss": 3.5266,
      "step": 762
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018087534305936734,
      "loss": 3.5291,
      "step": 763
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018084645384948723,
      "loss": 3.465,
      "step": 764
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001808175646396071,
      "loss": 3.6005,
      "step": 765
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000180788675429727,
      "loss": 3.5141,
      "step": 766
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001807597862198469,
      "loss": 3.5216,
      "step": 767
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001807308970099668,
      "loss": 3.4973,
      "step": 768
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018070200780008667,
      "loss": 3.3215,
      "step": 769
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018067311859020657,
      "loss": 3.4744,
      "step": 770
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018064422938032646,
      "loss": 3.5823,
      "step": 771
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018061534017044635,
      "loss": 3.5777,
      "step": 772
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018058645096056622,
      "loss": 3.506,
      "step": 773
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018055756175068611,
      "loss": 3.4742,
      "step": 774
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.000180528672540806,
      "loss": 3.5748,
      "step": 775
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001804997833309259,
      "loss": 3.6195,
      "step": 776
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001804708941210458,
      "loss": 3.3161,
      "step": 777
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001804420049111657,
      "loss": 3.5649,
      "step": 778
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018041311570128558,
      "loss": 3.5283,
      "step": 779
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018038422649140548,
      "loss": 3.6222,
      "step": 780
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018035533728152537,
      "loss": 3.5523,
      "step": 781
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018032644807164524,
      "loss": 3.5396,
      "step": 782
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018029755886176513,
      "loss": 3.5446,
      "step": 783
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018026866965188503,
      "loss": 3.6128,
      "step": 784
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018023978044200492,
      "loss": 3.4507,
      "step": 785
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001802108912321248,
      "loss": 3.4899,
      "step": 786
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018018200202224468,
      "loss": 3.6131,
      "step": 787
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001801531128123646,
      "loss": 3.5835,
      "step": 788
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001801242236024845,
      "loss": 3.5657,
      "step": 789
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001800953343926044,
      "loss": 3.5208,
      "step": 790
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018006644518272426,
      "loss": 3.3454,
      "step": 791
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018003755597284415,
      "loss": 3.5561,
      "step": 792
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00018000866676296404,
      "loss": 3.5146,
      "step": 793
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017997977755308394,
      "loss": 3.5223,
      "step": 794
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001799508883432038,
      "loss": 3.4998,
      "step": 795
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001799219991333237,
      "loss": 3.3141,
      "step": 796
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001798931099234436,
      "loss": 3.3507,
      "step": 797
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001798642207135635,
      "loss": 3.6048,
      "step": 798
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017983533150368338,
      "loss": 3.6106,
      "step": 799
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017980644229380327,
      "loss": 3.4502,
      "step": 800
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017977755308392317,
      "loss": 3.514,
      "step": 801
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017974866387404306,
      "loss": 3.6228,
      "step": 802
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017971977466416295,
      "loss": 3.5165,
      "step": 803
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017969088545428282,
      "loss": 3.4641,
      "step": 804
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.00017966199624440271,
      "loss": 3.5434,
      "step": 805
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001796331070345226,
      "loss": 3.5272,
      "step": 806
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0001796042178246425,
      "loss": 3.4221,
      "step": 807
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001795753286147624,
      "loss": 3.4645,
      "step": 808
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001795464394048823,
      "loss": 3.6072,
      "step": 809
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017951755019500218,
      "loss": 3.6087,
      "step": 810
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017948866098512208,
      "loss": 3.5988,
      "step": 811
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017945977177524197,
      "loss": 3.6459,
      "step": 812
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017943088256536184,
      "loss": 3.4854,
      "step": 813
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017940199335548173,
      "loss": 3.4805,
      "step": 814
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017937310414560163,
      "loss": 3.435,
      "step": 815
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017934421493572152,
      "loss": 3.5022,
      "step": 816
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001793153257258414,
      "loss": 3.6624,
      "step": 817
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017928643651596128,
      "loss": 3.5145,
      "step": 818
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001792575473060812,
      "loss": 3.4939,
      "step": 819
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001792286580962011,
      "loss": 3.4377,
      "step": 820
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017919976888632096,
      "loss": 3.5569,
      "step": 821
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017917087967644086,
      "loss": 3.5119,
      "step": 822
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017914199046656075,
      "loss": 3.5576,
      "step": 823
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017911310125668064,
      "loss": 3.4825,
      "step": 824
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017908421204680054,
      "loss": 3.5241,
      "step": 825
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001790553228369204,
      "loss": 3.4232,
      "step": 826
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001790264336270403,
      "loss": 3.429,
      "step": 827
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001789975444171602,
      "loss": 3.5378,
      "step": 828
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001789686552072801,
      "loss": 3.604,
      "step": 829
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017893976599739998,
      "loss": 3.6098,
      "step": 830
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017891087678751987,
      "loss": 3.4572,
      "step": 831
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017888198757763977,
      "loss": 3.4124,
      "step": 832
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017885309836775966,
      "loss": 3.5226,
      "step": 833
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017882420915787953,
      "loss": 3.5553,
      "step": 834
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017879531994799942,
      "loss": 3.5273,
      "step": 835
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017876643073811932,
      "loss": 3.5051,
      "step": 836
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001787375415282392,
      "loss": 3.5559,
      "step": 837
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001787086523183591,
      "loss": 3.605,
      "step": 838
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017867976310847897,
      "loss": 3.4009,
      "step": 839
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001786508738985989,
      "loss": 3.4324,
      "step": 840
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017862198468871878,
      "loss": 3.5615,
      "step": 841
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017859309547883868,
      "loss": 3.458,
      "step": 842
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017856420626895854,
      "loss": 3.4252,
      "step": 843
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017853531705907844,
      "loss": 3.4661,
      "step": 844
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017850642784919833,
      "loss": 3.4558,
      "step": 845
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017847753863931823,
      "loss": 3.5175,
      "step": 846
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017844864942943812,
      "loss": 3.5751,
      "step": 847
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000178419760219558,
      "loss": 3.5493,
      "step": 848
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017839087100967788,
      "loss": 3.6275,
      "step": 849
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001783619817997978,
      "loss": 3.5884,
      "step": 850
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001783330925899177,
      "loss": 3.519,
      "step": 851
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017830420338003756,
      "loss": 3.6041,
      "step": 852
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017827531417015746,
      "loss": 3.4436,
      "step": 853
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017824642496027735,
      "loss": 3.5615,
      "step": 854
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017821753575039724,
      "loss": 3.4636,
      "step": 855
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001781886465405171,
      "loss": 3.4988,
      "step": 856
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.000178159757330637,
      "loss": 3.4776,
      "step": 857
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001781308681207569,
      "loss": 3.5114,
      "step": 858
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001781019789108768,
      "loss": 3.5242,
      "step": 859
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017807308970099669,
      "loss": 3.3103,
      "step": 860
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017804420049111658,
      "loss": 3.5707,
      "step": 861
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017801531128123647,
      "loss": 3.5229,
      "step": 862
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017798642207135637,
      "loss": 3.548,
      "step": 863
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017795753286147626,
      "loss": 3.4077,
      "step": 864
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017792864365159613,
      "loss": 3.5361,
      "step": 865
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017789975444171602,
      "loss": 3.569,
      "step": 866
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017787086523183592,
      "loss": 3.5805,
      "step": 867
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001778419760219558,
      "loss": 3.5906,
      "step": 868
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001778130868120757,
      "loss": 3.5841,
      "step": 869
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017778419760219557,
      "loss": 3.3957,
      "step": 870
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0001777553083923155,
      "loss": 3.4768,
      "step": 871
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017772641918243538,
      "loss": 3.4825,
      "step": 872
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017769752997255528,
      "loss": 3.4059,
      "step": 873
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017766864076267514,
      "loss": 3.4614,
      "step": 874
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017763975155279504,
      "loss": 3.4827,
      "step": 875
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017761086234291493,
      "loss": 3.559,
      "step": 876
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.00017758197313303483,
      "loss": 3.5241,
      "step": 877
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001775530839231547,
      "loss": 3.5867,
      "step": 878
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001775241947132746,
      "loss": 3.5094,
      "step": 879
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017749530550339448,
      "loss": 3.3131,
      "step": 880
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017746641629351437,
      "loss": 3.4676,
      "step": 881
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017743752708363427,
      "loss": 3.6489,
      "step": 882
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017740863787375416,
      "loss": 3.5398,
      "step": 883
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017737974866387406,
      "loss": 3.5125,
      "step": 884
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017735085945399395,
      "loss": 3.5213,
      "step": 885
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017732197024411384,
      "loss": 3.4799,
      "step": 886
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001772930810342337,
      "loss": 3.4404,
      "step": 887
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001772641918243536,
      "loss": 3.478,
      "step": 888
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001772353026144735,
      "loss": 3.6084,
      "step": 889
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001772064134045934,
      "loss": 3.5364,
      "step": 890
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017717752419471329,
      "loss": 3.6162,
      "step": 891
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017714863498483318,
      "loss": 3.4878,
      "step": 892
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017711974577495307,
      "loss": 3.444,
      "step": 893
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017709085656507297,
      "loss": 3.5809,
      "step": 894
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017706196735519286,
      "loss": 3.5722,
      "step": 895
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017703307814531273,
      "loss": 3.5038,
      "step": 896
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017700418893543262,
      "loss": 3.5314,
      "step": 897
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017697529972555252,
      "loss": 3.3688,
      "step": 898
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001769464105156724,
      "loss": 3.6036,
      "step": 899
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017691752130579228,
      "loss": 3.615,
      "step": 900
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017688863209591217,
      "loss": 3.3827,
      "step": 901
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017685974288603206,
      "loss": 3.538,
      "step": 902
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017683085367615198,
      "loss": 3.4969,
      "step": 903
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017680196446627185,
      "loss": 3.5288,
      "step": 904
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017677307525639175,
      "loss": 3.6388,
      "step": 905
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017674418604651164,
      "loss": 3.5413,
      "step": 906
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017671529683663153,
      "loss": 3.5144,
      "step": 907
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017668640762675143,
      "loss": 3.3827,
      "step": 908
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001766575184168713,
      "loss": 3.4117,
      "step": 909
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001766286292069912,
      "loss": 3.5603,
      "step": 910
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017659973999711108,
      "loss": 3.446,
      "step": 911
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017657085078723097,
      "loss": 3.5327,
      "step": 912
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017654196157735087,
      "loss": 3.3939,
      "step": 913
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017651307236747076,
      "loss": 3.3279,
      "step": 914
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017648418315759066,
      "loss": 3.5229,
      "step": 915
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017645529394771055,
      "loss": 3.4836,
      "step": 916
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017642640473783044,
      "loss": 3.5151,
      "step": 917
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001763975155279503,
      "loss": 3.5036,
      "step": 918
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001763686263180702,
      "loss": 3.3613,
      "step": 919
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001763397371081901,
      "loss": 3.416,
      "step": 920
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017631084789831,
      "loss": 3.5282,
      "step": 921
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017628195868842986,
      "loss": 3.4353,
      "step": 922
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017625306947854975,
      "loss": 3.5206,
      "step": 923
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017622418026866967,
      "loss": 3.5056,
      "step": 924
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017619529105878957,
      "loss": 3.6753,
      "step": 925
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017616640184890943,
      "loss": 3.5445,
      "step": 926
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017613751263902933,
      "loss": 3.54,
      "step": 927
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017610862342914922,
      "loss": 3.5076,
      "step": 928
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017607973421926912,
      "loss": 3.341,
      "step": 929
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.000176050845009389,
      "loss": 3.5414,
      "step": 930
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017602195579950888,
      "loss": 3.5658,
      "step": 931
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017599306658962877,
      "loss": 3.4999,
      "step": 932
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017596417737974866,
      "loss": 3.5007,
      "step": 933
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017593528816986858,
      "loss": 3.5981,
      "step": 934
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017590639895998845,
      "loss": 3.4566,
      "step": 935
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017587750975010835,
      "loss": 3.6219,
      "step": 936
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017584862054022824,
      "loss": 3.4128,
      "step": 937
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017581973133034813,
      "loss": 3.3593,
      "step": 938
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017579084212046803,
      "loss": 3.4622,
      "step": 939
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001757619529105879,
      "loss": 3.6345,
      "step": 940
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0001757330637007078,
      "loss": 3.5757,
      "step": 941
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017570417449082768,
      "loss": 3.4232,
      "step": 942
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017567528528094757,
      "loss": 3.2727,
      "step": 943
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017564639607106744,
      "loss": 3.5383,
      "step": 944
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017561750686118736,
      "loss": 3.4219,
      "step": 945
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017558861765130726,
      "loss": 3.3755,
      "step": 946
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017555972844142715,
      "loss": 3.4804,
      "step": 947
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.00017553083923154702,
      "loss": 3.5328,
      "step": 948
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001755019500216669,
      "loss": 3.4785,
      "step": 949
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001754730608117868,
      "loss": 3.4262,
      "step": 950
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001754441716019067,
      "loss": 3.5284,
      "step": 951
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001754152823920266,
      "loss": 3.5174,
      "step": 952
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017538639318214646,
      "loss": 3.5466,
      "step": 953
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017535750397226635,
      "loss": 3.5061,
      "step": 954
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017532861476238627,
      "loss": 3.3919,
      "step": 955
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017529972555250617,
      "loss": 3.4249,
      "step": 956
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017527083634262603,
      "loss": 3.5617,
      "step": 957
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017524194713274593,
      "loss": 3.5016,
      "step": 958
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017521305792286582,
      "loss": 3.4626,
      "step": 959
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017518416871298572,
      "loss": 3.3283,
      "step": 960
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017515527950310558,
      "loss": 3.5361,
      "step": 961
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017512639029322548,
      "loss": 3.5714,
      "step": 962
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017509750108334537,
      "loss": 3.6037,
      "step": 963
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017506861187346526,
      "loss": 3.4375,
      "step": 964
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017503972266358516,
      "loss": 3.4013,
      "step": 965
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017501083345370505,
      "loss": 3.3052,
      "step": 966
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017498194424382495,
      "loss": 3.5285,
      "step": 967
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017495305503394484,
      "loss": 3.477,
      "step": 968
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017492416582406473,
      "loss": 3.4421,
      "step": 969
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001748952766141846,
      "loss": 3.5492,
      "step": 970
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001748663874043045,
      "loss": 3.5483,
      "step": 971
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001748374981944244,
      "loss": 3.4165,
      "step": 972
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017480860898454428,
      "loss": 3.5144,
      "step": 973
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017477971977466418,
      "loss": 3.6121,
      "step": 974
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017475083056478404,
      "loss": 3.6716,
      "step": 975
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017472194135490396,
      "loss": 3.5422,
      "step": 976
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017469305214502386,
      "loss": 3.5777,
      "step": 977
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017466416293514375,
      "loss": 3.406,
      "step": 978
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017463527372526362,
      "loss": 3.5384,
      "step": 979
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001746063845153835,
      "loss": 3.4563,
      "step": 980
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001745774953055034,
      "loss": 3.481,
      "step": 981
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001745486060956233,
      "loss": 3.3857,
      "step": 982
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017451971688574317,
      "loss": 3.43,
      "step": 983
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017449082767586306,
      "loss": 3.4773,
      "step": 984
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017446193846598295,
      "loss": 3.5036,
      "step": 985
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017443304925610285,
      "loss": 3.4525,
      "step": 986
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017440416004622274,
      "loss": 3.3283,
      "step": 987
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017437527083634263,
      "loss": 3.5363,
      "step": 988
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017434638162646253,
      "loss": 3.5391,
      "step": 989
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017431749241658242,
      "loss": 3.582,
      "step": 990
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017428860320670232,
      "loss": 3.6314,
      "step": 991
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017425971399682218,
      "loss": 3.53,
      "step": 992
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017423082478694208,
      "loss": 3.3956,
      "step": 993
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017420193557706197,
      "loss": 3.4508,
      "step": 994
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017417304636718186,
      "loss": 3.512,
      "step": 995
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017414415715730176,
      "loss": 3.3145,
      "step": 996
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017411526794742165,
      "loss": 3.5319,
      "step": 997
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017408637873754155,
      "loss": 3.4918,
      "step": 998
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017405748952766144,
      "loss": 3.4096,
      "step": 999
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017402860031778133,
      "loss": 3.6008,
      "step": 1000
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001739997111079012,
      "loss": 3.2475,
      "step": 1001
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001739708218980211,
      "loss": 3.6255,
      "step": 1002
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.000173941932688141,
      "loss": 3.491,
      "step": 1003
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017391304347826088,
      "loss": 3.3948,
      "step": 1004
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017388415426838075,
      "loss": 3.5865,
      "step": 1005
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017385526505850064,
      "loss": 3.5071,
      "step": 1006
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017382637584862056,
      "loss": 3.546,
      "step": 1007
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017379748663874046,
      "loss": 3.5369,
      "step": 1008
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017376859742886032,
      "loss": 3.4211,
      "step": 1009
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017373970821898022,
      "loss": 3.4982,
      "step": 1010
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001737108190091001,
      "loss": 3.4012,
      "step": 1011
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017368192979922,
      "loss": 3.4594,
      "step": 1012
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0001736530405893399,
      "loss": 3.5041,
      "step": 1013
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017362415137945977,
      "loss": 3.497,
      "step": 1014
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017359526216957966,
      "loss": 3.3829,
      "step": 1015
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017356637295969955,
      "loss": 3.4956,
      "step": 1016
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017353748374981945,
      "loss": 3.6394,
      "step": 1017
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.00017350859453993934,
      "loss": 3.629,
      "step": 1018
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017347970533005923,
      "loss": 3.5818,
      "step": 1019
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017345081612017913,
      "loss": 3.4767,
      "step": 1020
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017342192691029902,
      "loss": 3.6218,
      "step": 1021
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017339303770041892,
      "loss": 3.555,
      "step": 1022
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017336414849053878,
      "loss": 3.697,
      "step": 1023
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017333525928065868,
      "loss": 3.5714,
      "step": 1024
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017330637007077857,
      "loss": 3.4656,
      "step": 1025
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017327748086089846,
      "loss": 3.488,
      "step": 1026
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017324859165101833,
      "loss": 3.5766,
      "step": 1027
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017321970244113825,
      "loss": 3.6,
      "step": 1028
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017319081323125815,
      "loss": 3.662,
      "step": 1029
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017316192402137804,
      "loss": 3.4725,
      "step": 1030
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001731330348114979,
      "loss": 3.4792,
      "step": 1031
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001731041456016178,
      "loss": 3.143,
      "step": 1032
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001730752563917377,
      "loss": 3.5497,
      "step": 1033
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001730463671818576,
      "loss": 3.362,
      "step": 1034
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017301747797197748,
      "loss": 3.5541,
      "step": 1035
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017298858876209735,
      "loss": 3.4184,
      "step": 1036
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017295969955221724,
      "loss": 3.4563,
      "step": 1037
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017293081034233714,
      "loss": 3.4454,
      "step": 1038
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017290192113245706,
      "loss": 3.497,
      "step": 1039
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017287303192257692,
      "loss": 3.5273,
      "step": 1040
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017284414271269682,
      "loss": 3.3877,
      "step": 1041
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001728152535028167,
      "loss": 3.5384,
      "step": 1042
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001727863642929366,
      "loss": 3.5564,
      "step": 1043
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001727574750830565,
      "loss": 3.5066,
      "step": 1044
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017272858587317637,
      "loss": 3.3032,
      "step": 1045
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017269969666329626,
      "loss": 3.4389,
      "step": 1046
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017267080745341615,
      "loss": 3.5787,
      "step": 1047
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017264191824353605,
      "loss": 3.442,
      "step": 1048
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017261302903365594,
      "loss": 3.5452,
      "step": 1049
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017258413982377583,
      "loss": 3.3751,
      "step": 1050
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017255525061389573,
      "loss": 3.316,
      "step": 1051
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017252636140401562,
      "loss": 3.5407,
      "step": 1052
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001724974721941355,
      "loss": 3.3702,
      "step": 1053
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017246858298425538,
      "loss": 3.4369,
      "step": 1054
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017243969377437528,
      "loss": 3.438,
      "step": 1055
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017241080456449517,
      "loss": 3.4168,
      "step": 1056
    },
    {
      "epoch": 0.15,
      "eval_loss": 3.6160545349121094,
      "eval_runtime": 472.1217,
      "eval_samples_per_second": 43.393,
      "eval_steps_per_second": 14.464,
      "step": 1056
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017238191535461506,
      "loss": 3.4049,
      "step": 1057
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017235302614473493,
      "loss": 3.5561,
      "step": 1058
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017232413693485482,
      "loss": 3.5493,
      "step": 1059
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017229524772497475,
      "loss": 3.5245,
      "step": 1060
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017226635851509464,
      "loss": 3.565,
      "step": 1061
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001722374693052145,
      "loss": 3.5078,
      "step": 1062
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001722085800953344,
      "loss": 3.5372,
      "step": 1063
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001721796908854543,
      "loss": 3.4221,
      "step": 1064
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001721508016755742,
      "loss": 3.5337,
      "step": 1065
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017212191246569408,
      "loss": 3.3991,
      "step": 1066
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017209302325581395,
      "loss": 3.5483,
      "step": 1067
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017206413404593384,
      "loss": 3.4667,
      "step": 1068
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017203524483605374,
      "loss": 3.612,
      "step": 1069
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017200635562617363,
      "loss": 3.485,
      "step": 1070
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017197746641629352,
      "loss": 3.516,
      "step": 1071
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017194857720641342,
      "loss": 3.5625,
      "step": 1072
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001719196879965333,
      "loss": 3.4549,
      "step": 1073
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001718907987866532,
      "loss": 3.6844,
      "step": 1074
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017186190957677307,
      "loss": 3.5497,
      "step": 1075
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017183302036689297,
      "loss": 3.3467,
      "step": 1076
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017180413115701286,
      "loss": 3.5433,
      "step": 1077
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017177524194713275,
      "loss": 3.42,
      "step": 1078
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017174635273725265,
      "loss": 3.4682,
      "step": 1079
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017171746352737251,
      "loss": 3.4201,
      "step": 1080
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017168857431749243,
      "loss": 3.5717,
      "step": 1081
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017165968510761233,
      "loss": 3.453,
      "step": 1082
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017163079589773222,
      "loss": 3.6388,
      "step": 1083
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0001716019066878521,
      "loss": 3.5073,
      "step": 1084
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017157301747797198,
      "loss": 3.529,
      "step": 1085
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017154412826809188,
      "loss": 3.5095,
      "step": 1086
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017151523905821177,
      "loss": 3.4548,
      "step": 1087
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.00017148634984833164,
      "loss": 3.3342,
      "step": 1088
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017145746063845153,
      "loss": 3.3946,
      "step": 1089
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017142857142857143,
      "loss": 3.5994,
      "step": 1090
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017139968221869135,
      "loss": 3.5959,
      "step": 1091
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001713707930088112,
      "loss": 3.507,
      "step": 1092
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001713419037989311,
      "loss": 3.6017,
      "step": 1093
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.000171313014589051,
      "loss": 3.3432,
      "step": 1094
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001712841253791709,
      "loss": 3.4669,
      "step": 1095
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001712552361692908,
      "loss": 3.4304,
      "step": 1096
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017122634695941065,
      "loss": 3.5036,
      "step": 1097
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017119745774953055,
      "loss": 3.3196,
      "step": 1098
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017116856853965044,
      "loss": 3.4165,
      "step": 1099
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017113967932977034,
      "loss": 3.3619,
      "step": 1100
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017111079011989023,
      "loss": 3.546,
      "step": 1101
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017108190091001012,
      "loss": 3.3974,
      "step": 1102
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017105301170013002,
      "loss": 3.4734,
      "step": 1103
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001710241224902499,
      "loss": 3.4852,
      "step": 1104
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001709952332803698,
      "loss": 3.4784,
      "step": 1105
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017096634407048967,
      "loss": 3.5011,
      "step": 1106
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017093745486060957,
      "loss": 3.6002,
      "step": 1107
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017090856565072946,
      "loss": 3.4947,
      "step": 1108
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017087967644084935,
      "loss": 3.5663,
      "step": 1109
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017085078723096922,
      "loss": 3.411,
      "step": 1110
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017082189802108911,
      "loss": 3.5298,
      "step": 1111
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017079300881120904,
      "loss": 3.4539,
      "step": 1112
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017076411960132893,
      "loss": 3.3371,
      "step": 1113
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001707352303914488,
      "loss": 3.4276,
      "step": 1114
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001707063411815687,
      "loss": 3.5796,
      "step": 1115
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017067745197168858,
      "loss": 3.4728,
      "step": 1116
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017064856276180848,
      "loss": 3.4893,
      "step": 1117
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017061967355192837,
      "loss": 3.575,
      "step": 1118
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017059078434204824,
      "loss": 3.5082,
      "step": 1119
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017056189513216813,
      "loss": 3.3607,
      "step": 1120
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017053300592228803,
      "loss": 3.49,
      "step": 1121
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017050411671240792,
      "loss": 3.4581,
      "step": 1122
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001704752275025278,
      "loss": 3.5894,
      "step": 1123
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001704463382926477,
      "loss": 3.3818,
      "step": 1124
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001704174490827676,
      "loss": 3.7091,
      "step": 1125
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001703885598728875,
      "loss": 3.5166,
      "step": 1126
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001703596706630074,
      "loss": 3.547,
      "step": 1127
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017033078145312726,
      "loss": 3.392,
      "step": 1128
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017030189224324715,
      "loss": 3.6263,
      "step": 1129
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017027300303336704,
      "loss": 3.5654,
      "step": 1130
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017024411382348694,
      "loss": 3.547,
      "step": 1131
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001702152246136068,
      "loss": 3.5592,
      "step": 1132
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017018633540372672,
      "loss": 3.4518,
      "step": 1133
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017015744619384662,
      "loss": 3.55,
      "step": 1134
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001701285569839665,
      "loss": 3.3945,
      "step": 1135
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017009966777408638,
      "loss": 3.4937,
      "step": 1136
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017007077856420627,
      "loss": 3.5241,
      "step": 1137
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017004188935432617,
      "loss": 3.4706,
      "step": 1138
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00017001300014444606,
      "loss": 3.4276,
      "step": 1139
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016998411093456595,
      "loss": 3.569,
      "step": 1140
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016995522172468582,
      "loss": 3.5449,
      "step": 1141
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016992633251480571,
      "loss": 3.3942,
      "step": 1142
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001698974433049256,
      "loss": 3.5298,
      "step": 1143
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016986855409504553,
      "loss": 3.5164,
      "step": 1144
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001698396648851654,
      "loss": 3.5107,
      "step": 1145
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001698107756752853,
      "loss": 3.5288,
      "step": 1146
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016978188646540518,
      "loss": 3.2389,
      "step": 1147
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016975299725552508,
      "loss": 3.3991,
      "step": 1148
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016972410804564497,
      "loss": 3.3714,
      "step": 1149
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016969521883576484,
      "loss": 3.5075,
      "step": 1150
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016966632962588473,
      "loss": 3.4373,
      "step": 1151
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016963744041600463,
      "loss": 3.5701,
      "step": 1152
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016960855120612452,
      "loss": 3.4354,
      "step": 1153
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001695796619962444,
      "loss": 3.3822,
      "step": 1154
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001695507727863643,
      "loss": 3.3253,
      "step": 1155
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001695218835764842,
      "loss": 3.5329,
      "step": 1156
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0001694929943666041,
      "loss": 3.5455,
      "step": 1157
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.00016946410515672396,
      "loss": 3.4045,
      "step": 1158
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016943521594684386,
      "loss": 3.5935,
      "step": 1159
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016940632673696375,
      "loss": 3.4058,
      "step": 1160
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016937743752708364,
      "loss": 3.5842,
      "step": 1161
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016934854831720354,
      "loss": 3.429,
      "step": 1162
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001693196591073234,
      "loss": 3.4422,
      "step": 1163
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001692907698974433,
      "loss": 3.5908,
      "step": 1164
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016926188068756322,
      "loss": 3.6408,
      "step": 1165
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001692329914776831,
      "loss": 3.4388,
      "step": 1166
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016920410226780298,
      "loss": 3.5309,
      "step": 1167
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016917521305792287,
      "loss": 3.3306,
      "step": 1168
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016914632384804277,
      "loss": 3.5048,
      "step": 1169
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016911743463816266,
      "loss": 3.6547,
      "step": 1170
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016908854542828255,
      "loss": 3.5393,
      "step": 1171
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016905965621840242,
      "loss": 3.4526,
      "step": 1172
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016903076700852231,
      "loss": 3.5734,
      "step": 1173
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001690018777986422,
      "loss": 3.4974,
      "step": 1174
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001689729885887621,
      "loss": 3.5531,
      "step": 1175
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.000168944099378882,
      "loss": 3.6911,
      "step": 1176
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001689152101690019,
      "loss": 3.5546,
      "step": 1177
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016888632095912178,
      "loss": 3.5638,
      "step": 1178
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016885743174924168,
      "loss": 3.5208,
      "step": 1179
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016882854253936154,
      "loss": 3.5458,
      "step": 1180
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016879965332948144,
      "loss": 3.4993,
      "step": 1181
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016877076411960133,
      "loss": 3.5327,
      "step": 1182
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016874187490972123,
      "loss": 3.6325,
      "step": 1183
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016871298569984112,
      "loss": 3.5104,
      "step": 1184
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016868409648996099,
      "loss": 3.3472,
      "step": 1185
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001686552072800809,
      "loss": 3.2308,
      "step": 1186
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001686263180702008,
      "loss": 3.5748,
      "step": 1187
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001685974288603207,
      "loss": 3.5129,
      "step": 1188
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016856853965044056,
      "loss": 3.2358,
      "step": 1189
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016853965044056046,
      "loss": 3.6463,
      "step": 1190
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016851076123068035,
      "loss": 3.4268,
      "step": 1191
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016848187202080024,
      "loss": 3.5219,
      "step": 1192
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016845298281092014,
      "loss": 3.5516,
      "step": 1193
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016842409360104,
      "loss": 3.5834,
      "step": 1194
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001683952043911599,
      "loss": 3.4691,
      "step": 1195
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016836631518127982,
      "loss": 3.5435,
      "step": 1196
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016833742597139969,
      "loss": 3.4956,
      "step": 1197
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016830853676151958,
      "loss": 3.5069,
      "step": 1198
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016827964755163947,
      "loss": 3.411,
      "step": 1199
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016825075834175937,
      "loss": 3.5438,
      "step": 1200
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016822186913187926,
      "loss": 3.5069,
      "step": 1201
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016819297992199913,
      "loss": 3.5991,
      "step": 1202
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016816409071211902,
      "loss": 3.3631,
      "step": 1203
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016813520150223891,
      "loss": 3.5341,
      "step": 1204
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001681063122923588,
      "loss": 3.4704,
      "step": 1205
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001680774230824787,
      "loss": 3.3993,
      "step": 1206
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001680485338725986,
      "loss": 3.3834,
      "step": 1207
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001680196446627185,
      "loss": 3.5037,
      "step": 1208
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016799075545283838,
      "loss": 3.5927,
      "step": 1209
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016796186624295828,
      "loss": 3.625,
      "step": 1210
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016793297703307814,
      "loss": 3.5295,
      "step": 1211
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016790408782319804,
      "loss": 3.3316,
      "step": 1212
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016787519861331793,
      "loss": 3.5151,
      "step": 1213
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016784630940343783,
      "loss": 3.4836,
      "step": 1214
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016781742019355772,
      "loss": 3.5249,
      "step": 1215
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016778853098367759,
      "loss": 3.3766,
      "step": 1216
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001677596417737975,
      "loss": 3.5304,
      "step": 1217
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001677307525639174,
      "loss": 3.4835,
      "step": 1218
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016770186335403727,
      "loss": 3.7327,
      "step": 1219
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016767297414415716,
      "loss": 3.4814,
      "step": 1220
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016764408493427706,
      "loss": 3.4791,
      "step": 1221
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016761519572439695,
      "loss": 3.5619,
      "step": 1222
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016758630651451684,
      "loss": 3.5676,
      "step": 1223
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001675574173046367,
      "loss": 3.3664,
      "step": 1224
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001675285280947566,
      "loss": 3.3978,
      "step": 1225
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0001674996388848765,
      "loss": 3.5553,
      "step": 1226
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016747074967499642,
      "loss": 3.383,
      "step": 1227
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016744186046511629,
      "loss": 3.5232,
      "step": 1228
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.00016741297125523618,
      "loss": 3.3742,
      "step": 1229
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016738408204535607,
      "loss": 3.4612,
      "step": 1230
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016735519283547597,
      "loss": 3.5645,
      "step": 1231
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016732630362559586,
      "loss": 3.4287,
      "step": 1232
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016729741441571573,
      "loss": 3.5151,
      "step": 1233
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016726852520583562,
      "loss": 3.5459,
      "step": 1234
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016723963599595551,
      "loss": 3.6089,
      "step": 1235
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001672107467860754,
      "loss": 3.5577,
      "step": 1236
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016718185757619528,
      "loss": 3.6152,
      "step": 1237
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001671529683663152,
      "loss": 3.5544,
      "step": 1238
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001671240791564351,
      "loss": 3.519,
      "step": 1239
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016709518994655498,
      "loss": 3.4751,
      "step": 1240
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016706630073667485,
      "loss": 3.4899,
      "step": 1241
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016703741152679474,
      "loss": 3.3328,
      "step": 1242
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016700852231691464,
      "loss": 3.3789,
      "step": 1243
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016697963310703453,
      "loss": 3.4779,
      "step": 1244
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016695074389715443,
      "loss": 3.4797,
      "step": 1245
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001669218546872743,
      "loss": 3.4116,
      "step": 1246
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001668929654773942,
      "loss": 3.5402,
      "step": 1247
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001668640762675141,
      "loss": 3.4977,
      "step": 1248
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000166835187057634,
      "loss": 3.5899,
      "step": 1249
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016680629784775387,
      "loss": 3.4811,
      "step": 1250
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016677740863787376,
      "loss": 3.5071,
      "step": 1251
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016674851942799366,
      "loss": 3.5884,
      "step": 1252
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016671963021811355,
      "loss": 3.5284,
      "step": 1253
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016669074100823344,
      "loss": 3.3557,
      "step": 1254
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001666618517983533,
      "loss": 3.3826,
      "step": 1255
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001666329625884732,
      "loss": 3.5119,
      "step": 1256
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001666040733785931,
      "loss": 3.4175,
      "step": 1257
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000166575184168713,
      "loss": 3.4374,
      "step": 1258
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016654629495883289,
      "loss": 3.4862,
      "step": 1259
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016651740574895278,
      "loss": 3.3808,
      "step": 1260
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016648851653907267,
      "loss": 3.2596,
      "step": 1261
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016645962732919257,
      "loss": 3.4759,
      "step": 1262
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016643073811931243,
      "loss": 3.5809,
      "step": 1263
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016640184890943233,
      "loss": 3.5571,
      "step": 1264
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016637295969955222,
      "loss": 3.4506,
      "step": 1265
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016634407048967212,
      "loss": 3.496,
      "step": 1266
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.000166315181279792,
      "loss": 3.5072,
      "step": 1267
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016628629206991188,
      "loss": 3.4787,
      "step": 1268
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001662574028600318,
      "loss": 3.4776,
      "step": 1269
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001662285136501517,
      "loss": 3.4604,
      "step": 1270
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016619962444027158,
      "loss": 3.4504,
      "step": 1271
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016617073523039145,
      "loss": 3.4307,
      "step": 1272
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016614184602051134,
      "loss": 3.4938,
      "step": 1273
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016611295681063124,
      "loss": 3.3841,
      "step": 1274
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016608406760075113,
      "loss": 3.0949,
      "step": 1275
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016605517839087103,
      "loss": 3.5167,
      "step": 1276
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001660262891809909,
      "loss": 3.3093,
      "step": 1277
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001659973999711108,
      "loss": 3.3595,
      "step": 1278
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016596851076123068,
      "loss": 3.4556,
      "step": 1279
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001659396215513506,
      "loss": 3.564,
      "step": 1280
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016591073234147047,
      "loss": 3.6636,
      "step": 1281
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016588184313159036,
      "loss": 3.5133,
      "step": 1282
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016585295392171026,
      "loss": 3.5039,
      "step": 1283
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016582406471183015,
      "loss": 3.4521,
      "step": 1284
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016579517550195002,
      "loss": 3.4918,
      "step": 1285
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001657662862920699,
      "loss": 3.3308,
      "step": 1286
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001657373970821898,
      "loss": 3.5266,
      "step": 1287
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001657085078723097,
      "loss": 3.5503,
      "step": 1288
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001656796186624296,
      "loss": 3.4942,
      "step": 1289
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016565072945254949,
      "loss": 3.3854,
      "step": 1290
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016562184024266938,
      "loss": 3.5463,
      "step": 1291
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016559295103278927,
      "loss": 3.3519,
      "step": 1292
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016556406182290917,
      "loss": 3.3959,
      "step": 1293
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016553517261302903,
      "loss": 3.3282,
      "step": 1294
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016550628340314893,
      "loss": 3.4241,
      "step": 1295
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016547739419326882,
      "loss": 3.5177,
      "step": 1296
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016544850498338872,
      "loss": 3.4661,
      "step": 1297
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0001654196157735086,
      "loss": 3.4302,
      "step": 1298
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.00016539072656362848,
      "loss": 3.5691,
      "step": 1299
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016536183735374837,
      "loss": 3.47,
      "step": 1300
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001653329481438683,
      "loss": 3.4886,
      "step": 1301
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016530405893398816,
      "loss": 3.6058,
      "step": 1302
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016527516972410805,
      "loss": 3.2861,
      "step": 1303
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016524628051422794,
      "loss": 3.3734,
      "step": 1304
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016521739130434784,
      "loss": 3.581,
      "step": 1305
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016518850209446773,
      "loss": 3.4347,
      "step": 1306
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001651596128845876,
      "loss": 3.4264,
      "step": 1307
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001651307236747075,
      "loss": 3.5391,
      "step": 1308
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001651018344648274,
      "loss": 3.3605,
      "step": 1309
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016507294525494728,
      "loss": 3.4065,
      "step": 1310
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016504405604506717,
      "loss": 3.4079,
      "step": 1311
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016501516683518707,
      "loss": 3.5497,
      "step": 1312
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016498627762530696,
      "loss": 3.4103,
      "step": 1313
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016495738841542686,
      "loss": 3.4212,
      "step": 1314
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016492849920554675,
      "loss": 3.5213,
      "step": 1315
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016489960999566662,
      "loss": 3.3493,
      "step": 1316
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001648707207857865,
      "loss": 3.5141,
      "step": 1317
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001648418315759064,
      "loss": 3.3057,
      "step": 1318
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001648129423660263,
      "loss": 3.5041,
      "step": 1319
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001647840531561462,
      "loss": 3.5368,
      "step": 1320
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016475516394626606,
      "loss": 3.422,
      "step": 1321
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016472627473638598,
      "loss": 3.5449,
      "step": 1322
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016469738552650587,
      "loss": 3.5309,
      "step": 1323
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016466849631662574,
      "loss": 3.4928,
      "step": 1324
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016463960710674563,
      "loss": 3.5239,
      "step": 1325
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016461071789686553,
      "loss": 3.4141,
      "step": 1326
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016458182868698542,
      "loss": 3.5027,
      "step": 1327
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016455293947710532,
      "loss": 3.4508,
      "step": 1328
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016452405026722518,
      "loss": 3.4924,
      "step": 1329
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016449516105734508,
      "loss": 3.4917,
      "step": 1330
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016446627184746497,
      "loss": 3.5257,
      "step": 1331
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001644373826375849,
      "loss": 3.5661,
      "step": 1332
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016440849342770476,
      "loss": 3.4663,
      "step": 1333
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016437960421782465,
      "loss": 3.4531,
      "step": 1334
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016435071500794455,
      "loss": 3.4828,
      "step": 1335
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016432182579806444,
      "loss": 3.4431,
      "step": 1336
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016429293658818433,
      "loss": 3.4394,
      "step": 1337
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001642640473783042,
      "loss": 3.4915,
      "step": 1338
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001642351581684241,
      "loss": 3.5472,
      "step": 1339
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000164206268958544,
      "loss": 3.3655,
      "step": 1340
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016417737974866388,
      "loss": 3.4857,
      "step": 1341
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016414849053878377,
      "loss": 3.4078,
      "step": 1342
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016411960132890367,
      "loss": 3.5025,
      "step": 1343
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016409071211902356,
      "loss": 3.4253,
      "step": 1344
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016406182290914346,
      "loss": 3.5449,
      "step": 1345
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016403293369926332,
      "loss": 3.5968,
      "step": 1346
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016400404448938322,
      "loss": 3.2201,
      "step": 1347
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001639751552795031,
      "loss": 3.5532,
      "step": 1348
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.000163946266069623,
      "loss": 3.51,
      "step": 1349
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001639173768597429,
      "loss": 3.4733,
      "step": 1350
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016388848764986276,
      "loss": 3.603,
      "step": 1351
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016385959843998266,
      "loss": 3.4945,
      "step": 1352
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016383070923010258,
      "loss": 3.5108,
      "step": 1353
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016380182002022247,
      "loss": 3.5857,
      "step": 1354
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016377293081034234,
      "loss": 3.376,
      "step": 1355
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016374404160046223,
      "loss": 3.5224,
      "step": 1356
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016371515239058213,
      "loss": 3.4412,
      "step": 1357
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016368626318070202,
      "loss": 3.4796,
      "step": 1358
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016365737397082192,
      "loss": 3.5574,
      "step": 1359
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016362848476094178,
      "loss": 3.3681,
      "step": 1360
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016359959555106168,
      "loss": 3.462,
      "step": 1361
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016357070634118157,
      "loss": 3.3938,
      "step": 1362
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016354181713130146,
      "loss": 3.628,
      "step": 1363
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016351292792142136,
      "loss": 3.5116,
      "step": 1364
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016348403871154125,
      "loss": 3.5233,
      "step": 1365
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016345514950166115,
      "loss": 3.2241,
      "step": 1366
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.00016342626029178104,
      "loss": 3.4676,
      "step": 1367
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001633973710819009,
      "loss": 3.5448,
      "step": 1368
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0001633684818720208,
      "loss": 3.5828,
      "step": 1369
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001633395926621407,
      "loss": 3.5154,
      "step": 1370
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001633107034522606,
      "loss": 3.4541,
      "step": 1371
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016328181424238048,
      "loss": 3.4759,
      "step": 1372
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016325292503250035,
      "loss": 3.4711,
      "step": 1373
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016322403582262027,
      "loss": 3.563,
      "step": 1374
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016319514661274016,
      "loss": 3.3796,
      "step": 1375
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016316625740286006,
      "loss": 3.4592,
      "step": 1376
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016313736819297992,
      "loss": 3.4986,
      "step": 1377
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016310847898309982,
      "loss": 3.4143,
      "step": 1378
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001630795897732197,
      "loss": 3.6112,
      "step": 1379
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001630507005633396,
      "loss": 3.3096,
      "step": 1380
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001630218113534595,
      "loss": 3.4938,
      "step": 1381
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016299292214357937,
      "loss": 3.4479,
      "step": 1382
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016296403293369926,
      "loss": 3.4046,
      "step": 1383
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016293514372381915,
      "loss": 3.4154,
      "step": 1384
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016290625451393907,
      "loss": 3.5072,
      "step": 1385
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016287736530405894,
      "loss": 3.4145,
      "step": 1386
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016284847609417883,
      "loss": 3.4102,
      "step": 1387
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016281958688429873,
      "loss": 3.5517,
      "step": 1388
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016279069767441862,
      "loss": 3.4998,
      "step": 1389
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001627618084645385,
      "loss": 3.5467,
      "step": 1390
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016273291925465838,
      "loss": 3.4933,
      "step": 1391
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016270403004477828,
      "loss": 3.5235,
      "step": 1392
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016267514083489817,
      "loss": 3.5461,
      "step": 1393
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016264625162501806,
      "loss": 3.3848,
      "step": 1394
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016261736241513796,
      "loss": 3.3974,
      "step": 1395
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016258847320525785,
      "loss": 3.37,
      "step": 1396
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016255958399537775,
      "loss": 3.6363,
      "step": 1397
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016253069478549764,
      "loss": 3.4066,
      "step": 1398
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001625018055756175,
      "loss": 3.345,
      "step": 1399
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001624729163657374,
      "loss": 3.4497,
      "step": 1400
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001624440271558573,
      "loss": 3.5384,
      "step": 1401
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001624151379459772,
      "loss": 3.5991,
      "step": 1402
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016238624873609708,
      "loss": 3.4692,
      "step": 1403
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016235735952621695,
      "loss": 3.4696,
      "step": 1404
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016232847031633687,
      "loss": 3.5949,
      "step": 1405
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016229958110645676,
      "loss": 3.3968,
      "step": 1406
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016227069189657666,
      "loss": 3.5565,
      "step": 1407
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016224180268669652,
      "loss": 3.4906,
      "step": 1408
    },
    {
      "epoch": 0.2,
      "eval_loss": 3.605428695678711,
      "eval_runtime": 471.784,
      "eval_samples_per_second": 43.425,
      "eval_steps_per_second": 14.475,
      "step": 1408
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016221291347681642,
      "loss": 3.3684,
      "step": 1409
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001621840242669363,
      "loss": 3.4983,
      "step": 1410
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001621551350570562,
      "loss": 3.5674,
      "step": 1411
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016212624584717607,
      "loss": 3.3104,
      "step": 1412
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016209735663729597,
      "loss": 3.5129,
      "step": 1413
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016206846742741586,
      "loss": 3.2757,
      "step": 1414
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016203957821753575,
      "loss": 3.5879,
      "step": 1415
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016201068900765565,
      "loss": 3.5024,
      "step": 1416
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016198179979777554,
      "loss": 3.6142,
      "step": 1417
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016195291058789543,
      "loss": 3.5676,
      "step": 1418
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016192402137801533,
      "loss": 3.5295,
      "step": 1419
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016189513216813522,
      "loss": 3.5358,
      "step": 1420
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001618662429582551,
      "loss": 3.4825,
      "step": 1421
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016183735374837498,
      "loss": 3.3527,
      "step": 1422
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016180846453849488,
      "loss": 3.5883,
      "step": 1423
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016177957532861477,
      "loss": 3.407,
      "step": 1424
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016175068611873466,
      "loss": 3.5245,
      "step": 1425
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016172179690885456,
      "loss": 3.4443,
      "step": 1426
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016169290769897445,
      "loss": 3.3916,
      "step": 1427
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016166401848909435,
      "loss": 3.3604,
      "step": 1428
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001616351292792142,
      "loss": 3.475,
      "step": 1429
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001616062400693341,
      "loss": 3.6084,
      "step": 1430
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.000161577350859454,
      "loss": 3.5751,
      "step": 1431
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001615484616495739,
      "loss": 3.485,
      "step": 1432
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0001615195724396938,
      "loss": 3.5544,
      "step": 1433
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016149068322981365,
      "loss": 3.5604,
      "step": 1434
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016146179401993355,
      "loss": 3.5081,
      "step": 1435
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016143290481005344,
      "loss": 3.4819,
      "step": 1436
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016140401560017336,
      "loss": 3.4616,
      "step": 1437
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016137512639029323,
      "loss": 3.5589,
      "step": 1438
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.00016134623718041312,
      "loss": 3.4721,
      "step": 1439
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016131734797053302,
      "loss": 3.5011,
      "step": 1440
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001612884587606529,
      "loss": 3.5185,
      "step": 1441
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001612595695507728,
      "loss": 3.5604,
      "step": 1442
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016123068034089267,
      "loss": 3.577,
      "step": 1443
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016120179113101257,
      "loss": 3.3364,
      "step": 1444
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016117290192113246,
      "loss": 3.4346,
      "step": 1445
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016114401271125235,
      "loss": 3.5262,
      "step": 1446
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016111512350137225,
      "loss": 3.4903,
      "step": 1447
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016108623429149214,
      "loss": 3.2734,
      "step": 1448
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016105734508161203,
      "loss": 3.5319,
      "step": 1449
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016102845587173193,
      "loss": 3.4755,
      "step": 1450
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001609995666618518,
      "loss": 3.3927,
      "step": 1451
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001609706774519717,
      "loss": 3.523,
      "step": 1452
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016094178824209158,
      "loss": 3.448,
      "step": 1453
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016091289903221148,
      "loss": 3.5125,
      "step": 1454
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016088400982233137,
      "loss": 3.3653,
      "step": 1455
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016085512061245124,
      "loss": 3.5488,
      "step": 1456
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016082623140257113,
      "loss": 3.3857,
      "step": 1457
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016079734219269105,
      "loss": 3.4681,
      "step": 1458
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016076845298281095,
      "loss": 3.5075,
      "step": 1459
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001607395637729308,
      "loss": 3.4557,
      "step": 1460
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001607106745630507,
      "loss": 3.5808,
      "step": 1461
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001606817853531706,
      "loss": 3.6319,
      "step": 1462
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001606528961432905,
      "loss": 3.3235,
      "step": 1463
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001606240069334104,
      "loss": 3.4253,
      "step": 1464
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016059511772353025,
      "loss": 3.4317,
      "step": 1465
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016056622851365015,
      "loss": 3.3315,
      "step": 1466
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016053733930377004,
      "loss": 3.389,
      "step": 1467
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016050845009388996,
      "loss": 3.5869,
      "step": 1468
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016047956088400983,
      "loss": 3.4756,
      "step": 1469
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016045067167412972,
      "loss": 3.5431,
      "step": 1470
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016042178246424962,
      "loss": 3.5496,
      "step": 1471
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001603928932543695,
      "loss": 3.4815,
      "step": 1472
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016036400404448938,
      "loss": 3.4222,
      "step": 1473
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016033511483460927,
      "loss": 3.3973,
      "step": 1474
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016030622562472917,
      "loss": 3.5076,
      "step": 1475
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016027733641484906,
      "loss": 3.5157,
      "step": 1476
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016024844720496895,
      "loss": 3.3839,
      "step": 1477
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016021955799508882,
      "loss": 3.4792,
      "step": 1478
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016019066878520874,
      "loss": 3.4945,
      "step": 1479
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016016177957532863,
      "loss": 3.5246,
      "step": 1480
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016013289036544853,
      "loss": 3.3951,
      "step": 1481
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001601040011555684,
      "loss": 3.548,
      "step": 1482
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001600751119456883,
      "loss": 3.5085,
      "step": 1483
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016004622273580818,
      "loss": 3.3503,
      "step": 1484
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00016001733352592808,
      "loss": 3.5932,
      "step": 1485
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015998844431604797,
      "loss": 3.4716,
      "step": 1486
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015995955510616784,
      "loss": 3.6247,
      "step": 1487
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015993066589628773,
      "loss": 3.5502,
      "step": 1488
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015990177668640765,
      "loss": 3.4689,
      "step": 1489
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015987288747652755,
      "loss": 3.4942,
      "step": 1490
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001598439982666474,
      "loss": 3.4845,
      "step": 1491
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001598151090567673,
      "loss": 3.4916,
      "step": 1492
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001597862198468872,
      "loss": 3.3864,
      "step": 1493
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001597573306370071,
      "loss": 3.4457,
      "step": 1494
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015972844142712696,
      "loss": 3.3781,
      "step": 1495
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015969955221724685,
      "loss": 3.3533,
      "step": 1496
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015967066300736675,
      "loss": 3.5256,
      "step": 1497
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015964177379748664,
      "loss": 3.5881,
      "step": 1498
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015961288458760654,
      "loss": 3.4377,
      "step": 1499
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015958399537772643,
      "loss": 3.4979,
      "step": 1500
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015955510616784632,
      "loss": 3.2714,
      "step": 1501
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015952621695796622,
      "loss": 3.5759,
      "step": 1502
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0001594973277480861,
      "loss": 3.5093,
      "step": 1503
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015946843853820598,
      "loss": 3.2263,
      "step": 1504
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015943954932832587,
      "loss": 3.417,
      "step": 1505
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015941066011844577,
      "loss": 3.3492,
      "step": 1506
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015938177090856566,
      "loss": 3.4252,
      "step": 1507
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015935288169868555,
      "loss": 3.3727,
      "step": 1508
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015932399248880542,
      "loss": 3.5917,
      "step": 1509
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.00015929510327892534,
      "loss": 3.4071,
      "step": 1510
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015926621406904523,
      "loss": 3.4108,
      "step": 1511
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015923732485916513,
      "loss": 3.5325,
      "step": 1512
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000159208435649285,
      "loss": 3.4885,
      "step": 1513
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001591795464394049,
      "loss": 3.4643,
      "step": 1514
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015915065722952478,
      "loss": 3.6002,
      "step": 1515
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015912176801964468,
      "loss": 3.3506,
      "step": 1516
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015909287880976454,
      "loss": 3.3864,
      "step": 1517
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015906398959988444,
      "loss": 3.4929,
      "step": 1518
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015903510039000433,
      "loss": 3.508,
      "step": 1519
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015900621118012423,
      "loss": 3.4869,
      "step": 1520
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015897732197024412,
      "loss": 3.4724,
      "step": 1521
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.000158948432760364,
      "loss": 3.4725,
      "step": 1522
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001589195435504839,
      "loss": 3.4134,
      "step": 1523
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001588906543406038,
      "loss": 3.3603,
      "step": 1524
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001588617651307237,
      "loss": 3.4585,
      "step": 1525
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015883287592084356,
      "loss": 3.5225,
      "step": 1526
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015880398671096345,
      "loss": 3.4008,
      "step": 1527
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015877509750108335,
      "loss": 3.5919,
      "step": 1528
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015874620829120324,
      "loss": 3.5007,
      "step": 1529
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015871731908132314,
      "loss": 3.538,
      "step": 1530
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015868842987144303,
      "loss": 3.413,
      "step": 1531
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015865954066156292,
      "loss": 3.5105,
      "step": 1532
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015863065145168282,
      "loss": 3.3884,
      "step": 1533
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001586017622418027,
      "loss": 3.5485,
      "step": 1534
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015857287303192258,
      "loss": 3.5247,
      "step": 1535
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015854398382204247,
      "loss": 3.441,
      "step": 1536
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015851509461216237,
      "loss": 3.351,
      "step": 1537
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015848620540228226,
      "loss": 3.3742,
      "step": 1538
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015845731619240213,
      "loss": 3.4401,
      "step": 1539
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015842842698252202,
      "loss": 3.5104,
      "step": 1540
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015839953777264191,
      "loss": 3.3604,
      "step": 1541
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015837064856276184,
      "loss": 3.5065,
      "step": 1542
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001583417593528817,
      "loss": 3.3764,
      "step": 1543
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001583128701430016,
      "loss": 3.5065,
      "step": 1544
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001582839809331215,
      "loss": 3.4282,
      "step": 1545
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015825509172324138,
      "loss": 3.4819,
      "step": 1546
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015822620251336128,
      "loss": 3.4324,
      "step": 1547
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015819731330348114,
      "loss": 3.4524,
      "step": 1548
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015816842409360104,
      "loss": 3.5542,
      "step": 1549
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015813953488372093,
      "loss": 3.6157,
      "step": 1550
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015811064567384083,
      "loss": 3.4819,
      "step": 1551
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015808175646396072,
      "loss": 3.3803,
      "step": 1552
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001580528672540806,
      "loss": 3.2802,
      "step": 1553
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001580239780442005,
      "loss": 3.4394,
      "step": 1554
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001579950888343204,
      "loss": 3.3469,
      "step": 1555
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015796619962444027,
      "loss": 3.495,
      "step": 1556
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015793731041456016,
      "loss": 3.5226,
      "step": 1557
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015790842120468006,
      "loss": 3.4985,
      "step": 1558
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015787953199479995,
      "loss": 3.5033,
      "step": 1559
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015785064278491984,
      "loss": 3.3505,
      "step": 1560
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001578217535750397,
      "loss": 3.39,
      "step": 1561
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001577928643651596,
      "loss": 3.5471,
      "step": 1562
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015776397515527952,
      "loss": 3.4033,
      "step": 1563
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015773508594539942,
      "loss": 3.5035,
      "step": 1564
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015770619673551928,
      "loss": 3.5532,
      "step": 1565
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015767730752563918,
      "loss": 3.3897,
      "step": 1566
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015764841831575907,
      "loss": 3.5054,
      "step": 1567
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015761952910587897,
      "loss": 3.3783,
      "step": 1568
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015759063989599886,
      "loss": 3.588,
      "step": 1569
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015756175068611873,
      "loss": 3.4535,
      "step": 1570
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015753286147623862,
      "loss": 3.3799,
      "step": 1571
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015750397226635851,
      "loss": 3.396,
      "step": 1572
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015747508305647844,
      "loss": 3.385,
      "step": 1573
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001574461938465983,
      "loss": 3.1487,
      "step": 1574
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001574173046367182,
      "loss": 3.3268,
      "step": 1575
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0001573884154268381,
      "loss": 3.4515,
      "step": 1576
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015735952621695798,
      "loss": 3.5814,
      "step": 1577
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015733063700707785,
      "loss": 3.4728,
      "step": 1578
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015730174779719774,
      "loss": 3.5349,
      "step": 1579
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.00015727285858731764,
      "loss": 3.5144,
      "step": 1580
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015724396937743753,
      "loss": 3.4857,
      "step": 1581
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015721508016755743,
      "loss": 3.5092,
      "step": 1582
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001571861909576773,
      "loss": 3.3719,
      "step": 1583
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001571573017477972,
      "loss": 3.4875,
      "step": 1584
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001571284125379171,
      "loss": 3.1237,
      "step": 1585
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000157099523328037,
      "loss": 3.4557,
      "step": 1586
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015707063411815687,
      "loss": 3.3924,
      "step": 1587
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015704174490827676,
      "loss": 3.4948,
      "step": 1588
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015701285569839666,
      "loss": 3.505,
      "step": 1589
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015698396648851655,
      "loss": 3.4542,
      "step": 1590
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015695507727863644,
      "loss": 3.4866,
      "step": 1591
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001569261880687563,
      "loss": 3.435,
      "step": 1592
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001568972988588762,
      "loss": 3.4887,
      "step": 1593
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015686840964899612,
      "loss": 3.3567,
      "step": 1594
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015683952043911602,
      "loss": 3.4085,
      "step": 1595
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015681063122923588,
      "loss": 3.4561,
      "step": 1596
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015678174201935578,
      "loss": 3.5834,
      "step": 1597
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015675285280947567,
      "loss": 3.3371,
      "step": 1598
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015672396359959557,
      "loss": 3.374,
      "step": 1599
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015669507438971543,
      "loss": 3.419,
      "step": 1600
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015666618517983533,
      "loss": 3.5542,
      "step": 1601
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015663729596995522,
      "loss": 3.3417,
      "step": 1602
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015660840676007511,
      "loss": 3.3868,
      "step": 1603
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.000156579517550195,
      "loss": 3.4926,
      "step": 1604
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001565506283403149,
      "loss": 3.3727,
      "step": 1605
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001565217391304348,
      "loss": 3.4854,
      "step": 1606
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001564928499205547,
      "loss": 3.4923,
      "step": 1607
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015646396071067458,
      "loss": 3.4982,
      "step": 1608
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015643507150079445,
      "loss": 3.5484,
      "step": 1609
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015640618229091434,
      "loss": 3.4584,
      "step": 1610
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015637729308103424,
      "loss": 3.4635,
      "step": 1611
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015634840387115413,
      "loss": 3.336,
      "step": 1612
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015631951466127403,
      "loss": 3.4563,
      "step": 1613
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001562906254513939,
      "loss": 3.4882,
      "step": 1614
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001562617362415138,
      "loss": 3.4069,
      "step": 1615
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001562328470316337,
      "loss": 3.3776,
      "step": 1616
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001562039578217536,
      "loss": 3.4601,
      "step": 1617
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015617506861187347,
      "loss": 3.4655,
      "step": 1618
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015614617940199336,
      "loss": 3.4411,
      "step": 1619
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015611729019211326,
      "loss": 3.424,
      "step": 1620
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015608840098223315,
      "loss": 3.5681,
      "step": 1621
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015605951177235302,
      "loss": 3.5628,
      "step": 1622
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001560306225624729,
      "loss": 3.6016,
      "step": 1623
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001560017333525928,
      "loss": 3.3233,
      "step": 1624
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015597284414271272,
      "loss": 3.4938,
      "step": 1625
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001559439549328326,
      "loss": 3.5279,
      "step": 1626
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015591506572295249,
      "loss": 3.351,
      "step": 1627
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015588617651307238,
      "loss": 3.4287,
      "step": 1628
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015585728730319227,
      "loss": 3.5784,
      "step": 1629
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015582839809331217,
      "loss": 3.5075,
      "step": 1630
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015579950888343203,
      "loss": 3.2393,
      "step": 1631
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015577061967355193,
      "loss": 3.44,
      "step": 1632
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015574173046367182,
      "loss": 3.4562,
      "step": 1633
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015571284125379171,
      "loss": 3.4367,
      "step": 1634
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001556839520439116,
      "loss": 3.3648,
      "step": 1635
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001556550628340315,
      "loss": 3.4151,
      "step": 1636
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001556261736241514,
      "loss": 3.4663,
      "step": 1637
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001555972844142713,
      "loss": 3.3799,
      "step": 1638
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015556839520439118,
      "loss": 3.4718,
      "step": 1639
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015553950599451105,
      "loss": 3.3207,
      "step": 1640
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015551061678463094,
      "loss": 3.5074,
      "step": 1641
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015548172757475084,
      "loss": 3.4117,
      "step": 1642
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015545283836487073,
      "loss": 3.4783,
      "step": 1643
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001554239491549906,
      "loss": 3.477,
      "step": 1644
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001553950599451105,
      "loss": 3.41,
      "step": 1645
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001553661707352304,
      "loss": 3.4416,
      "step": 1646
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0001553372815253503,
      "loss": 3.5455,
      "step": 1647
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015530839231547017,
      "loss": 3.3198,
      "step": 1648
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015527950310559007,
      "loss": 3.5073,
      "step": 1649
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.00015525061389570996,
      "loss": 3.3613,
      "step": 1650
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015522172468582986,
      "loss": 3.4919,
      "step": 1651
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015519283547594975,
      "loss": 3.371,
      "step": 1652
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015516394626606962,
      "loss": 3.5187,
      "step": 1653
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001551350570561895,
      "loss": 3.3989,
      "step": 1654
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001551061678463094,
      "loss": 3.4935,
      "step": 1655
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001550772786364293,
      "loss": 3.5323,
      "step": 1656
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001550483894265492,
      "loss": 3.5203,
      "step": 1657
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015501950021666909,
      "loss": 3.395,
      "step": 1658
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015499061100678898,
      "loss": 3.4209,
      "step": 1659
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015496172179690887,
      "loss": 3.5168,
      "step": 1660
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015493283258702877,
      "loss": 3.5469,
      "step": 1661
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015490394337714863,
      "loss": 3.4716,
      "step": 1662
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015487505416726853,
      "loss": 3.4165,
      "step": 1663
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015484616495738842,
      "loss": 3.3731,
      "step": 1664
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015481727574750831,
      "loss": 3.5362,
      "step": 1665
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015478838653762818,
      "loss": 3.5687,
      "step": 1666
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001547594973277481,
      "loss": 3.5811,
      "step": 1667
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000154730608117868,
      "loss": 3.4635,
      "step": 1668
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001547017189079879,
      "loss": 3.3834,
      "step": 1669
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015467282969810776,
      "loss": 3.4669,
      "step": 1670
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015464394048822765,
      "loss": 3.4839,
      "step": 1671
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015461505127834754,
      "loss": 3.4275,
      "step": 1672
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015458616206846744,
      "loss": 3.5108,
      "step": 1673
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015455727285858733,
      "loss": 3.5041,
      "step": 1674
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001545283836487072,
      "loss": 3.4974,
      "step": 1675
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001544994944388271,
      "loss": 3.3863,
      "step": 1676
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000154470605228947,
      "loss": 3.4526,
      "step": 1677
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001544417160190669,
      "loss": 3.5786,
      "step": 1678
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015441282680918677,
      "loss": 3.3061,
      "step": 1679
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015438393759930667,
      "loss": 3.4446,
      "step": 1680
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015435504838942656,
      "loss": 3.486,
      "step": 1681
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015432615917954646,
      "loss": 3.4333,
      "step": 1682
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015429726996966632,
      "loss": 3.3177,
      "step": 1683
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015426838075978622,
      "loss": 3.5748,
      "step": 1684
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001542394915499061,
      "loss": 3.495,
      "step": 1685
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.000154210602340026,
      "loss": 3.4927,
      "step": 1686
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001541817131301459,
      "loss": 3.4502,
      "step": 1687
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001541528239202658,
      "loss": 3.4559,
      "step": 1688
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015412393471038569,
      "loss": 3.3388,
      "step": 1689
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015409504550050558,
      "loss": 3.437,
      "step": 1690
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015406615629062547,
      "loss": 3.3642,
      "step": 1691
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015403726708074534,
      "loss": 3.4297,
      "step": 1692
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015400837787086523,
      "loss": 3.3197,
      "step": 1693
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015397948866098513,
      "loss": 3.5311,
      "step": 1694
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015395059945110502,
      "loss": 3.3404,
      "step": 1695
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015392171024122492,
      "loss": 3.4481,
      "step": 1696
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015389282103134478,
      "loss": 3.4041,
      "step": 1697
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015386393182146468,
      "loss": 3.4734,
      "step": 1698
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001538350426115846,
      "loss": 3.5042,
      "step": 1699
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001538061534017045,
      "loss": 3.3428,
      "step": 1700
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015377726419182436,
      "loss": 3.5199,
      "step": 1701
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015374837498194425,
      "loss": 3.3486,
      "step": 1702
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015371948577206414,
      "loss": 3.3362,
      "step": 1703
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015369059656218404,
      "loss": 3.3543,
      "step": 1704
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001536617073523039,
      "loss": 3.4948,
      "step": 1705
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001536328181424238,
      "loss": 3.4326,
      "step": 1706
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001536039289325437,
      "loss": 3.3218,
      "step": 1707
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001535750397226636,
      "loss": 3.4999,
      "step": 1708
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015354615051278348,
      "loss": 3.4222,
      "step": 1709
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015351726130290337,
      "loss": 3.0983,
      "step": 1710
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015348837209302327,
      "loss": 3.447,
      "step": 1711
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015345948288314316,
      "loss": 3.4486,
      "step": 1712
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015343059367326306,
      "loss": 3.398,
      "step": 1713
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015340170446338292,
      "loss": 3.3366,
      "step": 1714
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015337281525350282,
      "loss": 3.5535,
      "step": 1715
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001533439260436227,
      "loss": 3.4162,
      "step": 1716
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001533150368337426,
      "loss": 3.3429,
      "step": 1717
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0001532861476238625,
      "loss": 3.4877,
      "step": 1718
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015325725841398236,
      "loss": 3.4085,
      "step": 1719
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.00015322836920410229,
      "loss": 3.5294,
      "step": 1720
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015319947999422218,
      "loss": 3.3905,
      "step": 1721
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015317059078434207,
      "loss": 3.4958,
      "step": 1722
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015314170157446194,
      "loss": 3.4739,
      "step": 1723
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015311281236458183,
      "loss": 3.4774,
      "step": 1724
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015308392315470173,
      "loss": 3.4016,
      "step": 1725
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015305503394482162,
      "loss": 3.4329,
      "step": 1726
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001530261447349415,
      "loss": 3.4621,
      "step": 1727
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015299725552506138,
      "loss": 3.4899,
      "step": 1728
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015296836631518128,
      "loss": 3.397,
      "step": 1729
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001529394771053012,
      "loss": 3.4158,
      "step": 1730
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015291058789542106,
      "loss": 3.455,
      "step": 1731
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015288169868554096,
      "loss": 3.5955,
      "step": 1732
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015285280947566085,
      "loss": 3.4992,
      "step": 1733
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015282392026578074,
      "loss": 3.1957,
      "step": 1734
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015279503105590064,
      "loss": 3.4251,
      "step": 1735
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001527661418460205,
      "loss": 3.3176,
      "step": 1736
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001527372526361404,
      "loss": 3.5188,
      "step": 1737
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001527083634262603,
      "loss": 3.4461,
      "step": 1738
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001526794742163802,
      "loss": 3.3749,
      "step": 1739
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015265058500650008,
      "loss": 3.4516,
      "step": 1740
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015262169579661997,
      "loss": 3.5357,
      "step": 1741
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015259280658673987,
      "loss": 3.4015,
      "step": 1742
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015256391737685976,
      "loss": 3.5292,
      "step": 1743
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015253502816697966,
      "loss": 3.4333,
      "step": 1744
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015250613895709952,
      "loss": 3.3426,
      "step": 1745
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015247724974721942,
      "loss": 3.4477,
      "step": 1746
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001524483605373393,
      "loss": 3.4709,
      "step": 1747
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001524194713274592,
      "loss": 3.4584,
      "step": 1748
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015239058211757907,
      "loss": 3.448,
      "step": 1749
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015236169290769896,
      "loss": 3.4082,
      "step": 1750
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015233280369781889,
      "loss": 3.4354,
      "step": 1751
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015230391448793878,
      "loss": 3.4206,
      "step": 1752
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015227502527805865,
      "loss": 3.3987,
      "step": 1753
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015224613606817854,
      "loss": 3.3487,
      "step": 1754
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015221724685829843,
      "loss": 3.5518,
      "step": 1755
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015218835764841833,
      "loss": 3.341,
      "step": 1756
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015215946843853822,
      "loss": 3.4783,
      "step": 1757
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001521305792286581,
      "loss": 3.3948,
      "step": 1758
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015210169001877798,
      "loss": 3.2644,
      "step": 1759
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015207280080889788,
      "loss": 3.4853,
      "step": 1760
    },
    {
      "epoch": 0.25,
      "eval_loss": 3.5950064659118652,
      "eval_runtime": 471.9526,
      "eval_samples_per_second": 43.409,
      "eval_steps_per_second": 14.47,
      "step": 1760
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015204391159901777,
      "loss": 3.4051,
      "step": 1761
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015201502238913766,
      "loss": 3.3596,
      "step": 1762
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015198613317925756,
      "loss": 3.4631,
      "step": 1763
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015195724396937745,
      "loss": 3.416,
      "step": 1764
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015192835475949735,
      "loss": 3.4119,
      "step": 1765
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015189946554961724,
      "loss": 3.3318,
      "step": 1766
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001518705763397371,
      "loss": 3.4591,
      "step": 1767
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.000151841687129857,
      "loss": 3.483,
      "step": 1768
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001518127979199769,
      "loss": 3.4671,
      "step": 1769
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001517839087100968,
      "loss": 3.3936,
      "step": 1770
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015175501950021665,
      "loss": 3.4857,
      "step": 1771
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015172613029033657,
      "loss": 3.3232,
      "step": 1772
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015169724108045647,
      "loss": 3.4324,
      "step": 1773
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015166835187057636,
      "loss": 3.4477,
      "step": 1774
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015163946266069623,
      "loss": 3.593,
      "step": 1775
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015161057345081612,
      "loss": 3.2523,
      "step": 1776
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015158168424093602,
      "loss": 3.4394,
      "step": 1777
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001515527950310559,
      "loss": 3.4362,
      "step": 1778
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001515239058211758,
      "loss": 3.5046,
      "step": 1779
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015149501661129567,
      "loss": 3.35,
      "step": 1780
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015146612740141556,
      "loss": 3.4668,
      "step": 1781
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015143723819153546,
      "loss": 3.2848,
      "step": 1782
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015140834898165538,
      "loss": 3.5605,
      "step": 1783
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015137945977177525,
      "loss": 3.3905,
      "step": 1784
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015135057056189514,
      "loss": 3.4506,
      "step": 1785
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015132168135201503,
      "loss": 3.4539,
      "step": 1786
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015129279214213493,
      "loss": 3.532,
      "step": 1787
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015126390293225482,
      "loss": 3.4072,
      "step": 1788
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0001512350137223747,
      "loss": 3.5478,
      "step": 1789
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.00015120612451249458,
      "loss": 3.5377,
      "step": 1790
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015117723530261448,
      "loss": 3.3438,
      "step": 1791
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015114834609273437,
      "loss": 3.4662,
      "step": 1792
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015111945688285426,
      "loss": 3.3743,
      "step": 1793
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015109056767297416,
      "loss": 3.4603,
      "step": 1794
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015106167846309405,
      "loss": 3.5661,
      "step": 1795
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015103278925321395,
      "loss": 3.6666,
      "step": 1796
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001510039000433338,
      "loss": 3.5021,
      "step": 1797
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001509750108334537,
      "loss": 3.4549,
      "step": 1798
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001509461216235736,
      "loss": 3.4754,
      "step": 1799
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001509172324136935,
      "loss": 3.3963,
      "step": 1800
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001508883432038134,
      "loss": 3.5162,
      "step": 1801
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015085945399393325,
      "loss": 3.4586,
      "step": 1802
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015083056478405317,
      "loss": 3.4553,
      "step": 1803
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015080167557417307,
      "loss": 3.4765,
      "step": 1804
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015077278636429296,
      "loss": 3.5084,
      "step": 1805
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015074389715441283,
      "loss": 3.3074,
      "step": 1806
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015071500794453272,
      "loss": 3.4487,
      "step": 1807
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015068611873465262,
      "loss": 3.4417,
      "step": 1808
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001506572295247725,
      "loss": 3.364,
      "step": 1809
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001506283403148924,
      "loss": 3.4581,
      "step": 1810
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015059945110501227,
      "loss": 3.4035,
      "step": 1811
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015057056189513217,
      "loss": 3.4665,
      "step": 1812
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015054167268525206,
      "loss": 3.4461,
      "step": 1813
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015051278347537195,
      "loss": 3.4914,
      "step": 1814
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015048389426549185,
      "loss": 3.3571,
      "step": 1815
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015045500505561174,
      "loss": 3.3196,
      "step": 1816
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015042611584573163,
      "loss": 3.4052,
      "step": 1817
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015039722663585153,
      "loss": 3.341,
      "step": 1818
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001503683374259714,
      "loss": 3.5346,
      "step": 1819
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001503394482160913,
      "loss": 3.2565,
      "step": 1820
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015031055900621118,
      "loss": 3.5817,
      "step": 1821
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015028166979633108,
      "loss": 3.5139,
      "step": 1822
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015025278058645097,
      "loss": 3.5845,
      "step": 1823
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015022389137657086,
      "loss": 3.4028,
      "step": 1824
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015019500216669076,
      "loss": 3.555,
      "step": 1825
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015016611295681065,
      "loss": 3.3109,
      "step": 1826
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00015013722374693055,
      "loss": 3.5086,
      "step": 1827
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001501083345370504,
      "loss": 3.3717,
      "step": 1828
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001500794453271703,
      "loss": 3.375,
      "step": 1829
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001500505561172902,
      "loss": 3.4272,
      "step": 1830
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001500216669074101,
      "loss": 3.5006,
      "step": 1831
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014999277769752996,
      "loss": 3.4847,
      "step": 1832
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014996388848764985,
      "loss": 3.4681,
      "step": 1833
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014993499927776975,
      "loss": 3.342,
      "step": 1834
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014990611006788967,
      "loss": 3.3611,
      "step": 1835
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014987722085800954,
      "loss": 3.3995,
      "step": 1836
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014984833164812943,
      "loss": 3.4514,
      "step": 1837
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014981944243824932,
      "loss": 3.5673,
      "step": 1838
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014979055322836922,
      "loss": 3.4605,
      "step": 1839
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001497616640184891,
      "loss": 3.5023,
      "step": 1840
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014973277480860898,
      "loss": 3.3265,
      "step": 1841
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014970388559872887,
      "loss": 3.4755,
      "step": 1842
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014967499638884877,
      "loss": 3.3242,
      "step": 1843
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014964610717896866,
      "loss": 3.3834,
      "step": 1844
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014961721796908855,
      "loss": 3.4341,
      "step": 1845
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014958832875920845,
      "loss": 3.5213,
      "step": 1846
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014955943954932834,
      "loss": 3.543,
      "step": 1847
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014953055033944823,
      "loss": 3.4442,
      "step": 1848
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014950166112956813,
      "loss": 3.5597,
      "step": 1849
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000149472771919688,
      "loss": 3.3681,
      "step": 1850
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001494438827098079,
      "loss": 3.3805,
      "step": 1851
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014941499349992778,
      "loss": 3.4571,
      "step": 1852
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014938610429004768,
      "loss": 3.444,
      "step": 1853
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014935721508016754,
      "loss": 3.4871,
      "step": 1854
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014932832587028744,
      "loss": 3.4849,
      "step": 1855
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014929943666040736,
      "loss": 3.2477,
      "step": 1856
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014927054745052725,
      "loss": 3.3903,
      "step": 1857
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.00014924165824064712,
      "loss": 3.5671,
      "step": 1858
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.000149212769030767,
      "loss": 3.2997,
      "step": 1859
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001491838798208869,
      "loss": 3.5099,
      "step": 1860
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0001491549906110068,
      "loss": 3.3259,
      "step": 1861
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001491261014011267,
      "loss": 3.4265,
      "step": 1862
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014909721219124656,
      "loss": 3.357,
      "step": 1863
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014906832298136645,
      "loss": 3.6089,
      "step": 1864
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014903943377148635,
      "loss": 3.4923,
      "step": 1865
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014901054456160627,
      "loss": 3.5827,
      "step": 1866
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014898165535172614,
      "loss": 3.4838,
      "step": 1867
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014895276614184603,
      "loss": 3.3623,
      "step": 1868
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014892387693196592,
      "loss": 3.4234,
      "step": 1869
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014889498772208582,
      "loss": 3.42,
      "step": 1870
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001488660985122057,
      "loss": 3.4147,
      "step": 1871
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014883720930232558,
      "loss": 3.4886,
      "step": 1872
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014880832009244547,
      "loss": 3.2949,
      "step": 1873
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014877943088256537,
      "loss": 3.4814,
      "step": 1874
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014875054167268526,
      "loss": 3.4084,
      "step": 1875
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014872165246280513,
      "loss": 3.4247,
      "step": 1876
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014869276325292505,
      "loss": 3.4906,
      "step": 1877
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014866387404304494,
      "loss": 3.4889,
      "step": 1878
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014863498483316483,
      "loss": 3.4885,
      "step": 1879
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001486060956232847,
      "loss": 3.4409,
      "step": 1880
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001485772064134046,
      "loss": 3.2834,
      "step": 1881
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001485483172035245,
      "loss": 3.4257,
      "step": 1882
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014851942799364438,
      "loss": 3.5979,
      "step": 1883
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014849053878376428,
      "loss": 3.3264,
      "step": 1884
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014846164957388414,
      "loss": 3.4631,
      "step": 1885
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014843276036400404,
      "loss": 3.4964,
      "step": 1886
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014840387115412396,
      "loss": 3.5215,
      "step": 1887
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014837498194424385,
      "loss": 3.4671,
      "step": 1888
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014834609273436372,
      "loss": 3.4544,
      "step": 1889
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001483172035244836,
      "loss": 3.4728,
      "step": 1890
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001482883143146035,
      "loss": 3.5478,
      "step": 1891
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001482594251047234,
      "loss": 3.3491,
      "step": 1892
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001482305358948433,
      "loss": 3.5396,
      "step": 1893
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014820164668496316,
      "loss": 3.4655,
      "step": 1894
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014817275747508305,
      "loss": 3.3832,
      "step": 1895
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014814386826520295,
      "loss": 3.485,
      "step": 1896
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014811497905532284,
      "loss": 3.5032,
      "step": 1897
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014808608984544274,
      "loss": 3.4827,
      "step": 1898
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014805720063556263,
      "loss": 3.4626,
      "step": 1899
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014802831142568252,
      "loss": 3.3802,
      "step": 1900
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014799942221580242,
      "loss": 3.5136,
      "step": 1901
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014797053300592228,
      "loss": 3.5037,
      "step": 1902
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014794164379604218,
      "loss": 3.5207,
      "step": 1903
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014791275458616207,
      "loss": 3.4731,
      "step": 1904
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014788386537628197,
      "loss": 3.4823,
      "step": 1905
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014785497616640186,
      "loss": 3.4794,
      "step": 1906
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014782608695652173,
      "loss": 3.4531,
      "step": 1907
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014779719774664165,
      "loss": 3.3548,
      "step": 1908
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014776830853676154,
      "loss": 3.4536,
      "step": 1909
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014773941932688143,
      "loss": 3.5259,
      "step": 1910
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001477105301170013,
      "loss": 3.4277,
      "step": 1911
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001476816409071212,
      "loss": 3.3827,
      "step": 1912
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001476527516972411,
      "loss": 3.4711,
      "step": 1913
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014762386248736098,
      "loss": 3.4863,
      "step": 1914
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014759497327748088,
      "loss": 3.5021,
      "step": 1915
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014756608406760074,
      "loss": 3.3306,
      "step": 1916
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014753719485772064,
      "loss": 3.5057,
      "step": 1917
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014750830564784053,
      "loss": 3.4257,
      "step": 1918
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014747941643796042,
      "loss": 3.3859,
      "step": 1919
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014745052722808032,
      "loss": 3.488,
      "step": 1920
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001474216380182002,
      "loss": 3.2959,
      "step": 1921
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0001473927488083201,
      "loss": 3.5709,
      "step": 1922
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014736385959844,
      "loss": 3.3738,
      "step": 1923
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014733497038855987,
      "loss": 3.4031,
      "step": 1924
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014730608117867976,
      "loss": 3.3902,
      "step": 1925
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014727719196879965,
      "loss": 3.3969,
      "step": 1926
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014724830275891955,
      "loss": 3.3293,
      "step": 1927
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014721941354903944,
      "loss": 3.442,
      "step": 1928
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014719052433915934,
      "loss": 3.4116,
      "step": 1929
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014716163512927923,
      "loss": 3.5437,
      "step": 1930
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.00014713274591939912,
      "loss": 3.4743,
      "step": 1931
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014710385670951902,
      "loss": 3.281,
      "step": 1932
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014707496749963888,
      "loss": 3.4105,
      "step": 1933
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014704607828975878,
      "loss": 3.6074,
      "step": 1934
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014701718907987867,
      "loss": 3.5347,
      "step": 1935
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014698829986999857,
      "loss": 3.4239,
      "step": 1936
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014695941066011846,
      "loss": 3.4338,
      "step": 1937
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014693052145023833,
      "loss": 3.4685,
      "step": 1938
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014690163224035822,
      "loss": 3.3967,
      "step": 1939
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014687274303047814,
      "loss": 3.4177,
      "step": 1940
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.000146843853820598,
      "loss": 3.4269,
      "step": 1941
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001468149646107179,
      "loss": 3.4652,
      "step": 1942
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001467860754008378,
      "loss": 3.4627,
      "step": 1943
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001467571861909577,
      "loss": 3.5027,
      "step": 1944
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014672829698107758,
      "loss": 3.5213,
      "step": 1945
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014669940777119745,
      "loss": 3.5287,
      "step": 1946
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014667051856131734,
      "loss": 3.5134,
      "step": 1947
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014664162935143724,
      "loss": 3.439,
      "step": 1948
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014661274014155713,
      "loss": 3.4804,
      "step": 1949
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014658385093167703,
      "loss": 3.2259,
      "step": 1950
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014655496172179692,
      "loss": 3.4906,
      "step": 1951
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001465260725119168,
      "loss": 3.4134,
      "step": 1952
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001464971833020367,
      "loss": 3.5899,
      "step": 1953
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001464682940921566,
      "loss": 3.4322,
      "step": 1954
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014643940488227647,
      "loss": 3.2959,
      "step": 1955
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014641051567239636,
      "loss": 3.3999,
      "step": 1956
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014638162646251625,
      "loss": 3.444,
      "step": 1957
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014635273725263615,
      "loss": 3.4835,
      "step": 1958
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014632384804275602,
      "loss": 3.4171,
      "step": 1959
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001462949588328759,
      "loss": 3.3343,
      "step": 1960
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014626606962299583,
      "loss": 3.4857,
      "step": 1961
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014623718041311572,
      "loss": 3.2975,
      "step": 1962
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001462082912032356,
      "loss": 3.3762,
      "step": 1963
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014617940199335548,
      "loss": 3.5087,
      "step": 1964
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014615051278347538,
      "loss": 3.5289,
      "step": 1965
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014612162357359527,
      "loss": 3.4127,
      "step": 1966
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014609273436371517,
      "loss": 3.4432,
      "step": 1967
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014606384515383503,
      "loss": 3.4058,
      "step": 1968
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014603495594395493,
      "loss": 3.5486,
      "step": 1969
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014600606673407482,
      "loss": 3.491,
      "step": 1970
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014597717752419474,
      "loss": 3.3858,
      "step": 1971
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001459482883143146,
      "loss": 3.3609,
      "step": 1972
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001459193991044345,
      "loss": 3.4312,
      "step": 1973
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001458905098945544,
      "loss": 3.403,
      "step": 1974
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001458616206846743,
      "loss": 3.2376,
      "step": 1975
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014583273147479418,
      "loss": 3.4582,
      "step": 1976
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014580384226491405,
      "loss": 3.5404,
      "step": 1977
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014577495305503394,
      "loss": 3.5013,
      "step": 1978
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014574606384515384,
      "loss": 3.4204,
      "step": 1979
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014571717463527373,
      "loss": 3.3112,
      "step": 1980
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014568828542539363,
      "loss": 3.562,
      "step": 1981
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014565939621551352,
      "loss": 3.2597,
      "step": 1982
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001456305070056334,
      "loss": 3.4568,
      "step": 1983
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001456016177957533,
      "loss": 3.5787,
      "step": 1984
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014557272858587317,
      "loss": 3.4654,
      "step": 1985
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014554383937599307,
      "loss": 3.5657,
      "step": 1986
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014551495016611296,
      "loss": 3.4461,
      "step": 1987
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014548606095623285,
      "loss": 3.5129,
      "step": 1988
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014545717174635275,
      "loss": 3.5618,
      "step": 1989
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014542828253647262,
      "loss": 3.324,
      "step": 1990
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001453993933265925,
      "loss": 3.3,
      "step": 1991
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014537050411671243,
      "loss": 3.395,
      "step": 1992
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014534161490683232,
      "loss": 3.5426,
      "step": 1993
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0001453127256969522,
      "loss": 3.4614,
      "step": 1994
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014528383648707208,
      "loss": 3.3017,
      "step": 1995
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014525494727719198,
      "loss": 3.4191,
      "step": 1996
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014522605806731187,
      "loss": 3.4152,
      "step": 1997
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014519716885743177,
      "loss": 3.2961,
      "step": 1998
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014516827964755163,
      "loss": 3.3462,
      "step": 1999
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014513939043767153,
      "loss": 3.1992,
      "step": 2000
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.00014511050122779142,
      "loss": 3.4572,
      "step": 2001
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014508161201791134,
      "loss": 3.3327,
      "step": 2002
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001450527228080312,
      "loss": 3.4493,
      "step": 2003
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001450238335981511,
      "loss": 3.4782,
      "step": 2004
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.000144994944388271,
      "loss": 3.5013,
      "step": 2005
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001449660551783909,
      "loss": 3.3044,
      "step": 2006
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014493716596851076,
      "loss": 3.65,
      "step": 2007
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014490827675863065,
      "loss": 3.4377,
      "step": 2008
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014487938754875054,
      "loss": 3.2571,
      "step": 2009
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014485049833887044,
      "loss": 3.4659,
      "step": 2010
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014482160912899033,
      "loss": 3.3477,
      "step": 2011
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001447927199191102,
      "loss": 3.4271,
      "step": 2012
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014476383070923012,
      "loss": 3.5187,
      "step": 2013
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014473494149935,
      "loss": 3.4195,
      "step": 2014
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001447060522894699,
      "loss": 3.4035,
      "step": 2015
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014467716307958977,
      "loss": 3.3835,
      "step": 2016
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014464827386970967,
      "loss": 3.5223,
      "step": 2017
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014461938465982956,
      "loss": 3.412,
      "step": 2018
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014459049544994946,
      "loss": 3.3719,
      "step": 2019
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014456160624006935,
      "loss": 3.5132,
      "step": 2020
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014453271703018922,
      "loss": 3.4407,
      "step": 2021
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001445038278203091,
      "loss": 3.5427,
      "step": 2022
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014447493861042903,
      "loss": 3.3198,
      "step": 2023
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001444460494005489,
      "loss": 3.4871,
      "step": 2024
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001444171601906688,
      "loss": 3.3314,
      "step": 2025
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014438827098078868,
      "loss": 3.4372,
      "step": 2026
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014435938177090858,
      "loss": 3.2933,
      "step": 2027
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014433049256102847,
      "loss": 3.4497,
      "step": 2028
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014430160335114834,
      "loss": 3.501,
      "step": 2029
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014427271414126823,
      "loss": 3.3996,
      "step": 2030
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014424382493138813,
      "loss": 3.5048,
      "step": 2031
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014421493572150802,
      "loss": 3.4347,
      "step": 2032
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014418604651162791,
      "loss": 3.2714,
      "step": 2033
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001441571573017478,
      "loss": 3.5157,
      "step": 2034
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001441282680918677,
      "loss": 3.4685,
      "step": 2035
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001440993788819876,
      "loss": 3.4565,
      "step": 2036
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001440704896721075,
      "loss": 3.4407,
      "step": 2037
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014404160046222736,
      "loss": 3.4024,
      "step": 2038
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014401271125234725,
      "loss": 3.4478,
      "step": 2039
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014398382204246714,
      "loss": 3.5271,
      "step": 2040
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014395493283258704,
      "loss": 3.5119,
      "step": 2041
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014392604362270693,
      "loss": 3.2467,
      "step": 2042
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001438971544128268,
      "loss": 3.504,
      "step": 2043
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014386826520294672,
      "loss": 3.4109,
      "step": 2044
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001438393759930666,
      "loss": 3.3956,
      "step": 2045
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014381048678318648,
      "loss": 3.4761,
      "step": 2046
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014378159757330637,
      "loss": 3.5453,
      "step": 2047
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014375270836342627,
      "loss": 3.317,
      "step": 2048
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014372381915354616,
      "loss": 3.4624,
      "step": 2049
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014369492994366606,
      "loss": 3.3344,
      "step": 2050
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014366604073378592,
      "loss": 3.3991,
      "step": 2051
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014363715152390582,
      "loss": 3.3717,
      "step": 2052
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001436082623140257,
      "loss": 3.4737,
      "step": 2053
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001435793731041456,
      "loss": 3.4253,
      "step": 2054
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001435504838942655,
      "loss": 3.25,
      "step": 2055
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001435215946843854,
      "loss": 3.4574,
      "step": 2056
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014349270547450529,
      "loss": 3.2754,
      "step": 2057
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014346381626462518,
      "loss": 3.391,
      "step": 2058
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014343492705474507,
      "loss": 3.5016,
      "step": 2059
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014340603784486494,
      "loss": 3.3742,
      "step": 2060
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014337714863498483,
      "loss": 3.3986,
      "step": 2061
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014334825942510473,
      "loss": 3.2683,
      "step": 2062
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014331937021522462,
      "loss": 3.3845,
      "step": 2063
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014329048100534451,
      "loss": 3.2227,
      "step": 2064
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001432615917954644,
      "loss": 3.3658,
      "step": 2065
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001432327025855843,
      "loss": 3.455,
      "step": 2066
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0001432038133757042,
      "loss": 3.3727,
      "step": 2067
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014317492416582406,
      "loss": 3.2789,
      "step": 2068
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014314603495594396,
      "loss": 3.4471,
      "step": 2069
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014311714574606385,
      "loss": 3.412,
      "step": 2070
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.00014308825653618374,
      "loss": 3.573,
      "step": 2071
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014305936732630364,
      "loss": 3.3721,
      "step": 2072
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001430304781164235,
      "loss": 3.5033,
      "step": 2073
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001430015889065434,
      "loss": 3.4696,
      "step": 2074
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001429726996966633,
      "loss": 3.4221,
      "step": 2075
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001429438104867832,
      "loss": 3.5093,
      "step": 2076
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014291492127690308,
      "loss": 3.4657,
      "step": 2077
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014288603206702297,
      "loss": 3.3822,
      "step": 2078
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014285714285714287,
      "loss": 3.4654,
      "step": 2079
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014282825364726276,
      "loss": 3.3517,
      "step": 2080
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014279936443738266,
      "loss": 3.4709,
      "step": 2081
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014277047522750252,
      "loss": 3.4689,
      "step": 2082
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014274158601762242,
      "loss": 3.4866,
      "step": 2083
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001427126968077423,
      "loss": 3.266,
      "step": 2084
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001426838075978622,
      "loss": 3.4905,
      "step": 2085
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001426549183879821,
      "loss": 3.5003,
      "step": 2086
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.000142626029178102,
      "loss": 3.4256,
      "step": 2087
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014259713996822189,
      "loss": 3.4049,
      "step": 2088
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014256825075834178,
      "loss": 3.5358,
      "step": 2089
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014253936154846165,
      "loss": 3.4838,
      "step": 2090
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014251047233858154,
      "loss": 3.3808,
      "step": 2091
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014248158312870143,
      "loss": 3.3242,
      "step": 2092
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014245269391882133,
      "loss": 3.336,
      "step": 2093
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014242380470894122,
      "loss": 3.3955,
      "step": 2094
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001423949154990611,
      "loss": 3.5222,
      "step": 2095
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014236602628918098,
      "loss": 3.4184,
      "step": 2096
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001423371370793009,
      "loss": 3.4204,
      "step": 2097
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001423082478694208,
      "loss": 3.4188,
      "step": 2098
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014227935865954066,
      "loss": 3.4482,
      "step": 2099
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014225046944966056,
      "loss": 3.4424,
      "step": 2100
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014222158023978045,
      "loss": 3.3663,
      "step": 2101
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014219269102990034,
      "loss": 3.3478,
      "step": 2102
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014216380182002024,
      "loss": 3.2278,
      "step": 2103
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001421349126101401,
      "loss": 3.4806,
      "step": 2104
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014210602340026,
      "loss": 3.5108,
      "step": 2105
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001420771341903799,
      "loss": 3.5468,
      "step": 2106
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014204824498049981,
      "loss": 3.4712,
      "step": 2107
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014201935577061968,
      "loss": 3.4026,
      "step": 2108
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014199046656073957,
      "loss": 3.4801,
      "step": 2109
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014196157735085947,
      "loss": 3.4062,
      "step": 2110
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014193268814097936,
      "loss": 3.4245,
      "step": 2111
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014190379893109923,
      "loss": 3.475,
      "step": 2112
    },
    {
      "epoch": 0.3,
      "eval_loss": 3.580632209777832,
      "eval_runtime": 471.7483,
      "eval_samples_per_second": 43.428,
      "eval_steps_per_second": 14.476,
      "step": 2112
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014187490972121912,
      "loss": 3.4145,
      "step": 2113
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014184602051133902,
      "loss": 3.3238,
      "step": 2114
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001418171313014589,
      "loss": 3.3204,
      "step": 2115
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001417882420915788,
      "loss": 3.2878,
      "step": 2116
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014175935288169867,
      "loss": 3.4137,
      "step": 2117
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001417304636718186,
      "loss": 3.4031,
      "step": 2118
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014170157446193849,
      "loss": 3.4523,
      "step": 2119
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014167268525205838,
      "loss": 3.3077,
      "step": 2120
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014164379604217825,
      "loss": 3.4158,
      "step": 2121
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014161490683229814,
      "loss": 3.3561,
      "step": 2122
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014158601762241803,
      "loss": 3.3478,
      "step": 2123
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014155712841253793,
      "loss": 3.4943,
      "step": 2124
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014152823920265782,
      "loss": 3.387,
      "step": 2125
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001414993499927777,
      "loss": 3.5522,
      "step": 2126
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014147046078289758,
      "loss": 3.4068,
      "step": 2127
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001414415715730175,
      "loss": 3.3601,
      "step": 2128
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001414126823631374,
      "loss": 3.4129,
      "step": 2129
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014138379315325726,
      "loss": 3.4063,
      "step": 2130
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014135490394337716,
      "loss": 3.2681,
      "step": 2131
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014132601473349705,
      "loss": 3.456,
      "step": 2132
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014129712552361694,
      "loss": 3.4014,
      "step": 2133
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001412682363137368,
      "loss": 3.2214,
      "step": 2134
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001412393471038567,
      "loss": 3.2428,
      "step": 2135
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001412104578939766,
      "loss": 3.3887,
      "step": 2136
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001411815686840965,
      "loss": 3.5157,
      "step": 2137
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0001411526794742164,
      "loss": 3.4656,
      "step": 2138
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014112379026433628,
      "loss": 3.4659,
      "step": 2139
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014109490105445617,
      "loss": 3.4634,
      "step": 2140
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014106601184457607,
      "loss": 3.4335,
      "step": 2141
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.00014103712263469596,
      "loss": 3.4129,
      "step": 2142
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014100823342481583,
      "loss": 3.38,
      "step": 2143
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014097934421493572,
      "loss": 3.5137,
      "step": 2144
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014095045500505562,
      "loss": 3.5617,
      "step": 2145
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001409215657951755,
      "loss": 3.395,
      "step": 2146
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001408926765852954,
      "loss": 3.5115,
      "step": 2147
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014086378737541527,
      "loss": 3.4942,
      "step": 2148
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001408348981655352,
      "loss": 3.5737,
      "step": 2149
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014080600895565509,
      "loss": 3.5475,
      "step": 2150
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014077711974577495,
      "loss": 3.3564,
      "step": 2151
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014074823053589485,
      "loss": 3.4725,
      "step": 2152
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014071934132601474,
      "loss": 3.407,
      "step": 2153
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014069045211613463,
      "loss": 3.5516,
      "step": 2154
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014066156290625453,
      "loss": 3.5385,
      "step": 2155
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001406326736963744,
      "loss": 3.35,
      "step": 2156
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001406037844864943,
      "loss": 3.3215,
      "step": 2157
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014057489527661418,
      "loss": 3.4591,
      "step": 2158
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014054600606673408,
      "loss": 3.4866,
      "step": 2159
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014051711685685397,
      "loss": 3.4184,
      "step": 2160
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014048822764697386,
      "loss": 3.3715,
      "step": 2161
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014045933843709376,
      "loss": 3.5308,
      "step": 2162
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014043044922721365,
      "loss": 3.3307,
      "step": 2163
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014040156001733354,
      "loss": 3.4918,
      "step": 2164
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001403726708074534,
      "loss": 3.2267,
      "step": 2165
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001403437815975733,
      "loss": 3.3426,
      "step": 2166
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001403148923876932,
      "loss": 3.3965,
      "step": 2167
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001402860031778131,
      "loss": 3.5155,
      "step": 2168
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000140257113967933,
      "loss": 3.3793,
      "step": 2169
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014022822475805288,
      "loss": 3.5449,
      "step": 2170
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014019933554817277,
      "loss": 3.4593,
      "step": 2171
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014017044633829267,
      "loss": 3.3046,
      "step": 2172
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014014155712841254,
      "loss": 3.1801,
      "step": 2173
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014011266791853243,
      "loss": 3.4525,
      "step": 2174
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014008377870865232,
      "loss": 3.4199,
      "step": 2175
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00014005488949877222,
      "loss": 3.5184,
      "step": 2176
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001400260002888921,
      "loss": 3.4598,
      "step": 2177
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013999711107901198,
      "loss": 3.4862,
      "step": 2178
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013996822186913187,
      "loss": 3.4429,
      "step": 2179
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001399393326592518,
      "loss": 3.5418,
      "step": 2180
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013991044344937169,
      "loss": 3.4268,
      "step": 2181
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013988155423949155,
      "loss": 3.2958,
      "step": 2182
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013985266502961145,
      "loss": 3.4215,
      "step": 2183
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013982377581973134,
      "loss": 3.3225,
      "step": 2184
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013979488660985123,
      "loss": 3.3378,
      "step": 2185
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013976599739997113,
      "loss": 3.4642,
      "step": 2186
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.000139737108190091,
      "loss": 3.4571,
      "step": 2187
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001397082189802109,
      "loss": 3.3121,
      "step": 2188
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013967932977033078,
      "loss": 3.4845,
      "step": 2189
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013965044056045068,
      "loss": 3.3634,
      "step": 2190
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013962155135057057,
      "loss": 3.3096,
      "step": 2191
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013959266214069046,
      "loss": 3.3263,
      "step": 2192
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013956377293081036,
      "loss": 3.12,
      "step": 2193
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013953488372093025,
      "loss": 3.5795,
      "step": 2194
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013950599451105012,
      "loss": 3.4364,
      "step": 2195
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013947710530117,
      "loss": 3.3533,
      "step": 2196
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001394482160912899,
      "loss": 3.5455,
      "step": 2197
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001394193268814098,
      "loss": 3.3429,
      "step": 2198
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001393904376715297,
      "loss": 3.4561,
      "step": 2199
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013936154846164956,
      "loss": 3.2739,
      "step": 2200
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013933265925176948,
      "loss": 3.3866,
      "step": 2201
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013930377004188937,
      "loss": 3.369,
      "step": 2202
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013927488083200927,
      "loss": 3.4938,
      "step": 2203
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013924599162212914,
      "loss": 3.4248,
      "step": 2204
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013921710241224903,
      "loss": 3.4534,
      "step": 2205
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013918821320236892,
      "loss": 3.4099,
      "step": 2206
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013915932399248882,
      "loss": 3.4923,
      "step": 2207
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0001391304347826087,
      "loss": 3.4412,
      "step": 2208
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013910154557272858,
      "loss": 3.4822,
      "step": 2209
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013907265636284847,
      "loss": 3.3436,
      "step": 2210
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013904376715296836,
      "loss": 3.3572,
      "step": 2211
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.00013901487794308829,
      "loss": 3.3505,
      "step": 2212
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013898598873320815,
      "loss": 3.408,
      "step": 2213
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013895709952332805,
      "loss": 3.3081,
      "step": 2214
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013892821031344794,
      "loss": 3.3781,
      "step": 2215
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013889932110356783,
      "loss": 3.4703,
      "step": 2216
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001388704318936877,
      "loss": 3.6122,
      "step": 2217
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001388415426838076,
      "loss": 3.3728,
      "step": 2218
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001388126534739275,
      "loss": 3.5355,
      "step": 2219
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013878376426404738,
      "loss": 3.4331,
      "step": 2220
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013875487505416728,
      "loss": 3.377,
      "step": 2221
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013872598584428717,
      "loss": 3.5025,
      "step": 2222
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013869709663440706,
      "loss": 3.5106,
      "step": 2223
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013866820742452696,
      "loss": 3.5311,
      "step": 2224
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013863931821464685,
      "loss": 3.4397,
      "step": 2225
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013861042900476672,
      "loss": 3.3628,
      "step": 2226
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001385815397948866,
      "loss": 3.4772,
      "step": 2227
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001385526505850065,
      "loss": 3.5013,
      "step": 2228
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001385237613751264,
      "loss": 3.5671,
      "step": 2229
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001384948721652463,
      "loss": 3.498,
      "step": 2230
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013846598295536616,
      "loss": 3.3978,
      "step": 2231
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013843709374548605,
      "loss": 3.3056,
      "step": 2232
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013840820453560597,
      "loss": 3.3484,
      "step": 2233
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013837931532572587,
      "loss": 3.4505,
      "step": 2234
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013835042611584574,
      "loss": 3.4581,
      "step": 2235
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013832153690596563,
      "loss": 3.4996,
      "step": 2236
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013829264769608552,
      "loss": 3.2491,
      "step": 2237
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013826375848620542,
      "loss": 3.3702,
      "step": 2238
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013823486927632528,
      "loss": 3.4381,
      "step": 2239
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013820598006644518,
      "loss": 3.4453,
      "step": 2240
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013817709085656507,
      "loss": 3.4205,
      "step": 2241
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013814820164668497,
      "loss": 3.5151,
      "step": 2242
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013811931243680486,
      "loss": 3.4931,
      "step": 2243
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013809042322692475,
      "loss": 3.4466,
      "step": 2244
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013806153401704465,
      "loss": 3.3914,
      "step": 2245
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013803264480716454,
      "loss": 3.4542,
      "step": 2246
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013800375559728443,
      "loss": 3.3532,
      "step": 2247
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001379748663874043,
      "loss": 3.4223,
      "step": 2248
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001379459771775242,
      "loss": 3.3836,
      "step": 2249
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001379170879676441,
      "loss": 3.4089,
      "step": 2250
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013788819875776398,
      "loss": 3.4003,
      "step": 2251
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013785930954788388,
      "loss": 3.3875,
      "step": 2252
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013783042033800374,
      "loss": 3.512,
      "step": 2253
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013780153112812366,
      "loss": 3.3483,
      "step": 2254
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013777264191824356,
      "loss": 3.4381,
      "step": 2255
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013774375270836345,
      "loss": 3.4253,
      "step": 2256
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013771486349848332,
      "loss": 3.4555,
      "step": 2257
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001376859742886032,
      "loss": 3.4515,
      "step": 2258
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001376570850787231,
      "loss": 3.4872,
      "step": 2259
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.000137628195868843,
      "loss": 3.2705,
      "step": 2260
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013759930665896287,
      "loss": 3.4221,
      "step": 2261
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013757041744908276,
      "loss": 3.3614,
      "step": 2262
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013754152823920265,
      "loss": 3.3684,
      "step": 2263
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013751263902932258,
      "loss": 3.4306,
      "step": 2264
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013748374981944244,
      "loss": 3.2772,
      "step": 2265
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013745486060956234,
      "loss": 3.4369,
      "step": 2266
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013742597139968223,
      "loss": 3.3247,
      "step": 2267
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013739708218980212,
      "loss": 3.5047,
      "step": 2268
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013736819297992202,
      "loss": 3.3833,
      "step": 2269
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013733930377004188,
      "loss": 3.2106,
      "step": 2270
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013731041456016178,
      "loss": 3.3727,
      "step": 2271
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013728152535028167,
      "loss": 3.3919,
      "step": 2272
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013725263614040157,
      "loss": 3.3544,
      "step": 2273
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013722374693052146,
      "loss": 3.3658,
      "step": 2274
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013719485772064135,
      "loss": 3.4047,
      "step": 2275
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013716596851076125,
      "loss": 3.3533,
      "step": 2276
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013713707930088114,
      "loss": 3.3707,
      "step": 2277
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013710819009100103,
      "loss": 3.2264,
      "step": 2278
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001370793008811209,
      "loss": 3.39,
      "step": 2279
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001370504116712408,
      "loss": 3.457,
      "step": 2280
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0001370215224613607,
      "loss": 3.2658,
      "step": 2281
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.00013699263325148058,
      "loss": 3.3738,
      "step": 2282
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013696374404160045,
      "loss": 3.3054,
      "step": 2283
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013693485483172034,
      "loss": 3.4937,
      "step": 2284
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013690596562184026,
      "loss": 3.3848,
      "step": 2285
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013687707641196016,
      "loss": 3.5145,
      "step": 2286
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013684818720208002,
      "loss": 3.4478,
      "step": 2287
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013681929799219992,
      "loss": 3.5072,
      "step": 2288
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001367904087823198,
      "loss": 3.4196,
      "step": 2289
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001367615195724397,
      "loss": 3.3121,
      "step": 2290
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001367326303625596,
      "loss": 3.4513,
      "step": 2291
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013670374115267947,
      "loss": 3.4903,
      "step": 2292
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013667485194279936,
      "loss": 3.39,
      "step": 2293
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013664596273291925,
      "loss": 3.4293,
      "step": 2294
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013661707352303915,
      "loss": 3.3947,
      "step": 2295
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013658818431315904,
      "loss": 3.376,
      "step": 2296
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013655929510327894,
      "loss": 3.3409,
      "step": 2297
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013653040589339883,
      "loss": 3.2811,
      "step": 2298
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013650151668351872,
      "loss": 3.4879,
      "step": 2299
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001364726274736386,
      "loss": 3.5118,
      "step": 2300
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013644373826375848,
      "loss": 3.4129,
      "step": 2301
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013641484905387838,
      "loss": 3.4909,
      "step": 2302
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013638595984399827,
      "loss": 3.3344,
      "step": 2303
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013635707063411817,
      "loss": 3.2765,
      "step": 2304
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013632818142423803,
      "loss": 3.4694,
      "step": 2305
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013629929221435795,
      "loss": 3.3702,
      "step": 2306
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013627040300447785,
      "loss": 3.3879,
      "step": 2307
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013624151379459774,
      "loss": 3.4246,
      "step": 2308
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001362126245847176,
      "loss": 3.3766,
      "step": 2309
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001361837353748375,
      "loss": 3.4544,
      "step": 2310
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001361548461649574,
      "loss": 3.3527,
      "step": 2311
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001361259569550773,
      "loss": 3.3898,
      "step": 2312
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013609706774519718,
      "loss": 3.3793,
      "step": 2313
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013606817853531705,
      "loss": 3.3719,
      "step": 2314
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013603928932543694,
      "loss": 3.5055,
      "step": 2315
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013601040011555684,
      "loss": 3.4176,
      "step": 2316
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013598151090567676,
      "loss": 3.1551,
      "step": 2317
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013595262169579662,
      "loss": 3.4452,
      "step": 2318
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013592373248591652,
      "loss": 3.351,
      "step": 2319
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001358948432760364,
      "loss": 3.3997,
      "step": 2320
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001358659540661563,
      "loss": 3.2877,
      "step": 2321
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013583706485627617,
      "loss": 3.3394,
      "step": 2322
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013580817564639607,
      "loss": 3.5017,
      "step": 2323
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013577928643651596,
      "loss": 3.306,
      "step": 2324
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013575039722663585,
      "loss": 3.2135,
      "step": 2325
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013572150801675575,
      "loss": 3.1845,
      "step": 2326
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013569261880687564,
      "loss": 3.1219,
      "step": 2327
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013566372959699554,
      "loss": 3.4218,
      "step": 2328
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013563484038711543,
      "loss": 3.425,
      "step": 2329
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013560595117723532,
      "loss": 3.3029,
      "step": 2330
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001355770619673552,
      "loss": 3.3517,
      "step": 2331
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013554817275747508,
      "loss": 3.3517,
      "step": 2332
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013551928354759498,
      "loss": 3.444,
      "step": 2333
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013549039433771487,
      "loss": 3.3205,
      "step": 2334
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013546150512783477,
      "loss": 3.4313,
      "step": 2335
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013543261591795463,
      "loss": 3.4755,
      "step": 2336
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013540372670807453,
      "loss": 3.3698,
      "step": 2337
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013537483749819445,
      "loss": 3.5406,
      "step": 2338
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013534594828831434,
      "loss": 3.4552,
      "step": 2339
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001353170590784342,
      "loss": 3.3954,
      "step": 2340
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001352881698685541,
      "loss": 3.3485,
      "step": 2341
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000135259280658674,
      "loss": 3.4813,
      "step": 2342
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001352303914487939,
      "loss": 3.4158,
      "step": 2343
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013520150223891376,
      "loss": 3.416,
      "step": 2344
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013517261302903365,
      "loss": 3.4413,
      "step": 2345
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013514372381915354,
      "loss": 3.4087,
      "step": 2346
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013511483460927344,
      "loss": 3.4527,
      "step": 2347
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013508594539939333,
      "loss": 3.4639,
      "step": 2348
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013505705618951322,
      "loss": 3.2068,
      "step": 2349
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.00013502816697963312,
      "loss": 3.5575,
      "step": 2350
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.000134999277769753,
      "loss": 3.1795,
      "step": 2351
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0001349703885598729,
      "loss": 3.4778,
      "step": 2352
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013494149934999277,
      "loss": 3.4111,
      "step": 2353
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013491261014011267,
      "loss": 3.1266,
      "step": 2354
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013488372093023256,
      "loss": 3.3929,
      "step": 2355
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013485483172035245,
      "loss": 3.3329,
      "step": 2356
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013482594251047235,
      "loss": 3.4474,
      "step": 2357
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013479705330059222,
      "loss": 3.4609,
      "step": 2358
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013476816409071214,
      "loss": 3.3622,
      "step": 2359
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013473927488083203,
      "loss": 3.3481,
      "step": 2360
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013471038567095192,
      "loss": 3.4563,
      "step": 2361
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001346814964610718,
      "loss": 3.295,
      "step": 2362
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013465260725119168,
      "loss": 3.4981,
      "step": 2363
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013462371804131158,
      "loss": 3.4384,
      "step": 2364
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013459482883143147,
      "loss": 3.3174,
      "step": 2365
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013456593962155134,
      "loss": 3.431,
      "step": 2366
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013453705041167123,
      "loss": 3.4281,
      "step": 2367
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013450816120179113,
      "loss": 3.4862,
      "step": 2368
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013447927199191105,
      "loss": 3.4451,
      "step": 2369
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013445038278203091,
      "loss": 3.3016,
      "step": 2370
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001344214935721508,
      "loss": 3.3705,
      "step": 2371
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001343926043622707,
      "loss": 3.3258,
      "step": 2372
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001343637151523906,
      "loss": 3.3826,
      "step": 2373
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001343348259425105,
      "loss": 3.3718,
      "step": 2374
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013430593673263036,
      "loss": 3.4815,
      "step": 2375
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013427704752275025,
      "loss": 3.3265,
      "step": 2376
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013424815831287014,
      "loss": 3.4326,
      "step": 2377
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013421926910299004,
      "loss": 3.3151,
      "step": 2378
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013419037989310993,
      "loss": 3.324,
      "step": 2379
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013416149068322983,
      "loss": 3.2986,
      "step": 2380
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013413260147334972,
      "loss": 3.3388,
      "step": 2381
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001341037122634696,
      "loss": 3.2725,
      "step": 2382
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001340748230535895,
      "loss": 3.4595,
      "step": 2383
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013404593384370937,
      "loss": 3.4725,
      "step": 2384
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013401704463382927,
      "loss": 3.4385,
      "step": 2385
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013398815542394916,
      "loss": 3.3203,
      "step": 2386
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013395926621406905,
      "loss": 3.5133,
      "step": 2387
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013393037700418892,
      "loss": 3.4137,
      "step": 2388
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013390148779430882,
      "loss": 3.4752,
      "step": 2389
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013387259858442874,
      "loss": 3.3159,
      "step": 2390
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013384370937454863,
      "loss": 3.37,
      "step": 2391
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001338148201646685,
      "loss": 3.4084,
      "step": 2392
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001337859309547884,
      "loss": 3.4467,
      "step": 2393
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013375704174490828,
      "loss": 3.3266,
      "step": 2394
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013372815253502818,
      "loss": 3.2987,
      "step": 2395
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013369926332514807,
      "loss": 3.3875,
      "step": 2396
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013367037411526794,
      "loss": 3.417,
      "step": 2397
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013364148490538783,
      "loss": 3.4194,
      "step": 2398
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013361259569550773,
      "loss": 3.4125,
      "step": 2399
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013358370648562765,
      "loss": 3.3874,
      "step": 2400
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013355481727574751,
      "loss": 3.4618,
      "step": 2401
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001335259280658674,
      "loss": 3.4174,
      "step": 2402
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001334970388559873,
      "loss": 3.3945,
      "step": 2403
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001334681496461072,
      "loss": 3.4469,
      "step": 2404
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001334392604362271,
      "loss": 3.5016,
      "step": 2405
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013341037122634696,
      "loss": 3.4477,
      "step": 2406
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013338148201646685,
      "loss": 3.3503,
      "step": 2407
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013335259280658674,
      "loss": 3.3697,
      "step": 2408
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013332370359670664,
      "loss": 3.2187,
      "step": 2409
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001332948143868265,
      "loss": 3.3541,
      "step": 2410
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013326592517694643,
      "loss": 3.4176,
      "step": 2411
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013323703596706632,
      "loss": 3.3427,
      "step": 2412
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001332081467571862,
      "loss": 3.3391,
      "step": 2413
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013317925754730608,
      "loss": 3.392,
      "step": 2414
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013315036833742597,
      "loss": 3.3353,
      "step": 2415
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013312147912754587,
      "loss": 3.4673,
      "step": 2416
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013309258991766576,
      "loss": 3.315,
      "step": 2417
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013306370070778565,
      "loss": 3.1815,
      "step": 2418
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013303481149790552,
      "loss": 3.3533,
      "step": 2419
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013300592228802542,
      "loss": 3.2504,
      "step": 2420
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013297703307814534,
      "loss": 3.4406,
      "step": 2421
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.00013294814386826523,
      "loss": 3.3511,
      "step": 2422
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0001329192546583851,
      "loss": 3.4214,
      "step": 2423
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000132890365448505,
      "loss": 3.3204,
      "step": 2424
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013286147623862488,
      "loss": 3.4374,
      "step": 2425
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013283258702874478,
      "loss": 3.405,
      "step": 2426
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013280369781886465,
      "loss": 3.2408,
      "step": 2427
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013277480860898454,
      "loss": 3.3318,
      "step": 2428
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013274591939910443,
      "loss": 3.4731,
      "step": 2429
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013271703018922433,
      "loss": 3.4582,
      "step": 2430
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013268814097934422,
      "loss": 3.4412,
      "step": 2431
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013265925176946411,
      "loss": 3.3804,
      "step": 2432
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.000132630362559584,
      "loss": 3.4339,
      "step": 2433
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001326014733497039,
      "loss": 3.2806,
      "step": 2434
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001325725841398238,
      "loss": 3.2909,
      "step": 2435
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013254369492994366,
      "loss": 3.4289,
      "step": 2436
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013251480572006356,
      "loss": 3.1086,
      "step": 2437
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013248591651018345,
      "loss": 3.4366,
      "step": 2438
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013245702730030334,
      "loss": 3.4952,
      "step": 2439
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013242813809042324,
      "loss": 3.2876,
      "step": 2440
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001323992488805431,
      "loss": 3.3862,
      "step": 2441
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013237035967066303,
      "loss": 3.3508,
      "step": 2442
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013234147046078292,
      "loss": 3.5075,
      "step": 2443
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001323125812509028,
      "loss": 3.0807,
      "step": 2444
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013228369204102268,
      "loss": 3.3371,
      "step": 2445
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013225480283114257,
      "loss": 3.3085,
      "step": 2446
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013222591362126247,
      "loss": 3.4499,
      "step": 2447
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013219702441138236,
      "loss": 3.474,
      "step": 2448
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013216813520150223,
      "loss": 3.4154,
      "step": 2449
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013213924599162212,
      "loss": 3.5276,
      "step": 2450
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013211035678174202,
      "loss": 3.432,
      "step": 2451
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001320814675718619,
      "loss": 3.3362,
      "step": 2452
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001320525783619818,
      "loss": 3.3559,
      "step": 2453
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001320236891521017,
      "loss": 3.2714,
      "step": 2454
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001319947999422216,
      "loss": 3.2934,
      "step": 2455
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013196591073234148,
      "loss": 3.3839,
      "step": 2456
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013193702152246138,
      "loss": 3.3512,
      "step": 2457
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013190813231258125,
      "loss": 3.2705,
      "step": 2458
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013187924310270114,
      "loss": 3.4098,
      "step": 2459
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013185035389282103,
      "loss": 3.2906,
      "step": 2460
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013182146468294093,
      "loss": 3.389,
      "step": 2461
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013179257547306082,
      "loss": 3.354,
      "step": 2462
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013176368626318071,
      "loss": 3.4475,
      "step": 2463
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001317347970533006,
      "loss": 3.2476,
      "step": 2464
    },
    {
      "epoch": 0.35,
      "eval_loss": 3.5674850940704346,
      "eval_runtime": 473.0302,
      "eval_samples_per_second": 43.31,
      "eval_steps_per_second": 14.437,
      "step": 2464
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001317059078434205,
      "loss": 3.4002,
      "step": 2465
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001316770186335404,
      "loss": 3.429,
      "step": 2466
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013164812942366026,
      "loss": 3.3282,
      "step": 2467
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013161924021378016,
      "loss": 3.1434,
      "step": 2468
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013159035100390005,
      "loss": 3.3893,
      "step": 2469
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013156146179401994,
      "loss": 3.3807,
      "step": 2470
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001315325725841398,
      "loss": 3.374,
      "step": 2471
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001315036833742597,
      "loss": 3.3385,
      "step": 2472
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001314747941643796,
      "loss": 3.3285,
      "step": 2473
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013144590495449952,
      "loss": 3.4922,
      "step": 2474
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013141701574461939,
      "loss": 3.4045,
      "step": 2475
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013138812653473928,
      "loss": 3.3272,
      "step": 2476
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013135923732485917,
      "loss": 3.373,
      "step": 2477
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013133034811497907,
      "loss": 3.3136,
      "step": 2478
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013130145890509896,
      "loss": 3.2604,
      "step": 2479
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013127256969521883,
      "loss": 3.4602,
      "step": 2480
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013124368048533872,
      "loss": 3.361,
      "step": 2481
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013121479127545862,
      "loss": 3.3301,
      "step": 2482
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001311859020655785,
      "loss": 3.3257,
      "step": 2483
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001311570128556984,
      "loss": 3.3816,
      "step": 2484
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001311281236458183,
      "loss": 3.4853,
      "step": 2485
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001310992344359382,
      "loss": 3.4588,
      "step": 2486
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013107034522605809,
      "loss": 3.369,
      "step": 2487
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013104145601617798,
      "loss": 3.3041,
      "step": 2488
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013101256680629785,
      "loss": 3.4367,
      "step": 2489
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013098367759641774,
      "loss": 3.2189,
      "step": 2490
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013095478838653763,
      "loss": 3.2695,
      "step": 2491
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.00013092589917665753,
      "loss": 3.4266,
      "step": 2492
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0001308970099667774,
      "loss": 3.4851,
      "step": 2493
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001308681207568973,
      "loss": 3.4867,
      "step": 2494
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001308392315470172,
      "loss": 3.5863,
      "step": 2495
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001308103423371371,
      "loss": 3.3697,
      "step": 2496
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013078145312725697,
      "loss": 3.4082,
      "step": 2497
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013075256391737686,
      "loss": 3.4489,
      "step": 2498
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013072367470749676,
      "loss": 3.4199,
      "step": 2499
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013069478549761665,
      "loss": 3.354,
      "step": 2500
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013066589628773654,
      "loss": 3.4063,
      "step": 2501
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001306370070778564,
      "loss": 3.4643,
      "step": 2502
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001306081178679763,
      "loss": 3.3565,
      "step": 2503
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001305792286580962,
      "loss": 3.4586,
      "step": 2504
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013055033944821612,
      "loss": 3.4305,
      "step": 2505
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013052145023833599,
      "loss": 3.4861,
      "step": 2506
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013049256102845588,
      "loss": 3.2312,
      "step": 2507
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013046367181857577,
      "loss": 3.2431,
      "step": 2508
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013043478260869567,
      "loss": 3.5262,
      "step": 2509
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013040589339881556,
      "loss": 3.3597,
      "step": 2510
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013037700418893543,
      "loss": 3.3521,
      "step": 2511
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013034811497905532,
      "loss": 3.2393,
      "step": 2512
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013031922576917522,
      "loss": 3.494,
      "step": 2513
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001302903365592951,
      "loss": 3.5039,
      "step": 2514
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013026144734941498,
      "loss": 3.3433,
      "step": 2515
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001302325581395349,
      "loss": 3.4088,
      "step": 2516
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001302036689296548,
      "loss": 3.2848,
      "step": 2517
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013017477971977469,
      "loss": 3.4788,
      "step": 2518
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013014589050989455,
      "loss": 3.2091,
      "step": 2519
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013011700130001445,
      "loss": 3.3265,
      "step": 2520
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013008811209013434,
      "loss": 3.4179,
      "step": 2521
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013005922288025423,
      "loss": 3.4583,
      "step": 2522
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00013003033367037413,
      "loss": 3.4588,
      "step": 2523
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000130001444460494,
      "loss": 3.3187,
      "step": 2524
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001299725552506139,
      "loss": 3.3585,
      "step": 2525
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001299436660407338,
      "loss": 3.2796,
      "step": 2526
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001299147768308537,
      "loss": 3.3668,
      "step": 2527
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012988588762097357,
      "loss": 3.2533,
      "step": 2528
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012985699841109346,
      "loss": 3.5402,
      "step": 2529
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012982810920121336,
      "loss": 3.336,
      "step": 2530
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012979921999133325,
      "loss": 3.4623,
      "step": 2531
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012977033078145314,
      "loss": 3.4201,
      "step": 2532
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.000129741441571573,
      "loss": 3.3559,
      "step": 2533
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001297125523616929,
      "loss": 3.3933,
      "step": 2534
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001296836631518128,
      "loss": 3.4174,
      "step": 2535
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001296547739419327,
      "loss": 3.3951,
      "step": 2536
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001296258847320526,
      "loss": 3.2896,
      "step": 2537
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012959699552217248,
      "loss": 3.4864,
      "step": 2538
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012956810631229237,
      "loss": 3.4116,
      "step": 2539
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012953921710241227,
      "loss": 3.4263,
      "step": 2540
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012951032789253213,
      "loss": 3.4341,
      "step": 2541
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012948143868265203,
      "loss": 3.2924,
      "step": 2542
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012945254947277192,
      "loss": 3.3333,
      "step": 2543
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012942366026289182,
      "loss": 3.3644,
      "step": 2544
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001293947710530117,
      "loss": 3.3937,
      "step": 2545
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012936588184313158,
      "loss": 3.3669,
      "step": 2546
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001293369926332515,
      "loss": 3.4506,
      "step": 2547
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001293081034233714,
      "loss": 3.4005,
      "step": 2548
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012927921421349129,
      "loss": 3.347,
      "step": 2549
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012925032500361115,
      "loss": 3.3494,
      "step": 2550
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012922143579373105,
      "loss": 3.3931,
      "step": 2551
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012919254658385094,
      "loss": 3.4486,
      "step": 2552
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012916365737397083,
      "loss": 3.3955,
      "step": 2553
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001291347681640907,
      "loss": 3.323,
      "step": 2554
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001291058789542106,
      "loss": 3.4641,
      "step": 2555
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0001290769897443305,
      "loss": 3.4741,
      "step": 2556
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012904810053445038,
      "loss": 3.3527,
      "step": 2557
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012901921132457028,
      "loss": 3.4107,
      "step": 2558
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012899032211469017,
      "loss": 3.3844,
      "step": 2559
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012896143290481006,
      "loss": 3.3398,
      "step": 2560
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012893254369492996,
      "loss": 3.3155,
      "step": 2561
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012890365448504985,
      "loss": 3.4171,
      "step": 2562
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.00012887476527516972,
      "loss": 3.3677,
      "step": 2563
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001288458760652896,
      "loss": 3.1637,
      "step": 2564
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001288169868554095,
      "loss": 3.3501,
      "step": 2565
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001287880976455294,
      "loss": 3.2347,
      "step": 2566
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001287592084356493,
      "loss": 3.3296,
      "step": 2567
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001287303192257692,
      "loss": 3.4267,
      "step": 2568
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012870143001588908,
      "loss": 3.3299,
      "step": 2569
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012867254080600897,
      "loss": 3.3744,
      "step": 2570
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012864365159612887,
      "loss": 3.4453,
      "step": 2571
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012861476238624873,
      "loss": 3.4061,
      "step": 2572
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012858587317636863,
      "loss": 3.3093,
      "step": 2573
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012855698396648852,
      "loss": 3.1693,
      "step": 2574
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012852809475660842,
      "loss": 3.1817,
      "step": 2575
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012849920554672828,
      "loss": 3.4069,
      "step": 2576
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012847031633684818,
      "loss": 3.5068,
      "step": 2577
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001284414271269681,
      "loss": 3.3693,
      "step": 2578
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000128412537917088,
      "loss": 3.3957,
      "step": 2579
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012838364870720786,
      "loss": 3.36,
      "step": 2580
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012835475949732775,
      "loss": 3.3984,
      "step": 2581
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012832587028744765,
      "loss": 3.1863,
      "step": 2582
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012829698107756754,
      "loss": 3.3402,
      "step": 2583
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012826809186768743,
      "loss": 3.3369,
      "step": 2584
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001282392026578073,
      "loss": 3.3675,
      "step": 2585
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001282103134479272,
      "loss": 3.3442,
      "step": 2586
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001281814242380471,
      "loss": 3.4422,
      "step": 2587
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012815253502816698,
      "loss": 3.2933,
      "step": 2588
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012812364581828688,
      "loss": 3.4378,
      "step": 2589
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012809475660840677,
      "loss": 3.4258,
      "step": 2590
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012806586739852666,
      "loss": 3.2842,
      "step": 2591
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012803697818864656,
      "loss": 3.0408,
      "step": 2592
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012800808897876645,
      "loss": 3.2241,
      "step": 2593
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012797919976888632,
      "loss": 3.4057,
      "step": 2594
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001279503105590062,
      "loss": 3.3407,
      "step": 2595
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001279214213491261,
      "loss": 3.4498,
      "step": 2596
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.000127892532139246,
      "loss": 3.4478,
      "step": 2597
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012786364292936587,
      "loss": 3.3925,
      "step": 2598
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001278347537194858,
      "loss": 3.3481,
      "step": 2599
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012780586450960568,
      "loss": 3.3213,
      "step": 2600
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012777697529972557,
      "loss": 3.3323,
      "step": 2601
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012774808608984544,
      "loss": 3.3765,
      "step": 2602
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012771919687996534,
      "loss": 3.3472,
      "step": 2603
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012769030767008523,
      "loss": 3.376,
      "step": 2604
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012766141846020512,
      "loss": 3.356,
      "step": 2605
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012763252925032502,
      "loss": 3.4621,
      "step": 2606
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012760364004044488,
      "loss": 3.3554,
      "step": 2607
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012757475083056478,
      "loss": 3.439,
      "step": 2608
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012754586162068467,
      "loss": 3.3964,
      "step": 2609
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001275169724108046,
      "loss": 3.2209,
      "step": 2610
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012748808320092446,
      "loss": 3.471,
      "step": 2611
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012745919399104435,
      "loss": 3.2721,
      "step": 2612
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012743030478116425,
      "loss": 3.3398,
      "step": 2613
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012740141557128414,
      "loss": 3.3427,
      "step": 2614
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012737252636140403,
      "loss": 3.4032,
      "step": 2615
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001273436371515239,
      "loss": 3.2084,
      "step": 2616
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001273147479416438,
      "loss": 3.348,
      "step": 2617
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001272858587317637,
      "loss": 3.2441,
      "step": 2618
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012725696952188358,
      "loss": 3.391,
      "step": 2619
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012722808031200348,
      "loss": 3.4738,
      "step": 2620
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012719919110212337,
      "loss": 3.4117,
      "step": 2621
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012717030189224326,
      "loss": 3.2362,
      "step": 2622
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012714141268236316,
      "loss": 3.3843,
      "step": 2623
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012711252347248302,
      "loss": 3.2797,
      "step": 2624
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012708363426260292,
      "loss": 3.3762,
      "step": 2625
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001270547450527228,
      "loss": 3.3659,
      "step": 2626
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001270258558428427,
      "loss": 3.2781,
      "step": 2627
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0001269969666329626,
      "loss": 3.3275,
      "step": 2628
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012696807742308247,
      "loss": 3.3819,
      "step": 2629
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012693918821320236,
      "loss": 3.4597,
      "step": 2630
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012691029900332228,
      "loss": 3.4048,
      "step": 2631
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012688140979344217,
      "loss": 3.3594,
      "step": 2632
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.00012685252058356204,
      "loss": 3.4678,
      "step": 2633
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012682363137368194,
      "loss": 3.1888,
      "step": 2634
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012679474216380183,
      "loss": 3.4054,
      "step": 2635
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012676585295392172,
      "loss": 3.2659,
      "step": 2636
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012673696374404162,
      "loss": 3.1928,
      "step": 2637
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012670807453416148,
      "loss": 3.3957,
      "step": 2638
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012667918532428138,
      "loss": 3.4169,
      "step": 2639
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012665029611440127,
      "loss": 3.4016,
      "step": 2640
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012662140690452116,
      "loss": 3.3982,
      "step": 2641
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012659251769464106,
      "loss": 3.3423,
      "step": 2642
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012656362848476095,
      "loss": 3.439,
      "step": 2643
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012653473927488085,
      "loss": 3.3031,
      "step": 2644
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012650585006500074,
      "loss": 3.4652,
      "step": 2645
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001264769608551206,
      "loss": 3.3652,
      "step": 2646
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001264480716452405,
      "loss": 3.3633,
      "step": 2647
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001264191824353604,
      "loss": 3.3795,
      "step": 2648
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001263902932254803,
      "loss": 3.5622,
      "step": 2649
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012636140401560018,
      "loss": 3.3977,
      "step": 2650
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012633251480572005,
      "loss": 3.4627,
      "step": 2651
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012630362559583997,
      "loss": 3.4903,
      "step": 2652
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012627473638595986,
      "loss": 3.4199,
      "step": 2653
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012624584717607976,
      "loss": 3.3015,
      "step": 2654
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012621695796619962,
      "loss": 3.2783,
      "step": 2655
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012618806875631952,
      "loss": 3.3574,
      "step": 2656
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001261591795464394,
      "loss": 3.3708,
      "step": 2657
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001261302903365593,
      "loss": 3.5374,
      "step": 2658
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001261014011266792,
      "loss": 3.4267,
      "step": 2659
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012607251191679907,
      "loss": 3.4563,
      "step": 2660
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012604362270691896,
      "loss": 3.3468,
      "step": 2661
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012601473349703888,
      "loss": 3.2677,
      "step": 2662
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012598584428715875,
      "loss": 3.3923,
      "step": 2663
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012595695507727864,
      "loss": 3.3761,
      "step": 2664
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012592806586739854,
      "loss": 3.4044,
      "step": 2665
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012589917665751843,
      "loss": 3.2343,
      "step": 2666
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012587028744763832,
      "loss": 3.4267,
      "step": 2667
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001258413982377582,
      "loss": 3.4803,
      "step": 2668
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012581250902787808,
      "loss": 3.4012,
      "step": 2669
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012578361981799798,
      "loss": 3.3552,
      "step": 2670
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012575473060811787,
      "loss": 3.204,
      "step": 2671
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012572584139823777,
      "loss": 3.3883,
      "step": 2672
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012569695218835766,
      "loss": 3.4008,
      "step": 2673
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012566806297847755,
      "loss": 3.3178,
      "step": 2674
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012563917376859745,
      "loss": 3.0956,
      "step": 2675
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012561028455871734,
      "loss": 3.325,
      "step": 2676
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001255813953488372,
      "loss": 3.417,
      "step": 2677
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001255525061389571,
      "loss": 3.4145,
      "step": 2678
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000125523616929077,
      "loss": 3.5485,
      "step": 2679
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001254947277191969,
      "loss": 3.293,
      "step": 2680
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012546583850931676,
      "loss": 3.4376,
      "step": 2681
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012543694929943665,
      "loss": 3.427,
      "step": 2682
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012540806008955657,
      "loss": 3.4608,
      "step": 2683
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012537917087967646,
      "loss": 3.4159,
      "step": 2684
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012535028166979633,
      "loss": 3.3954,
      "step": 2685
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012532139245991622,
      "loss": 3.4478,
      "step": 2686
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012529250325003612,
      "loss": 3.5245,
      "step": 2687
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.000125263614040156,
      "loss": 3.43,
      "step": 2688
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001252347248302759,
      "loss": 3.5288,
      "step": 2689
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012520583562039577,
      "loss": 3.4635,
      "step": 2690
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012517694641051567,
      "loss": 3.2762,
      "step": 2691
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012514805720063556,
      "loss": 3.207,
      "step": 2692
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012511916799075545,
      "loss": 3.318,
      "step": 2693
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012509027878087535,
      "loss": 3.4903,
      "step": 2694
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012506138957099524,
      "loss": 3.4562,
      "step": 2695
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012503250036111514,
      "loss": 3.3621,
      "step": 2696
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012500361115123503,
      "loss": 3.3899,
      "step": 2697
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012497472194135492,
      "loss": 3.3823,
      "step": 2698
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0001249458327314748,
      "loss": 3.4669,
      "step": 2699
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012491694352159468,
      "loss": 3.4761,
      "step": 2700
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012488805431171458,
      "loss": 3.3686,
      "step": 2701
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012485916510183447,
      "loss": 3.2844,
      "step": 2702
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012483027589195434,
      "loss": 3.4819,
      "step": 2703
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.00012480138668207426,
      "loss": 3.3464,
      "step": 2704
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012477249747219415,
      "loss": 3.3565,
      "step": 2705
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012474360826231405,
      "loss": 3.6032,
      "step": 2706
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001247147190524339,
      "loss": 3.3503,
      "step": 2707
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001246858298425538,
      "loss": 3.462,
      "step": 2708
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001246569406326737,
      "loss": 3.4015,
      "step": 2709
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001246280514227936,
      "loss": 3.3831,
      "step": 2710
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001245991622129135,
      "loss": 3.4502,
      "step": 2711
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012457027300303336,
      "loss": 3.5121,
      "step": 2712
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012454138379315325,
      "loss": 3.3088,
      "step": 2713
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012451249458327314,
      "loss": 3.405,
      "step": 2714
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012448360537339306,
      "loss": 3.4379,
      "step": 2715
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012445471616351293,
      "loss": 3.4802,
      "step": 2716
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012442582695363282,
      "loss": 3.3756,
      "step": 2717
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012439693774375272,
      "loss": 3.3297,
      "step": 2718
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001243680485338726,
      "loss": 3.4115,
      "step": 2719
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001243391593239925,
      "loss": 3.256,
      "step": 2720
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012431027011411237,
      "loss": 3.3898,
      "step": 2721
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012428138090423227,
      "loss": 3.2545,
      "step": 2722
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012425249169435216,
      "loss": 3.2719,
      "step": 2723
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012422360248447205,
      "loss": 3.3735,
      "step": 2724
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012419471327459195,
      "loss": 3.3794,
      "step": 2725
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012416582406471184,
      "loss": 3.4381,
      "step": 2726
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012413693485483174,
      "loss": 3.3723,
      "step": 2727
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012410804564495163,
      "loss": 3.153,
      "step": 2728
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001240791564350715,
      "loss": 3.3809,
      "step": 2729
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001240502672251914,
      "loss": 3.3281,
      "step": 2730
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012402137801531128,
      "loss": 3.1046,
      "step": 2731
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012399248880543118,
      "loss": 3.3417,
      "step": 2732
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012396359959555107,
      "loss": 3.3571,
      "step": 2733
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012393471038567094,
      "loss": 3.493,
      "step": 2734
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012390582117579083,
      "loss": 3.3214,
      "step": 2735
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012387693196591075,
      "loss": 3.4362,
      "step": 2736
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012384804275603065,
      "loss": 3.4042,
      "step": 2737
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012381915354615051,
      "loss": 3.4319,
      "step": 2738
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001237902643362704,
      "loss": 3.4458,
      "step": 2739
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001237613751263903,
      "loss": 3.4421,
      "step": 2740
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001237324859165102,
      "loss": 3.3754,
      "step": 2741
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001237035967066301,
      "loss": 3.347,
      "step": 2742
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012367470749674996,
      "loss": 3.4606,
      "step": 2743
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012364581828686985,
      "loss": 3.3032,
      "step": 2744
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012361692907698974,
      "loss": 3.3964,
      "step": 2745
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012358803986710964,
      "loss": 3.355,
      "step": 2746
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012355915065722953,
      "loss": 3.3482,
      "step": 2747
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012353026144734942,
      "loss": 3.4613,
      "step": 2748
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012350137223746932,
      "loss": 3.3546,
      "step": 2749
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001234724830275892,
      "loss": 3.2237,
      "step": 2750
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012344359381770908,
      "loss": 3.3636,
      "step": 2751
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012341470460782897,
      "loss": 3.2352,
      "step": 2752
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012338581539794887,
      "loss": 3.4965,
      "step": 2753
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012335692618806876,
      "loss": 3.2916,
      "step": 2754
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012332803697818865,
      "loss": 3.3528,
      "step": 2755
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012329914776830852,
      "loss": 3.2192,
      "step": 2756
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012327025855842844,
      "loss": 3.3678,
      "step": 2757
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012324136934854834,
      "loss": 3.2173,
      "step": 2758
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012321248013866823,
      "loss": 3.2887,
      "step": 2759
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001231835909287881,
      "loss": 3.3319,
      "step": 2760
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000123154701718908,
      "loss": 3.4748,
      "step": 2761
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012312581250902788,
      "loss": 3.2698,
      "step": 2762
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012309692329914778,
      "loss": 3.4107,
      "step": 2763
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012306803408926767,
      "loss": 3.3461,
      "step": 2764
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012303914487938754,
      "loss": 3.2252,
      "step": 2765
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012301025566950743,
      "loss": 3.225,
      "step": 2766
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012298136645962735,
      "loss": 3.4453,
      "step": 2767
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012295247724974722,
      "loss": 3.4721,
      "step": 2768
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012292358803986711,
      "loss": 3.3088,
      "step": 2769
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.000122894698829987,
      "loss": 3.4014,
      "step": 2770
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001228658096201069,
      "loss": 3.4508,
      "step": 2771
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0001228369204102268,
      "loss": 3.4795,
      "step": 2772
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012280803120034666,
      "loss": 3.3166,
      "step": 2773
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.00012277914199046656,
      "loss": 3.3391,
      "step": 2774
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012275025278058645,
      "loss": 3.2663,
      "step": 2775
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012272136357070634,
      "loss": 3.3781,
      "step": 2776
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012269247436082624,
      "loss": 3.3014,
      "step": 2777
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012266358515094613,
      "loss": 3.2865,
      "step": 2778
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012263469594106602,
      "loss": 3.2788,
      "step": 2779
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012260580673118592,
      "loss": 3.2579,
      "step": 2780
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001225769175213058,
      "loss": 3.3732,
      "step": 2781
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012254802831142568,
      "loss": 3.3919,
      "step": 2782
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012251913910154557,
      "loss": 3.3965,
      "step": 2783
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012249024989166547,
      "loss": 3.3287,
      "step": 2784
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012246136068178536,
      "loss": 3.4794,
      "step": 2785
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012243247147190525,
      "loss": 3.3119,
      "step": 2786
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012240358226202512,
      "loss": 3.2659,
      "step": 2787
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012237469305214504,
      "loss": 3.4235,
      "step": 2788
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012234580384226494,
      "loss": 3.4274,
      "step": 2789
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001223169146323848,
      "loss": 3.1281,
      "step": 2790
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001222880254225047,
      "loss": 3.4995,
      "step": 2791
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001222591362126246,
      "loss": 3.5272,
      "step": 2792
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012223024700274448,
      "loss": 3.2849,
      "step": 2793
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012220135779286438,
      "loss": 3.4098,
      "step": 2794
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012217246858298424,
      "loss": 3.42,
      "step": 2795
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012214357937310414,
      "loss": 3.2024,
      "step": 2796
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012211469016322403,
      "loss": 3.4517,
      "step": 2797
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012208580095334395,
      "loss": 3.2603,
      "step": 2798
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012205691174346383,
      "loss": 3.0616,
      "step": 2799
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012202802253358371,
      "loss": 3.447,
      "step": 2800
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012199913332370361,
      "loss": 3.2967,
      "step": 2801
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001219702441138235,
      "loss": 3.2181,
      "step": 2802
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012194135490394338,
      "loss": 3.2925,
      "step": 2803
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012191246569406328,
      "loss": 3.4124,
      "step": 2804
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012188357648418316,
      "loss": 3.2474,
      "step": 2805
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012185468727430305,
      "loss": 3.3938,
      "step": 2806
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012182579806442293,
      "loss": 3.3929,
      "step": 2807
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012179690885454282,
      "loss": 3.463,
      "step": 2808
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012176801964466273,
      "loss": 3.2457,
      "step": 2809
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012173913043478263,
      "loss": 3.3338,
      "step": 2810
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001217102412249025,
      "loss": 3.4464,
      "step": 2811
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001216813520150224,
      "loss": 3.3941,
      "step": 2812
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012165246280514229,
      "loss": 3.4187,
      "step": 2813
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012162357359526217,
      "loss": 3.4447,
      "step": 2814
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012159468438538207,
      "loss": 3.3622,
      "step": 2815
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012156579517550195,
      "loss": 3.2339,
      "step": 2816
    },
    {
      "epoch": 0.4,
      "eval_loss": 3.5397603511810303,
      "eval_runtime": 473.3374,
      "eval_samples_per_second": 43.282,
      "eval_steps_per_second": 14.427,
      "step": 2816
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012153690596562184,
      "loss": 3.2623,
      "step": 2817
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012150801675574172,
      "loss": 3.2769,
      "step": 2818
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012147912754586164,
      "loss": 3.2308,
      "step": 2819
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012145023833598152,
      "loss": 3.3513,
      "step": 2820
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012142134912610142,
      "loss": 3.3549,
      "step": 2821
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001213924599162213,
      "loss": 3.2788,
      "step": 2822
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012136357070634119,
      "loss": 3.5271,
      "step": 2823
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012133468149646108,
      "loss": 3.1716,
      "step": 2824
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012130579228658096,
      "loss": 3.3601,
      "step": 2825
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012127690307670086,
      "loss": 3.4085,
      "step": 2826
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012124801386682074,
      "loss": 3.2235,
      "step": 2827
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012121912465694063,
      "loss": 3.4152,
      "step": 2828
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012119023544706051,
      "loss": 3.291,
      "step": 2829
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012116134623718043,
      "loss": 3.322,
      "step": 2830
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012113245702730031,
      "loss": 3.3296,
      "step": 2831
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012110356781742021,
      "loss": 3.2322,
      "step": 2832
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012107467860754009,
      "loss": 3.4277,
      "step": 2833
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012104578939765998,
      "loss": 3.1975,
      "step": 2834
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012101690018777988,
      "loss": 3.3079,
      "step": 2835
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012098801097789976,
      "loss": 3.4171,
      "step": 2836
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012095912176801965,
      "loss": 3.206,
      "step": 2837
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012093023255813953,
      "loss": 3.3719,
      "step": 2838
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012090134334825942,
      "loss": 3.536,
      "step": 2839
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012087245413837933,
      "loss": 3.3528,
      "step": 2840
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012084356492849923,
      "loss": 3.3847,
      "step": 2841
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0001208146757186191,
      "loss": 2.9773,
      "step": 2842
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.000120785786508739,
      "loss": 3.3853,
      "step": 2843
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.00012075689729885888,
      "loss": 3.3528,
      "step": 2844
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012072800808897877,
      "loss": 3.3483,
      "step": 2845
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012069911887909865,
      "loss": 3.4658,
      "step": 2846
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012067022966921855,
      "loss": 3.346,
      "step": 2847
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012064134045933844,
      "loss": 3.2132,
      "step": 2848
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012061245124945832,
      "loss": 3.4038,
      "step": 2849
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012058356203957822,
      "loss": 3.3602,
      "step": 2850
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012055467282969812,
      "loss": 3.4381,
      "step": 2851
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012052578361981802,
      "loss": 3.4591,
      "step": 2852
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001204968944099379,
      "loss": 3.3465,
      "step": 2853
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012046800520005779,
      "loss": 3.3748,
      "step": 2854
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012043911599017767,
      "loss": 3.3602,
      "step": 2855
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012041022678029756,
      "loss": 3.3046,
      "step": 2856
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012038133757041745,
      "loss": 3.4129,
      "step": 2857
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012035244836053734,
      "loss": 3.3221,
      "step": 2858
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012032355915065723,
      "loss": 3.351,
      "step": 2859
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012029466994077711,
      "loss": 3.4366,
      "step": 2860
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012026578073089702,
      "loss": 3.3561,
      "step": 2861
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012023689152101691,
      "loss": 3.4635,
      "step": 2862
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012020800231113681,
      "loss": 3.3082,
      "step": 2863
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012017911310125669,
      "loss": 3.4234,
      "step": 2864
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012015022389137658,
      "loss": 3.3442,
      "step": 2865
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012012133468149646,
      "loss": 3.3602,
      "step": 2866
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012009244547161636,
      "loss": 3.0126,
      "step": 2867
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012006355626173624,
      "loss": 3.2896,
      "step": 2868
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012003466705185613,
      "loss": 3.4567,
      "step": 2869
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00012000577784197602,
      "loss": 3.2361,
      "step": 2870
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001199768886320959,
      "loss": 3.4387,
      "step": 2871
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011994799942221581,
      "loss": 3.395,
      "step": 2872
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001199191102123357,
      "loss": 3.3643,
      "step": 2873
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001198902210024556,
      "loss": 3.4342,
      "step": 2874
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011986133179257548,
      "loss": 3.4494,
      "step": 2875
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011983244258269537,
      "loss": 3.3813,
      "step": 2876
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011980355337281525,
      "loss": 3.1537,
      "step": 2877
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011977466416293515,
      "loss": 3.4073,
      "step": 2878
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011974577495305503,
      "loss": 3.402,
      "step": 2879
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011971688574317492,
      "loss": 3.2713,
      "step": 2880
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011968799653329482,
      "loss": 3.454,
      "step": 2881
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011965910732341472,
      "loss": 3.2313,
      "step": 2882
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001196302181135346,
      "loss": 3.3605,
      "step": 2883
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001196013289036545,
      "loss": 3.3283,
      "step": 2884
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011957243969377439,
      "loss": 3.4797,
      "step": 2885
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011954355048389427,
      "loss": 3.4396,
      "step": 2886
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011951466127401417,
      "loss": 3.4563,
      "step": 2887
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011948577206413405,
      "loss": 3.4401,
      "step": 2888
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011945688285425394,
      "loss": 3.4105,
      "step": 2889
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011942799364437382,
      "loss": 3.2352,
      "step": 2890
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011939910443449371,
      "loss": 3.5398,
      "step": 2891
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011937021522461361,
      "loss": 3.2313,
      "step": 2892
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011934132601473351,
      "loss": 3.0719,
      "step": 2893
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001193124368048534,
      "loss": 3.3121,
      "step": 2894
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011928354759497329,
      "loss": 3.2686,
      "step": 2895
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011925465838509318,
      "loss": 3.3653,
      "step": 2896
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011922576917521306,
      "loss": 3.4363,
      "step": 2897
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011919687996533296,
      "loss": 3.525,
      "step": 2898
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011916799075545284,
      "loss": 3.3923,
      "step": 2899
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011913910154557273,
      "loss": 3.4238,
      "step": 2900
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011911021233569261,
      "loss": 3.2937,
      "step": 2901
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001190813231258125,
      "loss": 3.37,
      "step": 2902
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011905243391593241,
      "loss": 3.3827,
      "step": 2903
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001190235447060523,
      "loss": 3.4186,
      "step": 2904
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011899465549617219,
      "loss": 3.2927,
      "step": 2905
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011896576628629208,
      "loss": 3.2217,
      "step": 2906
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011893687707641197,
      "loss": 3.3386,
      "step": 2907
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011890798786653185,
      "loss": 3.3211,
      "step": 2908
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011887909865665175,
      "loss": 3.4936,
      "step": 2909
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011885020944677163,
      "loss": 3.3862,
      "step": 2910
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.00011882132023689152,
      "loss": 3.3776,
      "step": 2911
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001187924310270114,
      "loss": 3.3151,
      "step": 2912
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001187635418171313,
      "loss": 3.1704,
      "step": 2913
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0001187346526072512,
      "loss": 3.3042,
      "step": 2914
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001187057633973711,
      "loss": 3.0431,
      "step": 2915
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011867687418749098,
      "loss": 3.2402,
      "step": 2916
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011864798497761087,
      "loss": 3.3413,
      "step": 2917
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011861909576773077,
      "loss": 3.4283,
      "step": 2918
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011859020655785065,
      "loss": 3.2246,
      "step": 2919
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011856131734797054,
      "loss": 3.5293,
      "step": 2920
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011853242813809042,
      "loss": 3.309,
      "step": 2921
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011850353892821031,
      "loss": 3.4406,
      "step": 2922
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001184746497183302,
      "loss": 3.3076,
      "step": 2923
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011844576050845011,
      "loss": 3.3118,
      "step": 2924
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011841687129857,
      "loss": 3.3957,
      "step": 2925
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011838798208868989,
      "loss": 3.3812,
      "step": 2926
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011835909287880977,
      "loss": 3.3103,
      "step": 2927
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011833020366892966,
      "loss": 3.3207,
      "step": 2928
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011830131445904956,
      "loss": 3.3636,
      "step": 2929
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011827242524916944,
      "loss": 3.2148,
      "step": 2930
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011824353603928933,
      "loss": 3.4342,
      "step": 2931
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011821464682940921,
      "loss": 3.418,
      "step": 2932
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001181857576195291,
      "loss": 3.1815,
      "step": 2933
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011815686840964899,
      "loss": 3.3941,
      "step": 2934
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001181279791997689,
      "loss": 3.2534,
      "step": 2935
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011809908998988879,
      "loss": 3.2224,
      "step": 2936
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011807020078000868,
      "loss": 3.2472,
      "step": 2937
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011804131157012856,
      "loss": 3.3013,
      "step": 2938
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011801242236024845,
      "loss": 3.4542,
      "step": 2939
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011798353315036835,
      "loss": 3.3669,
      "step": 2940
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011795464394048823,
      "loss": 3.2054,
      "step": 2941
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011792575473060812,
      "loss": 3.4628,
      "step": 2942
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.000117896865520728,
      "loss": 3.314,
      "step": 2943
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001178679763108479,
      "loss": 3.2845,
      "step": 2944
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001178390871009678,
      "loss": 3.3936,
      "step": 2945
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001178101978910877,
      "loss": 3.3622,
      "step": 2946
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011778130868120758,
      "loss": 3.4131,
      "step": 2947
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011775241947132747,
      "loss": 3.2771,
      "step": 2948
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011772353026144735,
      "loss": 3.4104,
      "step": 2949
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011769464105156725,
      "loss": 3.4032,
      "step": 2950
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011766575184168714,
      "loss": 3.2302,
      "step": 2951
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011763686263180702,
      "loss": 3.3607,
      "step": 2952
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011760797342192691,
      "loss": 3.2546,
      "step": 2953
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001175790842120468,
      "loss": 3.2651,
      "step": 2954
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011755019500216669,
      "loss": 3.2851,
      "step": 2955
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001175213057922866,
      "loss": 3.4795,
      "step": 2956
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011749241658240649,
      "loss": 3.4549,
      "step": 2957
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011746352737252637,
      "loss": 3.3106,
      "step": 2958
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011743463816264626,
      "loss": 3.363,
      "step": 2959
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011740574895276614,
      "loss": 3.2841,
      "step": 2960
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011737685974288604,
      "loss": 3.4212,
      "step": 2961
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011734797053300593,
      "loss": 3.3983,
      "step": 2962
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011731908132312581,
      "loss": 3.2952,
      "step": 2963
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001172901921132457,
      "loss": 3.4007,
      "step": 2964
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011726130290336559,
      "loss": 3.3081,
      "step": 2965
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011723241369348549,
      "loss": 3.4531,
      "step": 2966
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011720352448360539,
      "loss": 3.3969,
      "step": 2967
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011717463527372528,
      "loss": 3.3213,
      "step": 2968
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011714574606384516,
      "loss": 3.3514,
      "step": 2969
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011711685685396505,
      "loss": 3.3203,
      "step": 2970
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011708796764408493,
      "loss": 3.3769,
      "step": 2971
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011705907843420483,
      "loss": 3.3929,
      "step": 2972
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011703018922432471,
      "loss": 3.4084,
      "step": 2973
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001170013000144446,
      "loss": 3.2764,
      "step": 2974
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001169724108045645,
      "loss": 3.3365,
      "step": 2975
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001169435215946844,
      "loss": 3.137,
      "step": 2976
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011691463238480428,
      "loss": 3.3314,
      "step": 2977
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011688574317492418,
      "loss": 3.2965,
      "step": 2978
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011685685396504407,
      "loss": 3.4489,
      "step": 2979
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011682796475516395,
      "loss": 3.2876,
      "step": 2980
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011679907554528385,
      "loss": 3.3736,
      "step": 2981
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011677018633540373,
      "loss": 3.4143,
      "step": 2982
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.00011674129712552362,
      "loss": 3.3062,
      "step": 2983
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0001167124079156435,
      "loss": 3.3008,
      "step": 2984
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001166835187057634,
      "loss": 3.5034,
      "step": 2985
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011665462949588329,
      "loss": 3.2881,
      "step": 2986
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001166257402860032,
      "loss": 3.3307,
      "step": 2987
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011659685107612308,
      "loss": 3.3547,
      "step": 2988
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011656796186624297,
      "loss": 3.1519,
      "step": 2989
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011653907265636286,
      "loss": 3.4225,
      "step": 2990
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011651018344648274,
      "loss": 3.3672,
      "step": 2991
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011648129423660264,
      "loss": 3.4003,
      "step": 2992
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011645240502672252,
      "loss": 3.4046,
      "step": 2993
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011642351581684241,
      "loss": 3.3142,
      "step": 2994
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011639462660696229,
      "loss": 3.3948,
      "step": 2995
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011636573739708219,
      "loss": 3.3539,
      "step": 2996
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011633684818720209,
      "loss": 3.4269,
      "step": 2997
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011630795897732199,
      "loss": 3.4374,
      "step": 2998
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011627906976744187,
      "loss": 3.212,
      "step": 2999
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011625018055756176,
      "loss": 3.3384,
      "step": 3000
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011622129134768165,
      "loss": 3.5069,
      "step": 3001
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011619240213780153,
      "loss": 3.2896,
      "step": 3002
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011616351292792143,
      "loss": 3.2173,
      "step": 3003
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011613462371804131,
      "loss": 3.2881,
      "step": 3004
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001161057345081612,
      "loss": 3.4238,
      "step": 3005
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011607684529828108,
      "loss": 3.3646,
      "step": 3006
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011604795608840098,
      "loss": 3.2412,
      "step": 3007
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011601906687852088,
      "loss": 3.3506,
      "step": 3008
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011599017766864078,
      "loss": 3.3297,
      "step": 3009
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011596128845876066,
      "loss": 3.3462,
      "step": 3010
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011593239924888055,
      "loss": 3.3547,
      "step": 3011
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011590351003900045,
      "loss": 3.4532,
      "step": 3012
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011587462082912033,
      "loss": 3.4124,
      "step": 3013
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011584573161924022,
      "loss": 3.3584,
      "step": 3014
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001158168424093601,
      "loss": 3.3526,
      "step": 3015
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011578795319948,
      "loss": 3.338,
      "step": 3016
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011575906398959987,
      "loss": 3.4493,
      "step": 3017
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001157301747797198,
      "loss": 3.3514,
      "step": 3018
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011570128556983968,
      "loss": 3.423,
      "step": 3019
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011567239635995957,
      "loss": 3.3087,
      "step": 3020
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011564350715007945,
      "loss": 3.3231,
      "step": 3021
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011561461794019934,
      "loss": 3.2946,
      "step": 3022
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011558572873031924,
      "loss": 3.3502,
      "step": 3023
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011555683952043912,
      "loss": 3.3376,
      "step": 3024
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011552795031055901,
      "loss": 3.4109,
      "step": 3025
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011549906110067889,
      "loss": 3.3582,
      "step": 3026
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011547017189079879,
      "loss": 3.2769,
      "step": 3027
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011544128268091867,
      "loss": 3.4471,
      "step": 3028
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011541239347103859,
      "loss": 3.3088,
      "step": 3029
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011538350426115847,
      "loss": 3.434,
      "step": 3030
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011535461505127836,
      "loss": 3.4577,
      "step": 3031
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011532572584139824,
      "loss": 3.4747,
      "step": 3032
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011529683663151814,
      "loss": 3.339,
      "step": 3033
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011526794742163803,
      "loss": 3.2112,
      "step": 3034
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011523905821175791,
      "loss": 3.2662,
      "step": 3035
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001152101690018778,
      "loss": 3.3228,
      "step": 3036
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011518127979199768,
      "loss": 3.4458,
      "step": 3037
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011515239058211758,
      "loss": 3.4888,
      "step": 3038
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011512350137223748,
      "loss": 3.4086,
      "step": 3039
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011509461216235738,
      "loss": 3.3899,
      "step": 3040
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011506572295247726,
      "loss": 3.2718,
      "step": 3041
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011503683374259715,
      "loss": 3.3835,
      "step": 3042
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011500794453271703,
      "loss": 3.3354,
      "step": 3043
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011497905532283693,
      "loss": 3.3969,
      "step": 3044
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011495016611295682,
      "loss": 3.2528,
      "step": 3045
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001149212769030767,
      "loss": 3.2295,
      "step": 3046
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0001148923876931966,
      "loss": 3.4159,
      "step": 3047
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011486349848331647,
      "loss": 3.4099,
      "step": 3048
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011483460927343637,
      "loss": 3.3859,
      "step": 3049
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011480572006355628,
      "loss": 3.381,
      "step": 3050
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011477683085367617,
      "loss": 3.1571,
      "step": 3051
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011474794164379605,
      "loss": 3.3107,
      "step": 3052
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011471905243391594,
      "loss": 3.3559,
      "step": 3053
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011469016322403582,
      "loss": 3.3384,
      "step": 3054
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.00011466127401415572,
      "loss": 3.2054,
      "step": 3055
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011463238480427561,
      "loss": 3.3614,
      "step": 3056
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011460349559439549,
      "loss": 3.418,
      "step": 3057
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011457460638451539,
      "loss": 3.2436,
      "step": 3058
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011454571717463527,
      "loss": 3.3524,
      "step": 3059
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011451682796475517,
      "loss": 3.5018,
      "step": 3060
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011448793875487507,
      "loss": 3.4198,
      "step": 3061
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011445904954499496,
      "loss": 3.183,
      "step": 3062
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011443016033511484,
      "loss": 3.4058,
      "step": 3063
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011440127112523474,
      "loss": 3.3522,
      "step": 3064
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011437238191535462,
      "loss": 3.2152,
      "step": 3065
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011434349270547451,
      "loss": 3.2856,
      "step": 3066
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001143146034955944,
      "loss": 3.3922,
      "step": 3067
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011428571428571428,
      "loss": 3.3337,
      "step": 3068
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011425682507583418,
      "loss": 3.3856,
      "step": 3069
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011422793586595406,
      "loss": 3.4281,
      "step": 3070
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011419904665607396,
      "loss": 3.2898,
      "step": 3071
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011417015744619386,
      "loss": 3.2965,
      "step": 3072
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011414126823631375,
      "loss": 3.3825,
      "step": 3073
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011411237902643363,
      "loss": 3.4231,
      "step": 3074
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011408348981655353,
      "loss": 3.4035,
      "step": 3075
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011405460060667341,
      "loss": 3.4066,
      "step": 3076
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001140257113967933,
      "loss": 3.2967,
      "step": 3077
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001139968221869132,
      "loss": 3.3218,
      "step": 3078
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011396793297703307,
      "loss": 3.461,
      "step": 3079
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011393904376715297,
      "loss": 3.4681,
      "step": 3080
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011391015455727288,
      "loss": 3.2937,
      "step": 3081
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011388126534739276,
      "loss": 3.3003,
      "step": 3082
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011385237613751265,
      "loss": 3.4808,
      "step": 3083
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011382348692763254,
      "loss": 3.3351,
      "step": 3084
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011379459771775242,
      "loss": 3.3666,
      "step": 3085
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011376570850787232,
      "loss": 3.3265,
      "step": 3086
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001137368192979922,
      "loss": 3.1792,
      "step": 3087
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011370793008811209,
      "loss": 3.3736,
      "step": 3088
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011367904087823199,
      "loss": 3.2193,
      "step": 3089
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011365015166835187,
      "loss": 3.3988,
      "step": 3090
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011362126245847176,
      "loss": 3.3794,
      "step": 3091
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011359237324859167,
      "loss": 3.2979,
      "step": 3092
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011356348403871155,
      "loss": 3.3949,
      "step": 3093
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011353459482883144,
      "loss": 3.3369,
      "step": 3094
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011350570561895134,
      "loss": 3.2751,
      "step": 3095
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011347681640907122,
      "loss": 3.4321,
      "step": 3096
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011344792719919111,
      "loss": 3.3877,
      "step": 3097
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011341903798931099,
      "loss": 3.3482,
      "step": 3098
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011339014877943088,
      "loss": 3.2839,
      "step": 3099
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011336125956955076,
      "loss": 3.3243,
      "step": 3100
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011333237035967066,
      "loss": 3.4563,
      "step": 3101
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011330348114979057,
      "loss": 3.4088,
      "step": 3102
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011327459193991046,
      "loss": 3.281,
      "step": 3103
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011324570273003034,
      "loss": 3.2996,
      "step": 3104
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011321681352015023,
      "loss": 3.1804,
      "step": 3105
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011318792431027013,
      "loss": 3.3312,
      "step": 3106
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011315903510039001,
      "loss": 3.4377,
      "step": 3107
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001131301458905099,
      "loss": 3.4095,
      "step": 3108
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011310125668062978,
      "loss": 3.3624,
      "step": 3109
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011307236747074968,
      "loss": 3.3228,
      "step": 3110
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011304347826086956,
      "loss": 3.43,
      "step": 3111
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011301458905098945,
      "loss": 3.3281,
      "step": 3112
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011298569984110936,
      "loss": 3.243,
      "step": 3113
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011295681063122925,
      "loss": 3.4537,
      "step": 3114
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011292792142134913,
      "loss": 3.393,
      "step": 3115
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011289903221146902,
      "loss": 3.3423,
      "step": 3116
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011287014300158892,
      "loss": 3.4138,
      "step": 3117
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0001128412537917088,
      "loss": 3.2931,
      "step": 3118
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011281236458182869,
      "loss": 3.4065,
      "step": 3119
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011278347537194857,
      "loss": 3.2166,
      "step": 3120
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011275458616206847,
      "loss": 3.3086,
      "step": 3121
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011272569695218835,
      "loss": 3.3501,
      "step": 3122
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011269680774230827,
      "loss": 3.5366,
      "step": 3123
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011266791853242815,
      "loss": 3.4074,
      "step": 3124
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.00011263902932254804,
      "loss": 3.1839,
      "step": 3125
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011261014011266792,
      "loss": 3.3956,
      "step": 3126
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011258125090278782,
      "loss": 3.3401,
      "step": 3127
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011255236169290771,
      "loss": 3.3493,
      "step": 3128
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011252347248302759,
      "loss": 3.38,
      "step": 3129
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011249458327314748,
      "loss": 3.2987,
      "step": 3130
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011246569406326736,
      "loss": 3.2619,
      "step": 3131
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011243680485338726,
      "loss": 3.3791,
      "step": 3132
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011240791564350714,
      "loss": 3.401,
      "step": 3133
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011237902643362706,
      "loss": 3.1383,
      "step": 3134
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011235013722374694,
      "loss": 3.3987,
      "step": 3135
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011232124801386683,
      "loss": 3.248,
      "step": 3136
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011229235880398671,
      "loss": 3.2962,
      "step": 3137
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011226346959410661,
      "loss": 3.2754,
      "step": 3138
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001122345803842265,
      "loss": 3.4079,
      "step": 3139
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011220569117434638,
      "loss": 3.4831,
      "step": 3140
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011217680196446628,
      "loss": 3.4135,
      "step": 3141
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011214791275458616,
      "loss": 3.3875,
      "step": 3142
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011211902354470605,
      "loss": 3.2953,
      "step": 3143
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011209013433482596,
      "loss": 3.4627,
      "step": 3144
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011206124512494585,
      "loss": 3.2988,
      "step": 3145
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011203235591506573,
      "loss": 3.2434,
      "step": 3146
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011200346670518562,
      "loss": 3.4888,
      "step": 3147
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001119745774953055,
      "loss": 3.3249,
      "step": 3148
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001119456882854254,
      "loss": 3.4025,
      "step": 3149
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011191679907554529,
      "loss": 3.4046,
      "step": 3150
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011188790986566517,
      "loss": 3.2726,
      "step": 3151
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011185902065578507,
      "loss": 3.3567,
      "step": 3152
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011183013144590495,
      "loss": 3.3017,
      "step": 3153
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011180124223602484,
      "loss": 3.3534,
      "step": 3154
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011177235302614475,
      "loss": 3.3808,
      "step": 3155
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011174346381626464,
      "loss": 3.2337,
      "step": 3156
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011171457460638452,
      "loss": 3.3304,
      "step": 3157
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011168568539650442,
      "loss": 3.2466,
      "step": 3158
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001116567961866243,
      "loss": 3.4152,
      "step": 3159
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011162790697674419,
      "loss": 3.2942,
      "step": 3160
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011159901776686408,
      "loss": 3.3444,
      "step": 3161
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011157012855698396,
      "loss": 3.3307,
      "step": 3162
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011154123934710386,
      "loss": 3.3356,
      "step": 3163
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011151235013722374,
      "loss": 3.4839,
      "step": 3164
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011148346092734366,
      "loss": 3.4511,
      "step": 3165
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011145457171746354,
      "loss": 3.3769,
      "step": 3166
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011142568250758343,
      "loss": 3.2655,
      "step": 3167
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011139679329770331,
      "loss": 3.3711,
      "step": 3168
    },
    {
      "epoch": 0.45,
      "eval_loss": 3.5232768058776855,
      "eval_runtime": 472.4363,
      "eval_samples_per_second": 43.365,
      "eval_steps_per_second": 14.455,
      "step": 3168
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011136790408782321,
      "loss": 3.407,
      "step": 3169
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011133901487794309,
      "loss": 3.4157,
      "step": 3170
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011131012566806298,
      "loss": 3.2203,
      "step": 3171
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011128123645818288,
      "loss": 3.3327,
      "step": 3172
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011125234724830276,
      "loss": 3.2629,
      "step": 3173
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011122345803842265,
      "loss": 3.3861,
      "step": 3174
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011119456882854256,
      "loss": 3.3472,
      "step": 3175
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011116567961866244,
      "loss": 3.4091,
      "step": 3176
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011113679040878233,
      "loss": 3.2157,
      "step": 3177
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011110790119890222,
      "loss": 3.3241,
      "step": 3178
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001110790119890221,
      "loss": 3.3768,
      "step": 3179
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.000111050122779142,
      "loss": 3.3307,
      "step": 3180
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011102123356926188,
      "loss": 3.4113,
      "step": 3181
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011099234435938177,
      "loss": 3.2805,
      "step": 3182
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011096345514950167,
      "loss": 3.4032,
      "step": 3183
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011093456593962155,
      "loss": 3.0521,
      "step": 3184
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011090567672974144,
      "loss": 3.2905,
      "step": 3185
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011087678751986135,
      "loss": 3.2763,
      "step": 3186
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011084789830998123,
      "loss": 3.2978,
      "step": 3187
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011081900910010112,
      "loss": 3.5079,
      "step": 3188
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011079011989022102,
      "loss": 3.3663,
      "step": 3189
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0001107612306803409,
      "loss": 3.3789,
      "step": 3190
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011073234147046079,
      "loss": 3.3341,
      "step": 3191
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011070345226058067,
      "loss": 3.3016,
      "step": 3192
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011067456305070056,
      "loss": 3.3406,
      "step": 3193
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011064567384082046,
      "loss": 3.239,
      "step": 3194
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.00011061678463094034,
      "loss": 3.3135,
      "step": 3195
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011058789542106025,
      "loss": 3.4249,
      "step": 3196
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011055900621118014,
      "loss": 3.3227,
      "step": 3197
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011053011700130002,
      "loss": 3.2321,
      "step": 3198
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011050122779141991,
      "loss": 3.3633,
      "step": 3199
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011047233858153981,
      "loss": 3.1829,
      "step": 3200
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011044344937165969,
      "loss": 3.4205,
      "step": 3201
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011041456016177958,
      "loss": 3.3592,
      "step": 3202
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011038567095189946,
      "loss": 3.295,
      "step": 3203
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011035678174201936,
      "loss": 3.4284,
      "step": 3204
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011032789253213925,
      "loss": 3.2576,
      "step": 3205
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011029900332225913,
      "loss": 3.2283,
      "step": 3206
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011027011411237904,
      "loss": 3.3789,
      "step": 3207
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011024122490249893,
      "loss": 3.3567,
      "step": 3208
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011021233569261881,
      "loss": 3.2874,
      "step": 3209
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001101834464827387,
      "loss": 3.2771,
      "step": 3210
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001101545572728586,
      "loss": 3.4531,
      "step": 3211
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011012566806297848,
      "loss": 3.4304,
      "step": 3212
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011009677885309837,
      "loss": 3.3653,
      "step": 3213
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011006788964321825,
      "loss": 3.198,
      "step": 3214
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011003900043333815,
      "loss": 3.4138,
      "step": 3215
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00011001011122345804,
      "loss": 3.3546,
      "step": 3216
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010998122201357795,
      "loss": 3.361,
      "step": 3217
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010995233280369783,
      "loss": 3.4561,
      "step": 3218
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010992344359381772,
      "loss": 3.3528,
      "step": 3219
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001098945543839376,
      "loss": 3.2822,
      "step": 3220
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001098656651740575,
      "loss": 3.138,
      "step": 3221
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010983677596417739,
      "loss": 3.3132,
      "step": 3222
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010980788675429727,
      "loss": 3.1742,
      "step": 3223
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010977899754441716,
      "loss": 3.4067,
      "step": 3224
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010975010833453704,
      "loss": 3.121,
      "step": 3225
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010972121912465694,
      "loss": 3.2398,
      "step": 3226
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010969232991477683,
      "loss": 3.3237,
      "step": 3227
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010966344070489674,
      "loss": 3.2556,
      "step": 3228
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010963455149501662,
      "loss": 3.3521,
      "step": 3229
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010960566228513651,
      "loss": 3.2995,
      "step": 3230
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001095767730752564,
      "loss": 3.2207,
      "step": 3231
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010954788386537629,
      "loss": 3.2197,
      "step": 3232
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010951899465549618,
      "loss": 3.2293,
      "step": 3233
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010949010544561606,
      "loss": 3.2608,
      "step": 3234
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010946121623573596,
      "loss": 3.0432,
      "step": 3235
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010943232702585584,
      "loss": 3.3144,
      "step": 3236
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010940343781597573,
      "loss": 3.304,
      "step": 3237
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010937454860609564,
      "loss": 3.3748,
      "step": 3238
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010934565939621553,
      "loss": 3.3572,
      "step": 3239
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010931677018633541,
      "loss": 3.3201,
      "step": 3240
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001092878809764553,
      "loss": 3.3465,
      "step": 3241
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010925899176657519,
      "loss": 3.4183,
      "step": 3242
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010923010255669508,
      "loss": 3.2054,
      "step": 3243
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010920121334681497,
      "loss": 3.3875,
      "step": 3244
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010917232413693485,
      "loss": 3.2239,
      "step": 3245
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010914343492705475,
      "loss": 3.4065,
      "step": 3246
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010911454571717463,
      "loss": 3.3298,
      "step": 3247
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010908565650729452,
      "loss": 3.503,
      "step": 3248
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010905676729741443,
      "loss": 3.1974,
      "step": 3249
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010902787808753432,
      "loss": 3.2414,
      "step": 3250
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001089989888776542,
      "loss": 3.2385,
      "step": 3251
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0001089700996677741,
      "loss": 3.3298,
      "step": 3252
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010894121045789398,
      "loss": 3.3801,
      "step": 3253
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010891232124801387,
      "loss": 3.1857,
      "step": 3254
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010888343203813376,
      "loss": 3.2238,
      "step": 3255
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010885454282825365,
      "loss": 3.3985,
      "step": 3256
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010882565361837354,
      "loss": 3.4367,
      "step": 3257
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010879676440849342,
      "loss": 3.287,
      "step": 3258
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010876787519861334,
      "loss": 3.4034,
      "step": 3259
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010873898598873322,
      "loss": 3.3362,
      "step": 3260
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010871009677885311,
      "loss": 3.3303,
      "step": 3261
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.000108681207568973,
      "loss": 3.3775,
      "step": 3262
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010865231835909289,
      "loss": 3.3738,
      "step": 3263
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010862342914921277,
      "loss": 3.2712,
      "step": 3264
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.00010859453993933266,
      "loss": 3.3352,
      "step": 3265
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010856565072945256,
      "loss": 3.3474,
      "step": 3266
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010853676151957244,
      "loss": 3.2761,
      "step": 3267
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010850787230969233,
      "loss": 3.2135,
      "step": 3268
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010847898309981221,
      "loss": 3.3481,
      "step": 3269
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010845009388993213,
      "loss": 3.3173,
      "step": 3270
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010842120468005201,
      "loss": 3.2898,
      "step": 3271
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001083923154701719,
      "loss": 3.2506,
      "step": 3272
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010836342626029179,
      "loss": 3.2026,
      "step": 3273
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010833453705041168,
      "loss": 3.3621,
      "step": 3274
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010830564784053156,
      "loss": 3.2253,
      "step": 3275
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010827675863065145,
      "loss": 3.3631,
      "step": 3276
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010824786942077135,
      "loss": 3.4034,
      "step": 3277
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010821898021089123,
      "loss": 3.3907,
      "step": 3278
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010819009100101112,
      "loss": 3.3238,
      "step": 3279
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010816120179113103,
      "loss": 3.4008,
      "step": 3280
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010813231258125092,
      "loss": 3.1676,
      "step": 3281
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001081034233713708,
      "loss": 3.1822,
      "step": 3282
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001080745341614907,
      "loss": 3.2965,
      "step": 3283
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010804564495161058,
      "loss": 3.406,
      "step": 3284
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010801675574173047,
      "loss": 3.2503,
      "step": 3285
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010798786653185035,
      "loss": 3.3469,
      "step": 3286
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010795897732197025,
      "loss": 3.4601,
      "step": 3287
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010793008811209014,
      "loss": 3.3476,
      "step": 3288
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010790119890221002,
      "loss": 3.1634,
      "step": 3289
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010787230969232991,
      "loss": 3.2858,
      "step": 3290
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010784342048244982,
      "loss": 3.3034,
      "step": 3291
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010781453127256971,
      "loss": 3.2368,
      "step": 3292
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001077856420626896,
      "loss": 3.2122,
      "step": 3293
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010775675285280949,
      "loss": 3.2798,
      "step": 3294
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010772786364292937,
      "loss": 3.2974,
      "step": 3295
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010769897443304926,
      "loss": 3.3698,
      "step": 3296
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010767008522316914,
      "loss": 3.3504,
      "step": 3297
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010764119601328904,
      "loss": 3.1445,
      "step": 3298
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010761230680340893,
      "loss": 3.3549,
      "step": 3299
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010758341759352881,
      "loss": 3.1997,
      "step": 3300
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010755452838364872,
      "loss": 3.2635,
      "step": 3301
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010752563917376861,
      "loss": 3.3632,
      "step": 3302
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010749674996388849,
      "loss": 3.2498,
      "step": 3303
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010746786075400839,
      "loss": 3.2627,
      "step": 3304
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010743897154412828,
      "loss": 3.3686,
      "step": 3305
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010741008233424816,
      "loss": 3.2816,
      "step": 3306
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010738119312436805,
      "loss": 3.2955,
      "step": 3307
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010735230391448793,
      "loss": 3.3908,
      "step": 3308
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010732341470460783,
      "loss": 3.282,
      "step": 3309
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010729452549472772,
      "loss": 3.3469,
      "step": 3310
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001072656362848476,
      "loss": 3.2559,
      "step": 3311
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010723674707496751,
      "loss": 3.3632,
      "step": 3312
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001072078578650874,
      "loss": 3.3206,
      "step": 3313
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010717896865520728,
      "loss": 3.3854,
      "step": 3314
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010715007944532718,
      "loss": 3.2409,
      "step": 3315
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010712119023544707,
      "loss": 3.3704,
      "step": 3316
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010709230102556695,
      "loss": 3.202,
      "step": 3317
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010706341181568685,
      "loss": 3.3145,
      "step": 3318
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010703452260580673,
      "loss": 3.2447,
      "step": 3319
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010700563339592662,
      "loss": 3.4438,
      "step": 3320
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010697674418604651,
      "loss": 3.2226,
      "step": 3321
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010694785497616642,
      "loss": 3.3073,
      "step": 3322
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001069189657662863,
      "loss": 3.31,
      "step": 3323
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001068900765564062,
      "loss": 3.1201,
      "step": 3324
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010686118734652608,
      "loss": 3.4242,
      "step": 3325
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010683229813664597,
      "loss": 3.2957,
      "step": 3326
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010680340892676586,
      "loss": 3.3738,
      "step": 3327
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010677451971688574,
      "loss": 3.1818,
      "step": 3328
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010674563050700564,
      "loss": 3.3446,
      "step": 3329
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010671674129712552,
      "loss": 3.3844,
      "step": 3330
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010668785208724541,
      "loss": 3.3296,
      "step": 3331
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0001066589628773653,
      "loss": 3.3716,
      "step": 3332
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010663007366748521,
      "loss": 3.1275,
      "step": 3333
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010660118445760509,
      "loss": 3.0919,
      "step": 3334
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010657229524772499,
      "loss": 3.3065,
      "step": 3335
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.00010654340603784487,
      "loss": 3.3737,
      "step": 3336
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010651451682796476,
      "loss": 3.3247,
      "step": 3337
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010648562761808465,
      "loss": 3.0945,
      "step": 3338
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010645673840820453,
      "loss": 3.3309,
      "step": 3339
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010642784919832443,
      "loss": 3.3925,
      "step": 3340
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010639895998844431,
      "loss": 3.3906,
      "step": 3341
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001063700707785642,
      "loss": 3.3281,
      "step": 3342
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010634118156868411,
      "loss": 3.3451,
      "step": 3343
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000106312292358804,
      "loss": 3.4958,
      "step": 3344
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010628340314892388,
      "loss": 3.3795,
      "step": 3345
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010625451393904378,
      "loss": 3.1946,
      "step": 3346
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010622562472916366,
      "loss": 3.2881,
      "step": 3347
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010619673551928355,
      "loss": 3.3135,
      "step": 3348
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010616784630940345,
      "loss": 3.3223,
      "step": 3349
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010613895709952333,
      "loss": 3.4084,
      "step": 3350
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010611006788964322,
      "loss": 3.1788,
      "step": 3351
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001060811786797631,
      "loss": 3.2679,
      "step": 3352
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.000106052289469883,
      "loss": 3.2749,
      "step": 3353
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001060234002600029,
      "loss": 3.2444,
      "step": 3354
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001059945110501228,
      "loss": 3.3425,
      "step": 3355
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010596562184024268,
      "loss": 3.2667,
      "step": 3356
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010593673263036257,
      "loss": 3.4639,
      "step": 3357
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010590784342048245,
      "loss": 3.283,
      "step": 3358
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010587895421060234,
      "loss": 3.3782,
      "step": 3359
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010585006500072224,
      "loss": 3.3153,
      "step": 3360
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010582117579084212,
      "loss": 3.3003,
      "step": 3361
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010579228658096201,
      "loss": 3.3667,
      "step": 3362
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010576339737108189,
      "loss": 3.1026,
      "step": 3363
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010573450816120181,
      "loss": 3.3164,
      "step": 3364
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010570561895132169,
      "loss": 3.3183,
      "step": 3365
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010567672974144159,
      "loss": 3.2863,
      "step": 3366
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010564784053156147,
      "loss": 3.4557,
      "step": 3367
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010561895132168136,
      "loss": 3.3675,
      "step": 3368
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010559006211180124,
      "loss": 3.2719,
      "step": 3369
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010556117290192113,
      "loss": 3.3039,
      "step": 3370
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010553228369204103,
      "loss": 3.3442,
      "step": 3371
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010550339448216091,
      "loss": 3.3411,
      "step": 3372
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001054745052722808,
      "loss": 3.2531,
      "step": 3373
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010544561606240071,
      "loss": 3.3079,
      "step": 3374
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001054167268525206,
      "loss": 3.2034,
      "step": 3375
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010538783764264048,
      "loss": 3.1568,
      "step": 3376
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010535894843276038,
      "loss": 3.3041,
      "step": 3377
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010533005922288026,
      "loss": 3.2839,
      "step": 3378
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010530117001300015,
      "loss": 3.2965,
      "step": 3379
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010527228080312003,
      "loss": 3.2,
      "step": 3380
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010524339159323993,
      "loss": 3.2672,
      "step": 3381
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010521450238335982,
      "loss": 3.3248,
      "step": 3382
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001051856131734797,
      "loss": 3.4735,
      "step": 3383
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001051567239635996,
      "loss": 3.4084,
      "step": 3384
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001051278347537195,
      "loss": 3.1859,
      "step": 3385
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001050989455438394,
      "loss": 3.2975,
      "step": 3386
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010507005633395928,
      "loss": 3.4183,
      "step": 3387
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010504116712407917,
      "loss": 3.276,
      "step": 3388
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010501227791419905,
      "loss": 3.2373,
      "step": 3389
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010498338870431894,
      "loss": 3.3277,
      "step": 3390
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010495449949443882,
      "loss": 3.2967,
      "step": 3391
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010492561028455872,
      "loss": 3.476,
      "step": 3392
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010489672107467861,
      "loss": 3.1755,
      "step": 3393
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010486783186479849,
      "loss": 3.3237,
      "step": 3394
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001048389426549184,
      "loss": 3.3224,
      "step": 3395
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010481005344503829,
      "loss": 3.212,
      "step": 3396
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010478116423515819,
      "loss": 3.2954,
      "step": 3397
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010475227502527807,
      "loss": 3.4549,
      "step": 3398
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010472338581539796,
      "loss": 3.2889,
      "step": 3399
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010469449660551784,
      "loss": 3.3868,
      "step": 3400
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010466560739563773,
      "loss": 3.3967,
      "step": 3401
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010463671818575762,
      "loss": 3.2687,
      "step": 3402
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010460782897587751,
      "loss": 3.2324,
      "step": 3403
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0001045789397659974,
      "loss": 3.3779,
      "step": 3404
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010455005055611728,
      "loss": 3.2455,
      "step": 3405
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.00010452116134623719,
      "loss": 3.3532,
      "step": 3406
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010449227213635708,
      "loss": 3.2354,
      "step": 3407
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010446338292647698,
      "loss": 3.3761,
      "step": 3408
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010443449371659686,
      "loss": 3.2817,
      "step": 3409
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010440560450671675,
      "loss": 3.2361,
      "step": 3410
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010437671529683663,
      "loss": 3.2874,
      "step": 3411
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010434782608695653,
      "loss": 3.3086,
      "step": 3412
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001043189368770764,
      "loss": 3.3358,
      "step": 3413
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001042900476671963,
      "loss": 3.2998,
      "step": 3414
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001042611584573162,
      "loss": 3.3185,
      "step": 3415
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001042322692474361,
      "loss": 3.2077,
      "step": 3416
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010420338003755598,
      "loss": 3.3911,
      "step": 3417
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010417449082767588,
      "loss": 3.3192,
      "step": 3418
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010414560161779577,
      "loss": 3.3712,
      "step": 3419
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010411671240791565,
      "loss": 3.2438,
      "step": 3420
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010408782319803554,
      "loss": 3.3327,
      "step": 3421
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010405893398815542,
      "loss": 3.2294,
      "step": 3422
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010403004477827532,
      "loss": 3.4021,
      "step": 3423
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001040011555683952,
      "loss": 3.3835,
      "step": 3424
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010397226635851509,
      "loss": 3.4587,
      "step": 3425
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010394337714863499,
      "loss": 3.4684,
      "step": 3426
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010391448793875489,
      "loss": 3.3067,
      "step": 3427
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010388559872887477,
      "loss": 3.1844,
      "step": 3428
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010385670951899467,
      "loss": 3.3517,
      "step": 3429
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010382782030911456,
      "loss": 3.3095,
      "step": 3430
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010379893109923444,
      "loss": 3.3435,
      "step": 3431
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010377004188935433,
      "loss": 3.2317,
      "step": 3432
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010374115267947422,
      "loss": 3.2868,
      "step": 3433
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010371226346959411,
      "loss": 3.3199,
      "step": 3434
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010368337425971399,
      "loss": 3.167,
      "step": 3435
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010365448504983388,
      "loss": 3.3248,
      "step": 3436
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010362559583995379,
      "loss": 3.262,
      "step": 3437
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010359670663007368,
      "loss": 3.3651,
      "step": 3438
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010356781742019356,
      "loss": 3.3711,
      "step": 3439
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010353892821031346,
      "loss": 3.2943,
      "step": 3440
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010351003900043334,
      "loss": 3.2476,
      "step": 3441
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010348114979055323,
      "loss": 3.1206,
      "step": 3442
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010345226058067313,
      "loss": 3.3442,
      "step": 3443
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.000103423371370793,
      "loss": 3.378,
      "step": 3444
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001033944821609129,
      "loss": 3.4165,
      "step": 3445
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010336559295103278,
      "loss": 3.3018,
      "step": 3446
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010333670374115267,
      "loss": 3.4229,
      "step": 3447
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010330781453127258,
      "loss": 3.3003,
      "step": 3448
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010327892532139248,
      "loss": 3.2404,
      "step": 3449
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010325003611151236,
      "loss": 3.361,
      "step": 3450
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010322114690163225,
      "loss": 3.2063,
      "step": 3451
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010319225769175213,
      "loss": 3.3266,
      "step": 3452
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010316336848187202,
      "loss": 3.3469,
      "step": 3453
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010313447927199192,
      "loss": 3.2305,
      "step": 3454
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001031055900621118,
      "loss": 3.2308,
      "step": 3455
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010307670085223169,
      "loss": 3.3362,
      "step": 3456
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010304781164235157,
      "loss": 3.3026,
      "step": 3457
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010301892243247149,
      "loss": 3.2693,
      "step": 3458
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010299003322259137,
      "loss": 3.36,
      "step": 3459
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010296114401271127,
      "loss": 3.1219,
      "step": 3460
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010293225480283115,
      "loss": 3.2911,
      "step": 3461
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010290336559295104,
      "loss": 3.4447,
      "step": 3462
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010287447638307092,
      "loss": 3.2468,
      "step": 3463
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010284558717319082,
      "loss": 3.3301,
      "step": 3464
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010281669796331071,
      "loss": 3.2482,
      "step": 3465
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010278780875343059,
      "loss": 3.4332,
      "step": 3466
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010275891954355048,
      "loss": 3.3375,
      "step": 3467
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010273003033367036,
      "loss": 3.2605,
      "step": 3468
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010270114112379028,
      "loss": 3.3278,
      "step": 3469
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010267225191391016,
      "loss": 3.1607,
      "step": 3470
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010264336270403006,
      "loss": 3.1191,
      "step": 3471
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010261447349414994,
      "loss": 3.3582,
      "step": 3472
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010258558428426983,
      "loss": 3.3878,
      "step": 3473
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.00010255669507438971,
      "loss": 3.3283,
      "step": 3474
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001025278058645096,
      "loss": 3.3213,
      "step": 3475
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0001024989166546295,
      "loss": 3.2,
      "step": 3476
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010247002744474938,
      "loss": 3.3431,
      "step": 3477
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010244113823486927,
      "loss": 3.3376,
      "step": 3478
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010241224902498918,
      "loss": 3.3633,
      "step": 3479
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010238335981510908,
      "loss": 3.3458,
      "step": 3480
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010235447060522896,
      "loss": 3.3621,
      "step": 3481
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010232558139534885,
      "loss": 3.3093,
      "step": 3482
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010229669218546873,
      "loss": 3.2172,
      "step": 3483
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010226780297558862,
      "loss": 3.3148,
      "step": 3484
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001022389137657085,
      "loss": 3.2161,
      "step": 3485
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001022100245558284,
      "loss": 3.3887,
      "step": 3486
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010218113534594829,
      "loss": 3.2185,
      "step": 3487
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010215224613606817,
      "loss": 3.2962,
      "step": 3488
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010212335692618807,
      "loss": 3.4699,
      "step": 3489
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010209446771630797,
      "loss": 3.3173,
      "step": 3490
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010206557850642787,
      "loss": 3.3064,
      "step": 3491
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010203668929654775,
      "loss": 3.2422,
      "step": 3492
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010200780008666764,
      "loss": 3.454,
      "step": 3493
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010197891087678752,
      "loss": 3.3527,
      "step": 3494
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010195002166690742,
      "loss": 3.2701,
      "step": 3495
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001019211324570273,
      "loss": 3.3632,
      "step": 3496
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010189224324714719,
      "loss": 3.2785,
      "step": 3497
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010186335403726708,
      "loss": 3.2947,
      "step": 3498
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010183446482738696,
      "loss": 3.31,
      "step": 3499
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010180557561750687,
      "loss": 3.2879,
      "step": 3500
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010177668640762676,
      "loss": 3.3353,
      "step": 3501
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010174779719774666,
      "loss": 3.2496,
      "step": 3502
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010171890798786654,
      "loss": 3.3229,
      "step": 3503
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010169001877798643,
      "loss": 3.2976,
      "step": 3504
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010166112956810631,
      "loss": 3.4325,
      "step": 3505
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010163224035822621,
      "loss": 3.1573,
      "step": 3506
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010160335114834609,
      "loss": 3.1567,
      "step": 3507
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010157446193846598,
      "loss": 3.2892,
      "step": 3508
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010154557272858587,
      "loss": 3.3654,
      "step": 3509
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010151668351870576,
      "loss": 3.2534,
      "step": 3510
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010148779430882566,
      "loss": 3.1937,
      "step": 3511
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010145890509894556,
      "loss": 3.2221,
      "step": 3512
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010143001588906545,
      "loss": 3.145,
      "step": 3513
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010140112667918533,
      "loss": 3.1539,
      "step": 3514
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010137223746930522,
      "loss": 3.34,
      "step": 3515
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001013433482594251,
      "loss": 3.4149,
      "step": 3516
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.000101314459049545,
      "loss": 3.3189,
      "step": 3517
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010128556983966488,
      "loss": 3.2401,
      "step": 3518
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010125668062978477,
      "loss": 3.2709,
      "step": 3519
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010122779141990467,
      "loss": 3.4195,
      "step": 3520
    },
    {
      "epoch": 0.5,
      "eval_loss": 3.5005860328674316,
      "eval_runtime": 472.0927,
      "eval_samples_per_second": 43.396,
      "eval_steps_per_second": 14.465,
      "step": 3520
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010119890221002457,
      "loss": 3.2958,
      "step": 3521
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010117001300014445,
      "loss": 3.4184,
      "step": 3522
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010114112379026435,
      "loss": 3.2575,
      "step": 3523
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010111223458038424,
      "loss": 3.3084,
      "step": 3524
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010108334537050412,
      "loss": 3.1578,
      "step": 3525
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010105445616062402,
      "loss": 3.3057,
      "step": 3526
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0001010255669507439,
      "loss": 3.3618,
      "step": 3527
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010099667774086379,
      "loss": 3.4165,
      "step": 3528
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010096778853098367,
      "loss": 3.0307,
      "step": 3529
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010093889932110356,
      "loss": 3.4035,
      "step": 3530
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010091001011122346,
      "loss": 3.2687,
      "step": 3531
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010088112090134337,
      "loss": 3.2818,
      "step": 3532
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010085223169146325,
      "loss": 3.4536,
      "step": 3533
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010082334248158314,
      "loss": 3.2161,
      "step": 3534
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010079445327170303,
      "loss": 3.1778,
      "step": 3535
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010076556406182291,
      "loss": 3.2902,
      "step": 3536
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010073667485194281,
      "loss": 3.3542,
      "step": 3537
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010070778564206269,
      "loss": 3.3223,
      "step": 3538
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010067889643218258,
      "loss": 3.2997,
      "step": 3539
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010065000722230246,
      "loss": 3.297,
      "step": 3540
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010062111801242236,
      "loss": 3.2426,
      "step": 3541
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010059222880254226,
      "loss": 3.3119,
      "step": 3542
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010056333959266216,
      "loss": 3.3011,
      "step": 3543
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010053445038278204,
      "loss": 3.4242,
      "step": 3544
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010050556117290193,
      "loss": 3.3547,
      "step": 3545
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.00010047667196302182,
      "loss": 3.1823,
      "step": 3546
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001004477827531417,
      "loss": 3.3671,
      "step": 3547
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001004188935432616,
      "loss": 3.2729,
      "step": 3548
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010039000433338148,
      "loss": 3.2731,
      "step": 3549
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010036111512350137,
      "loss": 3.3776,
      "step": 3550
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010033222591362125,
      "loss": 3.3522,
      "step": 3551
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010030333670374115,
      "loss": 3.2845,
      "step": 3552
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010027444749386105,
      "loss": 3.2711,
      "step": 3553
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010024555828398095,
      "loss": 3.3528,
      "step": 3554
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010021666907410083,
      "loss": 3.3548,
      "step": 3555
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010018777986422072,
      "loss": 3.1627,
      "step": 3556
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010015889065434062,
      "loss": 3.2985,
      "step": 3557
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0001001300014444605,
      "loss": 3.4028,
      "step": 3558
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010010111223458039,
      "loss": 3.318,
      "step": 3559
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010007222302470027,
      "loss": 3.2277,
      "step": 3560
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010004333381482016,
      "loss": 3.3456,
      "step": 3561
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.00010001444460494004,
      "loss": 3.2859,
      "step": 3562
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.998555539505995e-05,
      "loss": 3.3805,
      "step": 3563
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.995666618517983e-05,
      "loss": 3.2169,
      "step": 3564
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.992777697529973e-05,
      "loss": 3.2084,
      "step": 3565
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.989888776541962e-05,
      "loss": 3.4749,
      "step": 3566
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.986999855553951e-05,
      "loss": 3.2897,
      "step": 3567
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.98411093456594e-05,
      "loss": 3.3012,
      "step": 3568
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.981222013577929e-05,
      "loss": 3.3644,
      "step": 3569
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.978333092589918e-05,
      "loss": 3.2685,
      "step": 3570
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.975444171601908e-05,
      "loss": 3.4293,
      "step": 3571
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.972555250613897e-05,
      "loss": 3.4324,
      "step": 3572
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.969666329625885e-05,
      "loss": 3.3411,
      "step": 3573
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.966777408637874e-05,
      "loss": 3.3643,
      "step": 3574
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.963888487649862e-05,
      "loss": 3.2441,
      "step": 3575
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.960999566661853e-05,
      "loss": 3.3386,
      "step": 3576
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.958110645673841e-05,
      "loss": 3.3044,
      "step": 3577
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.95522172468583e-05,
      "loss": 3.2719,
      "step": 3578
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.952332803697819e-05,
      "loss": 3.4724,
      "step": 3579
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.949443882709808e-05,
      "loss": 3.4531,
      "step": 3580
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.946554961721797e-05,
      "loss": 3.2413,
      "step": 3581
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.943666040733787e-05,
      "loss": 3.3331,
      "step": 3582
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.940777119745776e-05,
      "loss": 3.2825,
      "step": 3583
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.937888198757764e-05,
      "loss": 3.1648,
      "step": 3584
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.934999277769753e-05,
      "loss": 3.3269,
      "step": 3585
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.932110356781741e-05,
      "loss": 3.2855,
      "step": 3586
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.929221435793732e-05,
      "loss": 3.2763,
      "step": 3587
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.92633251480572e-05,
      "loss": 3.3442,
      "step": 3588
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.92344359381771e-05,
      "loss": 3.3143,
      "step": 3589
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.920554672829698e-05,
      "loss": 3.1068,
      "step": 3590
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.917665751841687e-05,
      "loss": 3.3014,
      "step": 3591
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.914776830853676e-05,
      "loss": 3.2842,
      "step": 3592
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.911887909865666e-05,
      "loss": 3.3375,
      "step": 3593
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.908998988877655e-05,
      "loss": 3.2577,
      "step": 3594
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.906110067889643e-05,
      "loss": 3.3795,
      "step": 3595
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.903221146901633e-05,
      "loss": 3.3381,
      "step": 3596
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.900332225913622e-05,
      "loss": 3.305,
      "step": 3597
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.897443304925611e-05,
      "loss": 3.3266,
      "step": 3598
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.8945543839376e-05,
      "loss": 3.3164,
      "step": 3599
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.891665462949589e-05,
      "loss": 3.4941,
      "step": 3600
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.888776541961577e-05,
      "loss": 3.3784,
      "step": 3601
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.885887620973568e-05,
      "loss": 3.3169,
      "step": 3602
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.882998699985556e-05,
      "loss": 3.3696,
      "step": 3603
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.880109778997545e-05,
      "loss": 3.2779,
      "step": 3604
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.877220858009534e-05,
      "loss": 3.4196,
      "step": 3605
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.874331937021522e-05,
      "loss": 3.364,
      "step": 3606
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.871443016033512e-05,
      "loss": 3.365,
      "step": 3607
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.868554095045501e-05,
      "loss": 3.2851,
      "step": 3608
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.86566517405749e-05,
      "loss": 3.4338,
      "step": 3609
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.862776253069479e-05,
      "loss": 3.3938,
      "step": 3610
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.859887332081468e-05,
      "loss": 3.1677,
      "step": 3611
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.856998411093456e-05,
      "loss": 3.2969,
      "step": 3612
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.854109490105447e-05,
      "loss": 3.1708,
      "step": 3613
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.851220569117435e-05,
      "loss": 3.3359,
      "step": 3614
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.848331648129424e-05,
      "loss": 3.3044,
      "step": 3615
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.845442727141413e-05,
      "loss": 3.3036,
      "step": 3616
    },
    {
      "epoch": 0.51,
      "learning_rate": 9.842553806153401e-05,
      "loss": 3.1884,
      "step": 3617
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.839664885165392e-05,
      "loss": 3.3224,
      "step": 3618
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.83677596417738e-05,
      "loss": 3.3364,
      "step": 3619
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.83388704318937e-05,
      "loss": 3.1456,
      "step": 3620
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.830998122201358e-05,
      "loss": 3.3708,
      "step": 3621
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.828109201213347e-05,
      "loss": 3.351,
      "step": 3622
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.825220280225336e-05,
      "loss": 3.4185,
      "step": 3623
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.822331359237326e-05,
      "loss": 3.2952,
      "step": 3624
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.819442438249314e-05,
      "loss": 3.3222,
      "step": 3625
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.816553517261303e-05,
      "loss": 3.2495,
      "step": 3626
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.813664596273293e-05,
      "loss": 3.3842,
      "step": 3627
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.810775675285282e-05,
      "loss": 3.3091,
      "step": 3628
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.807886754297271e-05,
      "loss": 3.1676,
      "step": 3629
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.80499783330926e-05,
      "loss": 3.2684,
      "step": 3630
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.802108912321249e-05,
      "loss": 3.2345,
      "step": 3631
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.799219991333237e-05,
      "loss": 3.2182,
      "step": 3632
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.796331070345226e-05,
      "loss": 3.3009,
      "step": 3633
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.793442149357216e-05,
      "loss": 3.1447,
      "step": 3634
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.790553228369205e-05,
      "loss": 3.1446,
      "step": 3635
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.787664307381193e-05,
      "loss": 3.2587,
      "step": 3636
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.784775386393182e-05,
      "loss": 3.3582,
      "step": 3637
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.781886465405172e-05,
      "loss": 3.3572,
      "step": 3638
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.778997544417161e-05,
      "loss": 3.3,
      "step": 3639
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.77610862342915e-05,
      "loss": 3.276,
      "step": 3640
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.773219702441139e-05,
      "loss": 3.1791,
      "step": 3641
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.770330781453128e-05,
      "loss": 3.3582,
      "step": 3642
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.767441860465116e-05,
      "loss": 3.4701,
      "step": 3643
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.764552939477107e-05,
      "loss": 3.4079,
      "step": 3644
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.761664018489095e-05,
      "loss": 3.1479,
      "step": 3645
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.758775097501084e-05,
      "loss": 3.3389,
      "step": 3646
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.755886176513072e-05,
      "loss": 3.2733,
      "step": 3647
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.752997255525062e-05,
      "loss": 3.3105,
      "step": 3648
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.750108334537051e-05,
      "loss": 3.3759,
      "step": 3649
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.74721941354904e-05,
      "loss": 3.1932,
      "step": 3650
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.74433049256103e-05,
      "loss": 3.2846,
      "step": 3651
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.741441571573018e-05,
      "loss": 3.3393,
      "step": 3652
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.738552650585007e-05,
      "loss": 3.062,
      "step": 3653
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.735663729596995e-05,
      "loss": 3.3248,
      "step": 3654
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.732774808608986e-05,
      "loss": 3.3914,
      "step": 3655
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.729885887620974e-05,
      "loss": 3.3856,
      "step": 3656
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.726996966632963e-05,
      "loss": 3.2255,
      "step": 3657
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.724108045644951e-05,
      "loss": 3.2713,
      "step": 3658
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.72121912465694e-05,
      "loss": 3.2563,
      "step": 3659
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.71833020366893e-05,
      "loss": 3.351,
      "step": 3660
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.71544128268092e-05,
      "loss": 3.2914,
      "step": 3661
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.712552361692909e-05,
      "loss": 3.2493,
      "step": 3662
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.709663440704897e-05,
      "loss": 3.2843,
      "step": 3663
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.706774519716886e-05,
      "loss": 3.3055,
      "step": 3664
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.703885598728876e-05,
      "loss": 3.3493,
      "step": 3665
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.700996677740865e-05,
      "loss": 3.4551,
      "step": 3666
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.698107756752853e-05,
      "loss": 3.2609,
      "step": 3667
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.695218835764842e-05,
      "loss": 3.2567,
      "step": 3668
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.69232991477683e-05,
      "loss": 3.3386,
      "step": 3669
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.689440993788821e-05,
      "loss": 3.343,
      "step": 3670
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.686552072800809e-05,
      "loss": 3.2134,
      "step": 3671
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.683663151812799e-05,
      "loss": 3.4653,
      "step": 3672
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.680774230824788e-05,
      "loss": 3.305,
      "step": 3673
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.677885309836776e-05,
      "loss": 3.2433,
      "step": 3674
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.674996388848765e-05,
      "loss": 3.3959,
      "step": 3675
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.672107467860755e-05,
      "loss": 3.3096,
      "step": 3676
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.669218546872744e-05,
      "loss": 3.2794,
      "step": 3677
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.666329625884732e-05,
      "loss": 3.2811,
      "step": 3678
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.663440704896722e-05,
      "loss": 3.2929,
      "step": 3679
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.66055178390871e-05,
      "loss": 3.1585,
      "step": 3680
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.6576628629207e-05,
      "loss": 3.2812,
      "step": 3681
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.654773941932688e-05,
      "loss": 3.192,
      "step": 3682
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.651885020944678e-05,
      "loss": 3.3528,
      "step": 3683
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.648996099956667e-05,
      "loss": 3.041,
      "step": 3684
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.646107178968655e-05,
      "loss": 3.3006,
      "step": 3685
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.643218257980645e-05,
      "loss": 3.4209,
      "step": 3686
    },
    {
      "epoch": 0.52,
      "learning_rate": 9.640329336992634e-05,
      "loss": 3.3817,
      "step": 3687
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.637440416004623e-05,
      "loss": 3.3264,
      "step": 3688
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.634551495016611e-05,
      "loss": 3.2459,
      "step": 3689
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.6316625740286e-05,
      "loss": 3.3551,
      "step": 3690
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.62877365304059e-05,
      "loss": 3.2263,
      "step": 3691
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.62588473205258e-05,
      "loss": 3.1857,
      "step": 3692
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.622995811064567e-05,
      "loss": 3.3901,
      "step": 3693
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.620106890076557e-05,
      "loss": 3.348,
      "step": 3694
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.617217969088545e-05,
      "loss": 3.2752,
      "step": 3695
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.614329048100534e-05,
      "loss": 3.3257,
      "step": 3696
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.611440127112524e-05,
      "loss": 3.2138,
      "step": 3697
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.608551206124513e-05,
      "loss": 3.2224,
      "step": 3698
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.605662285136502e-05,
      "loss": 3.3144,
      "step": 3699
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.60277336414849e-05,
      "loss": 2.9823,
      "step": 3700
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.59988444316048e-05,
      "loss": 3.3617,
      "step": 3701
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.596995522172469e-05,
      "loss": 3.1095,
      "step": 3702
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.594106601184459e-05,
      "loss": 3.3501,
      "step": 3703
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.591217680196447e-05,
      "loss": 3.3588,
      "step": 3704
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.588328759208436e-05,
      "loss": 3.2918,
      "step": 3705
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.585439838220424e-05,
      "loss": 3.2578,
      "step": 3706
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.582550917232415e-05,
      "loss": 3.2849,
      "step": 3707
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.579661996244403e-05,
      "loss": 3.3586,
      "step": 3708
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.576773075256392e-05,
      "loss": 3.0917,
      "step": 3709
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.573884154268382e-05,
      "loss": 3.2508,
      "step": 3710
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.57099523328037e-05,
      "loss": 3.3001,
      "step": 3711
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.56810631229236e-05,
      "loss": 3.2744,
      "step": 3712
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.565217391304348e-05,
      "loss": 3.283,
      "step": 3713
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.562328470316338e-05,
      "loss": 3.2927,
      "step": 3714
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.559439549328326e-05,
      "loss": 3.3497,
      "step": 3715
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.556550628340315e-05,
      "loss": 3.1693,
      "step": 3716
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.553661707352305e-05,
      "loss": 3.2967,
      "step": 3717
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.550772786364294e-05,
      "loss": 3.2895,
      "step": 3718
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.547883865376282e-05,
      "loss": 3.2079,
      "step": 3719
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.544994944388271e-05,
      "loss": 3.3238,
      "step": 3720
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.542106023400261e-05,
      "loss": 3.2215,
      "step": 3721
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.539217102412249e-05,
      "loss": 3.2758,
      "step": 3722
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.53632818142424e-05,
      "loss": 3.4565,
      "step": 3723
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.533439260436227e-05,
      "loss": 3.1476,
      "step": 3724
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.530550339448217e-05,
      "loss": 3.1799,
      "step": 3725
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.527661418460205e-05,
      "loss": 3.2619,
      "step": 3726
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.524772497472194e-05,
      "loss": 3.3384,
      "step": 3727
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.521883576484184e-05,
      "loss": 3.3392,
      "step": 3728
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.518994655496173e-05,
      "loss": 3.2635,
      "step": 3729
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.516105734508161e-05,
      "loss": 3.2973,
      "step": 3730
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.51321681352015e-05,
      "loss": 3.1182,
      "step": 3731
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.51032789253214e-05,
      "loss": 3.366,
      "step": 3732
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.507438971544129e-05,
      "loss": 3.2735,
      "step": 3733
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.504550050556119e-05,
      "loss": 3.1093,
      "step": 3734
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.501661129568107e-05,
      "loss": 3.2478,
      "step": 3735
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.498772208580096e-05,
      "loss": 3.2874,
      "step": 3736
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.495883287592084e-05,
      "loss": 3.1797,
      "step": 3737
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.492994366604075e-05,
      "loss": 3.3518,
      "step": 3738
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.490105445616063e-05,
      "loss": 3.3398,
      "step": 3739
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.487216524628052e-05,
      "loss": 3.093,
      "step": 3740
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.48432760364004e-05,
      "loss": 3.443,
      "step": 3741
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.48143868265203e-05,
      "loss": 3.2541,
      "step": 3742
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.478549761664019e-05,
      "loss": 3.2678,
      "step": 3743
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.475660840676008e-05,
      "loss": 3.3353,
      "step": 3744
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.472771919687998e-05,
      "loss": 3.2779,
      "step": 3745
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.469882998699986e-05,
      "loss": 3.386,
      "step": 3746
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.466994077711975e-05,
      "loss": 3.1789,
      "step": 3747
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.464105156723963e-05,
      "loss": 3.2896,
      "step": 3748
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.461216235735954e-05,
      "loss": 3.3565,
      "step": 3749
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.458327314747942e-05,
      "loss": 3.1587,
      "step": 3750
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.455438393759931e-05,
      "loss": 3.3827,
      "step": 3751
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.45254947277192e-05,
      "loss": 3.3155,
      "step": 3752
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.449660551783909e-05,
      "loss": 3.2712,
      "step": 3753
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.446771630795898e-05,
      "loss": 3.2463,
      "step": 3754
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.443882709807888e-05,
      "loss": 3.1915,
      "step": 3755
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.440993788819877e-05,
      "loss": 3.1901,
      "step": 3756
    },
    {
      "epoch": 0.53,
      "learning_rate": 9.438104867831865e-05,
      "loss": 3.2365,
      "step": 3757
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.435215946843854e-05,
      "loss": 3.3428,
      "step": 3758
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.432327025855844e-05,
      "loss": 3.2661,
      "step": 3759
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.429438104867833e-05,
      "loss": 3.3797,
      "step": 3760
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.426549183879821e-05,
      "loss": 3.4258,
      "step": 3761
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.42366026289181e-05,
      "loss": 3.2706,
      "step": 3762
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.420771341903798e-05,
      "loss": 3.2997,
      "step": 3763
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.417882420915788e-05,
      "loss": 3.3774,
      "step": 3764
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.414993499927777e-05,
      "loss": 3.2793,
      "step": 3765
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.412104578939767e-05,
      "loss": 3.2433,
      "step": 3766
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.409215657951756e-05,
      "loss": 3.328,
      "step": 3767
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.406326736963744e-05,
      "loss": 3.2594,
      "step": 3768
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.403437815975733e-05,
      "loss": 3.3453,
      "step": 3769
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.400548894987723e-05,
      "loss": 3.2623,
      "step": 3770
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.397659973999712e-05,
      "loss": 3.3762,
      "step": 3771
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.3947710530117e-05,
      "loss": 3.2156,
      "step": 3772
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.39188213202369e-05,
      "loss": 3.2412,
      "step": 3773
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.388993211035678e-05,
      "loss": 3.3074,
      "step": 3774
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.386104290047668e-05,
      "loss": 3.2955,
      "step": 3775
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.383215369059656e-05,
      "loss": 3.3387,
      "step": 3776
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.380326448071646e-05,
      "loss": 3.2532,
      "step": 3777
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.377437527083635e-05,
      "loss": 3.4034,
      "step": 3778
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.374548606095623e-05,
      "loss": 3.2695,
      "step": 3779
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.371659685107614e-05,
      "loss": 3.3823,
      "step": 3780
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.368770764119602e-05,
      "loss": 3.3463,
      "step": 3781
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.365881843131591e-05,
      "loss": 3.2318,
      "step": 3782
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.36299292214358e-05,
      "loss": 3.1585,
      "step": 3783
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.360104001155569e-05,
      "loss": 3.2317,
      "step": 3784
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.357215080167557e-05,
      "loss": 3.2546,
      "step": 3785
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.354326159179548e-05,
      "loss": 3.3546,
      "step": 3786
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.351437238191536e-05,
      "loss": 3.291,
      "step": 3787
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.348548317203525e-05,
      "loss": 3.2058,
      "step": 3788
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.345659396215514e-05,
      "loss": 3.3634,
      "step": 3789
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.342770475227502e-05,
      "loss": 3.2855,
      "step": 3790
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.339881554239493e-05,
      "loss": 3.3437,
      "step": 3791
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.336992633251481e-05,
      "loss": 3.3228,
      "step": 3792
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.33410371226347e-05,
      "loss": 3.4097,
      "step": 3793
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.331214791275459e-05,
      "loss": 3.2167,
      "step": 3794
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.328325870287448e-05,
      "loss": 3.2635,
      "step": 3795
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.325436949299437e-05,
      "loss": 3.1602,
      "step": 3796
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.322548028311427e-05,
      "loss": 3.2665,
      "step": 3797
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.319659107323415e-05,
      "loss": 3.2843,
      "step": 3798
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.316770186335404e-05,
      "loss": 3.2841,
      "step": 3799
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.313881265347393e-05,
      "loss": 3.0637,
      "step": 3800
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.310992344359383e-05,
      "loss": 3.301,
      "step": 3801
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.308103423371371e-05,
      "loss": 3.3319,
      "step": 3802
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.30521450238336e-05,
      "loss": 3.1172,
      "step": 3803
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.30232558139535e-05,
      "loss": 3.3949,
      "step": 3804
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.299436660407338e-05,
      "loss": 3.3156,
      "step": 3805
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.296547739419327e-05,
      "loss": 3.2566,
      "step": 3806
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.293658818431316e-05,
      "loss": 3.171,
      "step": 3807
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.290769897443306e-05,
      "loss": 3.1897,
      "step": 3808
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.287880976455294e-05,
      "loss": 3.3065,
      "step": 3809
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.284992055467283e-05,
      "loss": 3.3067,
      "step": 3810
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.282103134479273e-05,
      "loss": 3.2692,
      "step": 3811
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.279214213491262e-05,
      "loss": 3.1924,
      "step": 3812
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.27632529250325e-05,
      "loss": 3.241,
      "step": 3813
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.27343637151524e-05,
      "loss": 3.3199,
      "step": 3814
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.270547450527229e-05,
      "loss": 3.2505,
      "step": 3815
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.267658529539217e-05,
      "loss": 3.3046,
      "step": 3816
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.264769608551208e-05,
      "loss": 3.2151,
      "step": 3817
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.261880687563196e-05,
      "loss": 3.2005,
      "step": 3818
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.258991766575185e-05,
      "loss": 3.3655,
      "step": 3819
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.256102845587173e-05,
      "loss": 3.1541,
      "step": 3820
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.253213924599162e-05,
      "loss": 3.2828,
      "step": 3821
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.250325003611152e-05,
      "loss": 3.3187,
      "step": 3822
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.247436082623141e-05,
      "loss": 3.2536,
      "step": 3823
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.244547161635129e-05,
      "loss": 3.3031,
      "step": 3824
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.241658240647119e-05,
      "loss": 3.1891,
      "step": 3825
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.238769319659108e-05,
      "loss": 3.0456,
      "step": 3826
    },
    {
      "epoch": 0.54,
      "learning_rate": 9.235880398671097e-05,
      "loss": 3.3972,
      "step": 3827
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.232991477683087e-05,
      "loss": 3.3103,
      "step": 3828
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.230102556695075e-05,
      "loss": 3.0148,
      "step": 3829
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.227213635707064e-05,
      "loss": 3.2918,
      "step": 3830
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.224324714719052e-05,
      "loss": 3.3179,
      "step": 3831
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.221435793731041e-05,
      "loss": 3.3337,
      "step": 3832
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.218546872743031e-05,
      "loss": 3.3814,
      "step": 3833
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.21565795175502e-05,
      "loss": 3.3344,
      "step": 3834
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.212769030767008e-05,
      "loss": 3.1587,
      "step": 3835
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.209880109778998e-05,
      "loss": 3.2824,
      "step": 3836
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.206991188790987e-05,
      "loss": 3.2514,
      "step": 3837
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.204102267802976e-05,
      "loss": 3.245,
      "step": 3838
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.201213346814966e-05,
      "loss": 3.1114,
      "step": 3839
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.198324425826954e-05,
      "loss": 3.1667,
      "step": 3840
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.195435504838943e-05,
      "loss": 3.2927,
      "step": 3841
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.192546583850931e-05,
      "loss": 3.1644,
      "step": 3842
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.189657662862922e-05,
      "loss": 3.3846,
      "step": 3843
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.18676874187491e-05,
      "loss": 3.3639,
      "step": 3844
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.1838798208869e-05,
      "loss": 3.2076,
      "step": 3845
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.180990899898887e-05,
      "loss": 3.2439,
      "step": 3846
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.178101978910877e-05,
      "loss": 3.2574,
      "step": 3847
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.175213057922866e-05,
      "loss": 3.3944,
      "step": 3848
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.172324136934856e-05,
      "loss": 3.3432,
      "step": 3849
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.169435215946845e-05,
      "loss": 3.451,
      "step": 3850
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.166546294958833e-05,
      "loss": 3.4377,
      "step": 3851
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.163657373970822e-05,
      "loss": 3.2292,
      "step": 3852
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.16076845298281e-05,
      "loss": 3.2505,
      "step": 3853
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.157879531994801e-05,
      "loss": 3.1625,
      "step": 3854
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.154990611006789e-05,
      "loss": 3.3417,
      "step": 3855
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.152101690018779e-05,
      "loss": 3.3231,
      "step": 3856
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.149212769030767e-05,
      "loss": 3.4253,
      "step": 3857
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.146323848042756e-05,
      "loss": 3.2481,
      "step": 3858
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.143434927054745e-05,
      "loss": 3.2928,
      "step": 3859
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.140546006066735e-05,
      "loss": 3.3022,
      "step": 3860
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.137657085078724e-05,
      "loss": 3.2664,
      "step": 3861
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.134768164090712e-05,
      "loss": 3.3661,
      "step": 3862
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.131879243102702e-05,
      "loss": 3.2142,
      "step": 3863
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.128990322114691e-05,
      "loss": 3.0478,
      "step": 3864
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.12610140112668e-05,
      "loss": 3.285,
      "step": 3865
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.123212480138668e-05,
      "loss": 3.1,
      "step": 3866
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.120323559150658e-05,
      "loss": 3.2835,
      "step": 3867
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.117434638162646e-05,
      "loss": 3.3589,
      "step": 3868
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.114545717174636e-05,
      "loss": 3.2728,
      "step": 3869
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.111656796186624e-05,
      "loss": 3.2323,
      "step": 3870
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.108767875198614e-05,
      "loss": 3.2699,
      "step": 3871
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.105878954210603e-05,
      "loss": 3.1105,
      "step": 3872
    },
    {
      "epoch": 0.55,
      "eval_loss": 3.4803874492645264,
      "eval_runtime": 471.7317,
      "eval_samples_per_second": 43.429,
      "eval_steps_per_second": 14.476,
      "step": 3872
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.102990033222591e-05,
      "loss": 3.3334,
      "step": 3873
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.10010111223458e-05,
      "loss": 3.2822,
      "step": 3874
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.09721219124657e-05,
      "loss": 3.1631,
      "step": 3875
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.09432327025856e-05,
      "loss": 3.2771,
      "step": 3876
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.091434349270547e-05,
      "loss": 3.2999,
      "step": 3877
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.088545428282537e-05,
      "loss": 3.19,
      "step": 3878
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.085656507294525e-05,
      "loss": 3.405,
      "step": 3879
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.082767586306516e-05,
      "loss": 3.3027,
      "step": 3880
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.079878665318504e-05,
      "loss": 3.1167,
      "step": 3881
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.076989744330493e-05,
      "loss": 3.2915,
      "step": 3882
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.074100823342482e-05,
      "loss": 3.2832,
      "step": 3883
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.07121190235447e-05,
      "loss": 3.032,
      "step": 3884
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.068322981366461e-05,
      "loss": 3.1557,
      "step": 3885
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.065434060378449e-05,
      "loss": 3.199,
      "step": 3886
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.062545139390439e-05,
      "loss": 3.2869,
      "step": 3887
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.059656218402427e-05,
      "loss": 3.3465,
      "step": 3888
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.056767297414416e-05,
      "loss": 3.2342,
      "step": 3889
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.053878376426405e-05,
      "loss": 3.2444,
      "step": 3890
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.050989455438395e-05,
      "loss": 3.198,
      "step": 3891
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.048100534450383e-05,
      "loss": 3.2915,
      "step": 3892
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.045211613462372e-05,
      "loss": 3.181,
      "step": 3893
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.042322692474362e-05,
      "loss": 3.3077,
      "step": 3894
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.03943377148635e-05,
      "loss": 3.298,
      "step": 3895
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.03654485049834e-05,
      "loss": 3.032,
      "step": 3896
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.033655929510328e-05,
      "loss": 3.2134,
      "step": 3897
    },
    {
      "epoch": 0.55,
      "learning_rate": 9.030767008522318e-05,
      "loss": 3.2998,
      "step": 3898
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.027878087534306e-05,
      "loss": 3.1998,
      "step": 3899
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.024989166546295e-05,
      "loss": 3.1195,
      "step": 3900
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.022100245558285e-05,
      "loss": 3.2613,
      "step": 3901
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.019211324570274e-05,
      "loss": 3.364,
      "step": 3902
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.016322403582262e-05,
      "loss": 3.2895,
      "step": 3903
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.013433482594251e-05,
      "loss": 3.2736,
      "step": 3904
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.01054456160624e-05,
      "loss": 3.3313,
      "step": 3905
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.00765564061823e-05,
      "loss": 3.2997,
      "step": 3906
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.00476671963022e-05,
      "loss": 3.3214,
      "step": 3907
    },
    {
      "epoch": 0.56,
      "learning_rate": 9.001877798642207e-05,
      "loss": 3.3161,
      "step": 3908
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.998988877654197e-05,
      "loss": 3.3021,
      "step": 3909
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.996099956666185e-05,
      "loss": 3.3702,
      "step": 3910
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.993211035678176e-05,
      "loss": 3.2741,
      "step": 3911
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.990322114690164e-05,
      "loss": 3.289,
      "step": 3912
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.987433193702153e-05,
      "loss": 3.3696,
      "step": 3913
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.984544272714141e-05,
      "loss": 3.3681,
      "step": 3914
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.98165535172613e-05,
      "loss": 3.2758,
      "step": 3915
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.97876643073812e-05,
      "loss": 3.2286,
      "step": 3916
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.975877509750109e-05,
      "loss": 3.2878,
      "step": 3917
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.972988588762099e-05,
      "loss": 3.4119,
      "step": 3918
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.970099667774087e-05,
      "loss": 3.2123,
      "step": 3919
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.967210746786076e-05,
      "loss": 3.0235,
      "step": 3920
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.964321825798064e-05,
      "loss": 3.3914,
      "step": 3921
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.961432904810055e-05,
      "loss": 3.1288,
      "step": 3922
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.958543983822043e-05,
      "loss": 3.238,
      "step": 3923
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.955655062834032e-05,
      "loss": 3.1957,
      "step": 3924
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.95276614184602e-05,
      "loss": 3.249,
      "step": 3925
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.94987722085801e-05,
      "loss": 3.181,
      "step": 3926
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.946988299869999e-05,
      "loss": 3.281,
      "step": 3927
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.944099378881988e-05,
      "loss": 3.2452,
      "step": 3928
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.941210457893976e-05,
      "loss": 3.2608,
      "step": 3929
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.938321536905966e-05,
      "loss": 3.3254,
      "step": 3930
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.935432615917955e-05,
      "loss": 3.3185,
      "step": 3931
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.932543694929945e-05,
      "loss": 3.2227,
      "step": 3932
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.929654773941934e-05,
      "loss": 3.2378,
      "step": 3933
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.926765852953922e-05,
      "loss": 3.3324,
      "step": 3934
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.923876931965911e-05,
      "loss": 3.4224,
      "step": 3935
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.9209880109779e-05,
      "loss": 3.2827,
      "step": 3936
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.91809908998989e-05,
      "loss": 3.2855,
      "step": 3937
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.915210169001878e-05,
      "loss": 3.0395,
      "step": 3938
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.912321248013867e-05,
      "loss": 3.3829,
      "step": 3939
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.909432327025856e-05,
      "loss": 3.2371,
      "step": 3940
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.906543406037845e-05,
      "loss": 3.2505,
      "step": 3941
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.903654485049834e-05,
      "loss": 3.3156,
      "step": 3942
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.900765564061824e-05,
      "loss": 3.0712,
      "step": 3943
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.897876643073813e-05,
      "loss": 3.3497,
      "step": 3944
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.894987722085801e-05,
      "loss": 3.2546,
      "step": 3945
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.89209880109779e-05,
      "loss": 3.153,
      "step": 3946
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.889209880109778e-05,
      "loss": 3.0865,
      "step": 3947
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.886320959121769e-05,
      "loss": 3.2698,
      "step": 3948
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.883432038133757e-05,
      "loss": 3.2959,
      "step": 3949
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.880543117145747e-05,
      "loss": 3.2559,
      "step": 3950
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.877654196157735e-05,
      "loss": 3.3895,
      "step": 3951
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.874765275169724e-05,
      "loss": 3.3017,
      "step": 3952
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.871876354181713e-05,
      "loss": 3.1077,
      "step": 3953
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.868987433193703e-05,
      "loss": 3.1145,
      "step": 3954
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.866098512205692e-05,
      "loss": 3.0593,
      "step": 3955
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.86320959121768e-05,
      "loss": 3.1752,
      "step": 3956
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.86032067022967e-05,
      "loss": 3.1846,
      "step": 3957
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.857431749241659e-05,
      "loss": 3.3596,
      "step": 3958
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.854542828253648e-05,
      "loss": 3.2812,
      "step": 3959
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.851653907265636e-05,
      "loss": 3.1094,
      "step": 3960
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.848764986277626e-05,
      "loss": 3.2295,
      "step": 3961
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.845876065289614e-05,
      "loss": 3.2622,
      "step": 3962
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.842987144301603e-05,
      "loss": 3.0867,
      "step": 3963
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.840098223313593e-05,
      "loss": 3.2774,
      "step": 3964
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.837209302325582e-05,
      "loss": 3.3415,
      "step": 3965
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.834320381337571e-05,
      "loss": 3.2803,
      "step": 3966
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.83143146034956e-05,
      "loss": 3.3382,
      "step": 3967
    },
    {
      "epoch": 0.56,
      "learning_rate": 8.828542539361549e-05,
      "loss": 3.2908,
      "step": 3968
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.825653618373538e-05,
      "loss": 3.2303,
      "step": 3969
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.822764697385528e-05,
      "loss": 3.2081,
      "step": 3970
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.819875776397516e-05,
      "loss": 3.1726,
      "step": 3971
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.816986855409505e-05,
      "loss": 3.2753,
      "step": 3972
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.814097934421493e-05,
      "loss": 3.1992,
      "step": 3973
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.811209013433484e-05,
      "loss": 3.1367,
      "step": 3974
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.808320092445472e-05,
      "loss": 3.3159,
      "step": 3975
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.805431171457461e-05,
      "loss": 3.3195,
      "step": 3976
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.80254225046945e-05,
      "loss": 3.3893,
      "step": 3977
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.799653329481438e-05,
      "loss": 3.2885,
      "step": 3978
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.796764408493429e-05,
      "loss": 3.2368,
      "step": 3979
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.793875487505417e-05,
      "loss": 3.2786,
      "step": 3980
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.790986566517407e-05,
      "loss": 3.1629,
      "step": 3981
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.788097645529395e-05,
      "loss": 3.1387,
      "step": 3982
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.785208724541384e-05,
      "loss": 3.2488,
      "step": 3983
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.782319803553372e-05,
      "loss": 3.2105,
      "step": 3984
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.779430882565363e-05,
      "loss": 3.2135,
      "step": 3985
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.776541961577351e-05,
      "loss": 3.1729,
      "step": 3986
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.77365304058934e-05,
      "loss": 3.286,
      "step": 3987
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.77076411960133e-05,
      "loss": 3.224,
      "step": 3988
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.767875198613318e-05,
      "loss": 3.298,
      "step": 3989
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.764986277625308e-05,
      "loss": 3.1093,
      "step": 3990
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.762097356637296e-05,
      "loss": 3.1439,
      "step": 3991
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.759208435649286e-05,
      "loss": 3.3523,
      "step": 3992
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.756319514661274e-05,
      "loss": 3.172,
      "step": 3993
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.753430593673263e-05,
      "loss": 3.3619,
      "step": 3994
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.750541672685253e-05,
      "loss": 3.366,
      "step": 3995
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.747652751697242e-05,
      "loss": 3.3856,
      "step": 3996
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.74476383070923e-05,
      "loss": 3.2243,
      "step": 3997
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.74187490972122e-05,
      "loss": 3.2988,
      "step": 3998
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.738985988733209e-05,
      "loss": 3.3289,
      "step": 3999
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.736097067745198e-05,
      "loss": 3.259,
      "step": 4000
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.733208146757188e-05,
      "loss": 3.258,
      "step": 4001
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.730319225769176e-05,
      "loss": 3.406,
      "step": 4002
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.727430304781165e-05,
      "loss": 3.3033,
      "step": 4003
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.724541383793153e-05,
      "loss": 3.2368,
      "step": 4004
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.721652462805142e-05,
      "loss": 3.0177,
      "step": 4005
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.718763541817132e-05,
      "loss": 3.2402,
      "step": 4006
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.715874620829121e-05,
      "loss": 3.1792,
      "step": 4007
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.712985699841109e-05,
      "loss": 3.092,
      "step": 4008
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.710096778853099e-05,
      "loss": 3.124,
      "step": 4009
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.707207857865088e-05,
      "loss": 3.3448,
      "step": 4010
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.704318936877077e-05,
      "loss": 3.331,
      "step": 4011
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.701430015889067e-05,
      "loss": 3.339,
      "step": 4012
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.698541094901055e-05,
      "loss": 3.1541,
      "step": 4013
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.695652173913044e-05,
      "loss": 3.3561,
      "step": 4014
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.692763252925032e-05,
      "loss": 3.1285,
      "step": 4015
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.689874331937023e-05,
      "loss": 3.4157,
      "step": 4016
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.686985410949011e-05,
      "loss": 3.2565,
      "step": 4017
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.684096489961e-05,
      "loss": 3.2633,
      "step": 4018
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.681207568972988e-05,
      "loss": 3.2736,
      "step": 4019
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.678318647984978e-05,
      "loss": 3.2032,
      "step": 4020
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.675429726996967e-05,
      "loss": 3.2755,
      "step": 4021
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.672540806008956e-05,
      "loss": 3.0297,
      "step": 4022
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.669651885020946e-05,
      "loss": 3.1787,
      "step": 4023
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.666762964032934e-05,
      "loss": 3.2807,
      "step": 4024
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.663874043044923e-05,
      "loss": 3.252,
      "step": 4025
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.660985122056913e-05,
      "loss": 3.2084,
      "step": 4026
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.658096201068902e-05,
      "loss": 3.2354,
      "step": 4027
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.65520728008089e-05,
      "loss": 3.2249,
      "step": 4028
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.65231835909288e-05,
      "loss": 3.2624,
      "step": 4029
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.649429438104867e-05,
      "loss": 3.2647,
      "step": 4030
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.646540517116857e-05,
      "loss": 3.3059,
      "step": 4031
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.643651596128846e-05,
      "loss": 3.3355,
      "step": 4032
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.640762675140836e-05,
      "loss": 3.2986,
      "step": 4033
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.637873754152825e-05,
      "loss": 3.4018,
      "step": 4034
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.634984833164813e-05,
      "loss": 3.2099,
      "step": 4035
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.632095912176802e-05,
      "loss": 3.2506,
      "step": 4036
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.629206991188792e-05,
      "loss": 3.0363,
      "step": 4037
    },
    {
      "epoch": 0.57,
      "learning_rate": 8.626318070200781e-05,
      "loss": 3.3057,
      "step": 4038
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.623429149212769e-05,
      "loss": 3.174,
      "step": 4039
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.620540228224759e-05,
      "loss": 3.1564,
      "step": 4040
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.617651307236747e-05,
      "loss": 3.2195,
      "step": 4041
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.614762386248737e-05,
      "loss": 3.3382,
      "step": 4042
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.611873465260725e-05,
      "loss": 3.2327,
      "step": 4043
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.608984544272715e-05,
      "loss": 3.2639,
      "step": 4044
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.606095623284704e-05,
      "loss": 3.3016,
      "step": 4045
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.603206702296692e-05,
      "loss": 3.2926,
      "step": 4046
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.600317781308681e-05,
      "loss": 3.277,
      "step": 4047
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.597428860320671e-05,
      "loss": 3.0697,
      "step": 4048
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.59453993933266e-05,
      "loss": 3.2062,
      "step": 4049
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.591651018344648e-05,
      "loss": 3.2135,
      "step": 4050
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.588762097356638e-05,
      "loss": 3.3941,
      "step": 4051
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.585873176368626e-05,
      "loss": 3.2611,
      "step": 4052
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.582984255380616e-05,
      "loss": 3.1288,
      "step": 4053
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.580095334392604e-05,
      "loss": 3.1586,
      "step": 4054
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.577206413404594e-05,
      "loss": 3.2621,
      "step": 4055
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.574317492416582e-05,
      "loss": 3.1419,
      "step": 4056
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.571428571428571e-05,
      "loss": 3.2366,
      "step": 4057
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.56853965044056e-05,
      "loss": 3.0991,
      "step": 4058
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.56565072945255e-05,
      "loss": 3.2461,
      "step": 4059
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.56276180846454e-05,
      "loss": 3.157,
      "step": 4060
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.559872887476527e-05,
      "loss": 3.279,
      "step": 4061
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.556983966488517e-05,
      "loss": 3.3706,
      "step": 4062
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.554095045500506e-05,
      "loss": 3.2767,
      "step": 4063
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.551206124512496e-05,
      "loss": 3.2843,
      "step": 4064
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.548317203524484e-05,
      "loss": 3.2912,
      "step": 4065
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.545428282536473e-05,
      "loss": 3.3588,
      "step": 4066
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.542539361548461e-05,
      "loss": 3.1223,
      "step": 4067
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.539650440560452e-05,
      "loss": 3.1939,
      "step": 4068
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.53676151957244e-05,
      "loss": 3.2885,
      "step": 4069
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.533872598584429e-05,
      "loss": 3.1524,
      "step": 4070
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.530983677596419e-05,
      "loss": 3.2319,
      "step": 4071
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.528094756608407e-05,
      "loss": 3.2579,
      "step": 4072
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.525205835620396e-05,
      "loss": 3.2752,
      "step": 4073
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.522316914632385e-05,
      "loss": 3.1005,
      "step": 4074
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.519427993644375e-05,
      "loss": 3.2904,
      "step": 4075
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.516539072656363e-05,
      "loss": 3.2652,
      "step": 4076
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.513650151668352e-05,
      "loss": 3.3645,
      "step": 4077
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.51076123068034e-05,
      "loss": 3.3647,
      "step": 4078
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.507872309692331e-05,
      "loss": 3.197,
      "step": 4079
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.504983388704319e-05,
      "loss": 3.2224,
      "step": 4080
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.502094467716308e-05,
      "loss": 3.2377,
      "step": 4081
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.499205546728298e-05,
      "loss": 3.2946,
      "step": 4082
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.496316625740286e-05,
      "loss": 3.2442,
      "step": 4083
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.493427704752276e-05,
      "loss": 3.2016,
      "step": 4084
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.490538783764264e-05,
      "loss": 3.2061,
      "step": 4085
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.487649862776254e-05,
      "loss": 3.3251,
      "step": 4086
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.484760941788242e-05,
      "loss": 3.3625,
      "step": 4087
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.481872020800231e-05,
      "loss": 3.0163,
      "step": 4088
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.47898309981222e-05,
      "loss": 3.1424,
      "step": 4089
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.47609417882421e-05,
      "loss": 3.3728,
      "step": 4090
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.473205257836198e-05,
      "loss": 3.2705,
      "step": 4091
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.470316336848187e-05,
      "loss": 3.3873,
      "step": 4092
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.467427415860177e-05,
      "loss": 3.3183,
      "step": 4093
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.464538494872165e-05,
      "loss": 3.2233,
      "step": 4094
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.461649573884156e-05,
      "loss": 3.2937,
      "step": 4095
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.458760652896144e-05,
      "loss": 3.3,
      "step": 4096
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.455871731908133e-05,
      "loss": 3.148,
      "step": 4097
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.452982810920121e-05,
      "loss": 3.2463,
      "step": 4098
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.45009388993211e-05,
      "loss": 3.2626,
      "step": 4099
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.4472049689441e-05,
      "loss": 3.2149,
      "step": 4100
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.444316047956089e-05,
      "loss": 3.2738,
      "step": 4101
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.441427126968077e-05,
      "loss": 3.258,
      "step": 4102
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.438538205980067e-05,
      "loss": 3.2738,
      "step": 4103
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.435649284992056e-05,
      "loss": 3.14,
      "step": 4104
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.432760364004045e-05,
      "loss": 3.409,
      "step": 4105
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.429871443016035e-05,
      "loss": 3.2291,
      "step": 4106
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.426982522028023e-05,
      "loss": 3.0735,
      "step": 4107
    },
    {
      "epoch": 0.58,
      "learning_rate": 8.424093601040012e-05,
      "loss": 3.3047,
      "step": 4108
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.421204680052e-05,
      "loss": 3.0758,
      "step": 4109
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.418315759063991e-05,
      "loss": 3.1981,
      "step": 4110
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.415426838075979e-05,
      "loss": 3.2431,
      "step": 4111
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.412537917087968e-05,
      "loss": 3.2672,
      "step": 4112
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.409648996099956e-05,
      "loss": 3.1203,
      "step": 4113
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.406760075111946e-05,
      "loss": 3.1629,
      "step": 4114
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.403871154123935e-05,
      "loss": 3.2416,
      "step": 4115
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.400982233135925e-05,
      "loss": 3.074,
      "step": 4116
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.398093312147914e-05,
      "loss": 3.3175,
      "step": 4117
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.395204391159902e-05,
      "loss": 3.2744,
      "step": 4118
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.392315470171891e-05,
      "loss": 3.289,
      "step": 4119
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.389426549183879e-05,
      "loss": 3.2475,
      "step": 4120
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.38653762819587e-05,
      "loss": 3.3102,
      "step": 4121
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.383648707207858e-05,
      "loss": 3.3482,
      "step": 4122
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.380759786219847e-05,
      "loss": 3.2501,
      "step": 4123
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.377870865231835e-05,
      "loss": 3.0875,
      "step": 4124
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.374981944243825e-05,
      "loss": 3.3484,
      "step": 4125
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.372093023255814e-05,
      "loss": 3.3203,
      "step": 4126
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.369204102267804e-05,
      "loss": 2.9373,
      "step": 4127
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.366315181279793e-05,
      "loss": 3.3178,
      "step": 4128
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.363426260291781e-05,
      "loss": 3.1093,
      "step": 4129
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.36053733930377e-05,
      "loss": 3.1304,
      "step": 4130
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.35764841831576e-05,
      "loss": 3.1829,
      "step": 4131
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.354759497327749e-05,
      "loss": 3.3268,
      "step": 4132
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.351870576339737e-05,
      "loss": 3.2223,
      "step": 4133
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.348981655351727e-05,
      "loss": 3.1491,
      "step": 4134
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.346092734363715e-05,
      "loss": 3.3163,
      "step": 4135
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.343203813375705e-05,
      "loss": 3.1891,
      "step": 4136
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.340314892387693e-05,
      "loss": 3.2958,
      "step": 4137
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.337425971399683e-05,
      "loss": 3.2614,
      "step": 4138
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.334537050411672e-05,
      "loss": 3.1945,
      "step": 4139
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.33164812942366e-05,
      "loss": 3.2473,
      "step": 4140
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.32875920843565e-05,
      "loss": 3.3148,
      "step": 4141
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.325870287447639e-05,
      "loss": 3.2322,
      "step": 4142
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.322981366459628e-05,
      "loss": 3.1535,
      "step": 4143
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.320092445471616e-05,
      "loss": 3.348,
      "step": 4144
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.317203524483606e-05,
      "loss": 3.3239,
      "step": 4145
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.314314603495594e-05,
      "loss": 3.2792,
      "step": 4146
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.311425682507585e-05,
      "loss": 3.249,
      "step": 4147
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.308536761519573e-05,
      "loss": 3.2805,
      "step": 4148
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.305647840531562e-05,
      "loss": 3.2893,
      "step": 4149
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.302758919543551e-05,
      "loss": 3.2423,
      "step": 4150
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.29986999855554e-05,
      "loss": 3.2868,
      "step": 4151
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.29698107756753e-05,
      "loss": 3.3046,
      "step": 4152
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.294092156579518e-05,
      "loss": 3.0475,
      "step": 4153
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.291203235591507e-05,
      "loss": 3.4062,
      "step": 4154
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.288314314603496e-05,
      "loss": 3.2152,
      "step": 4155
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.285425393615485e-05,
      "loss": 3.3383,
      "step": 4156
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.282536472627474e-05,
      "loss": 3.2178,
      "step": 4157
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.279647551639464e-05,
      "loss": 3.221,
      "step": 4158
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.276758630651452e-05,
      "loss": 3.2245,
      "step": 4159
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.273869709663441e-05,
      "loss": 3.2269,
      "step": 4160
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.27098078867543e-05,
      "loss": 3.2183,
      "step": 4161
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.268091867687418e-05,
      "loss": 3.1523,
      "step": 4162
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.265202946699408e-05,
      "loss": 3.1758,
      "step": 4163
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.262314025711397e-05,
      "loss": 3.2164,
      "step": 4164
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.259425104723387e-05,
      "loss": 3.2022,
      "step": 4165
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.256536183735375e-05,
      "loss": 3.2189,
      "step": 4166
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.253647262747364e-05,
      "loss": 3.2521,
      "step": 4167
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.250758341759353e-05,
      "loss": 3.2237,
      "step": 4168
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.247869420771343e-05,
      "loss": 3.2787,
      "step": 4169
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.244980499783331e-05,
      "loss": 3.1623,
      "step": 4170
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.24209157879532e-05,
      "loss": 3.2421,
      "step": 4171
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.23920265780731e-05,
      "loss": 3.239,
      "step": 4172
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.236313736819299e-05,
      "loss": 3.2289,
      "step": 4173
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.233424815831287e-05,
      "loss": 3.2067,
      "step": 4174
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.230535894843276e-05,
      "loss": 3.1906,
      "step": 4175
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.227646973855266e-05,
      "loss": 3.3086,
      "step": 4176
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.224758052867254e-05,
      "loss": 3.205,
      "step": 4177
    },
    {
      "epoch": 0.59,
      "learning_rate": 8.221869131879245e-05,
      "loss": 3.1586,
      "step": 4178
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.218980210891233e-05,
      "loss": 3.2744,
      "step": 4179
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.216091289903222e-05,
      "loss": 3.1454,
      "step": 4180
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.21320236891521e-05,
      "loss": 3.2559,
      "step": 4181
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.2103134479272e-05,
      "loss": 3.2432,
      "step": 4182
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.207424526939189e-05,
      "loss": 3.2066,
      "step": 4183
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.204535605951178e-05,
      "loss": 3.1526,
      "step": 4184
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.201646684963166e-05,
      "loss": 3.1424,
      "step": 4185
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.198757763975156e-05,
      "loss": 3.1921,
      "step": 4186
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.195868842987145e-05,
      "loss": 3.1171,
      "step": 4187
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.192979921999133e-05,
      "loss": 3.1883,
      "step": 4188
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.190091001011124e-05,
      "loss": 3.3401,
      "step": 4189
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.187202080023112e-05,
      "loss": 3.1864,
      "step": 4190
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.184313159035101e-05,
      "loss": 3.259,
      "step": 4191
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.181424238047089e-05,
      "loss": 3.1787,
      "step": 4192
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.178535317059078e-05,
      "loss": 3.1158,
      "step": 4193
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.175646396071068e-05,
      "loss": 3.2656,
      "step": 4194
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.172757475083057e-05,
      "loss": 3.2065,
      "step": 4195
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.169868554095045e-05,
      "loss": 3.2744,
      "step": 4196
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.166979633107035e-05,
      "loss": 3.0871,
      "step": 4197
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.164090712119024e-05,
      "loss": 3.3143,
      "step": 4198
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.161201791131013e-05,
      "loss": 3.1667,
      "step": 4199
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.158312870143003e-05,
      "loss": 3.2188,
      "step": 4200
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.155423949154991e-05,
      "loss": 3.1783,
      "step": 4201
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.15253502816698e-05,
      "loss": 3.2553,
      "step": 4202
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.149646107178968e-05,
      "loss": 3.3402,
      "step": 4203
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.146757186190958e-05,
      "loss": 3.2964,
      "step": 4204
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.143868265202947e-05,
      "loss": 3.258,
      "step": 4205
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.140979344214936e-05,
      "loss": 3.1899,
      "step": 4206
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.138090423226924e-05,
      "loss": 3.2875,
      "step": 4207
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.135201502238914e-05,
      "loss": 3.2679,
      "step": 4208
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.132312581250903e-05,
      "loss": 3.0585,
      "step": 4209
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.129423660262893e-05,
      "loss": 3.2907,
      "step": 4210
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.126534739274882e-05,
      "loss": 3.1095,
      "step": 4211
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.12364581828687e-05,
      "loss": 3.2634,
      "step": 4212
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.12075689729886e-05,
      "loss": 3.3216,
      "step": 4213
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.117867976310847e-05,
      "loss": 3.2927,
      "step": 4214
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.114979055322838e-05,
      "loss": 3.2301,
      "step": 4215
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.112090134334826e-05,
      "loss": 3.2072,
      "step": 4216
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.109201213346816e-05,
      "loss": 3.283,
      "step": 4217
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.106312292358804e-05,
      "loss": 3.1787,
      "step": 4218
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.103423371370793e-05,
      "loss": 3.2416,
      "step": 4219
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.100534450382782e-05,
      "loss": 3.0936,
      "step": 4220
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.097645529394772e-05,
      "loss": 3.2907,
      "step": 4221
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.094756608406761e-05,
      "loss": 3.1733,
      "step": 4222
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.091867687418749e-05,
      "loss": 3.381,
      "step": 4223
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.088978766430739e-05,
      "loss": 3.2477,
      "step": 4224
    },
    {
      "epoch": 0.6,
      "eval_loss": 3.454590082168579,
      "eval_runtime": 471.9004,
      "eval_samples_per_second": 43.414,
      "eval_steps_per_second": 14.471,
      "step": 4224
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.086089845442728e-05,
      "loss": 3.1189,
      "step": 4225
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.083200924454717e-05,
      "loss": 3.2583,
      "step": 4226
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.080312003466705e-05,
      "loss": 3.3358,
      "step": 4227
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.077423082478695e-05,
      "loss": 3.2218,
      "step": 4228
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.074534161490683e-05,
      "loss": 3.2638,
      "step": 4229
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.071645240502672e-05,
      "loss": 3.1001,
      "step": 4230
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.068756319514661e-05,
      "loss": 3.1898,
      "step": 4231
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.065867398526651e-05,
      "loss": 3.2099,
      "step": 4232
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.06297847753864e-05,
      "loss": 3.2063,
      "step": 4233
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.060089556550628e-05,
      "loss": 3.223,
      "step": 4234
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.057200635562618e-05,
      "loss": 3.2809,
      "step": 4235
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.054311714574607e-05,
      "loss": 3.2364,
      "step": 4236
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.051422793586596e-05,
      "loss": 3.132,
      "step": 4237
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.048533872598584e-05,
      "loss": 3.1899,
      "step": 4238
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.045644951610574e-05,
      "loss": 3.2914,
      "step": 4239
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.042756030622562e-05,
      "loss": 3.1883,
      "step": 4240
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.039867109634553e-05,
      "loss": 3.1069,
      "step": 4241
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.03697818864654e-05,
      "loss": 3.2493,
      "step": 4242
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.03408926765853e-05,
      "loss": 3.1019,
      "step": 4243
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.03120034667052e-05,
      "loss": 3.4186,
      "step": 4244
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.028311425682507e-05,
      "loss": 2.9726,
      "step": 4245
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.025422504694498e-05,
      "loss": 3.1249,
      "step": 4246
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.022533583706486e-05,
      "loss": 3.0704,
      "step": 4247
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.019644662718476e-05,
      "loss": 3.1305,
      "step": 4248
    },
    {
      "epoch": 0.6,
      "learning_rate": 8.016755741730464e-05,
      "loss": 3.2742,
      "step": 4249
    },
    {
      "epoch": 0.61,
      "learning_rate": 8.013866820742453e-05,
      "loss": 3.2609,
      "step": 4250
    },
    {
      "epoch": 0.61,
      "learning_rate": 8.010977899754441e-05,
      "loss": 3.2674,
      "step": 4251
    },
    {
      "epoch": 0.61,
      "learning_rate": 8.008088978766432e-05,
      "loss": 3.274,
      "step": 4252
    },
    {
      "epoch": 0.61,
      "learning_rate": 8.00520005777842e-05,
      "loss": 3.1981,
      "step": 4253
    },
    {
      "epoch": 0.61,
      "learning_rate": 8.002311136790409e-05,
      "loss": 3.1098,
      "step": 4254
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.999422215802399e-05,
      "loss": 3.3066,
      "step": 4255
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.996533294814387e-05,
      "loss": 3.2974,
      "step": 4256
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.993644373826377e-05,
      "loss": 3.2273,
      "step": 4257
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.990755452838365e-05,
      "loss": 3.3769,
      "step": 4258
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.987866531850355e-05,
      "loss": 3.2165,
      "step": 4259
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.984977610862343e-05,
      "loss": 3.2988,
      "step": 4260
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.982088689874332e-05,
      "loss": 3.1474,
      "step": 4261
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.979199768886321e-05,
      "loss": 3.1661,
      "step": 4262
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.976310847898311e-05,
      "loss": 2.9669,
      "step": 4263
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.973421926910299e-05,
      "loss": 3.1976,
      "step": 4264
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.970533005922288e-05,
      "loss": 3.2131,
      "step": 4265
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.967644084934278e-05,
      "loss": 3.2106,
      "step": 4266
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.964755163946267e-05,
      "loss": 3.2483,
      "step": 4267
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.961866242958256e-05,
      "loss": 3.3123,
      "step": 4268
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.958977321970244e-05,
      "loss": 3.2434,
      "step": 4269
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.956088400982234e-05,
      "loss": 3.2349,
      "step": 4270
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.953199479994222e-05,
      "loss": 3.2684,
      "step": 4271
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.950310559006211e-05,
      "loss": 3.3313,
      "step": 4272
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.9474216380182e-05,
      "loss": 3.1969,
      "step": 4273
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.94453271703019e-05,
      "loss": 3.1348,
      "step": 4274
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.941643796042178e-05,
      "loss": 3.153,
      "step": 4275
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.938754875054167e-05,
      "loss": 3.3107,
      "step": 4276
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.935865954066157e-05,
      "loss": 2.9463,
      "step": 4277
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.932977033078146e-05,
      "loss": 3.2237,
      "step": 4278
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.930088112090136e-05,
      "loss": 3.2952,
      "step": 4279
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.927199191102124e-05,
      "loss": 3.112,
      "step": 4280
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.924310270114113e-05,
      "loss": 3.2988,
      "step": 4281
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.921421349126101e-05,
      "loss": 3.2512,
      "step": 4282
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.918532428138092e-05,
      "loss": 3.35,
      "step": 4283
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.91564350715008e-05,
      "loss": 3.2903,
      "step": 4284
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.912754586162069e-05,
      "loss": 3.2621,
      "step": 4285
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.909865665174057e-05,
      "loss": 3.3227,
      "step": 4286
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.906976744186047e-05,
      "loss": 3.1815,
      "step": 4287
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.904087823198036e-05,
      "loss": 3.2774,
      "step": 4288
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.901198902210025e-05,
      "loss": 3.2155,
      "step": 4289
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.898309981222013e-05,
      "loss": 3.3794,
      "step": 4290
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.895421060234003e-05,
      "loss": 3.351,
      "step": 4291
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.892532139245992e-05,
      "loss": 3.1811,
      "step": 4292
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.88964321825798e-05,
      "loss": 3.2645,
      "step": 4293
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.886754297269971e-05,
      "loss": 3.2618,
      "step": 4294
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.883865376281959e-05,
      "loss": 3.2929,
      "step": 4295
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.880976455293948e-05,
      "loss": 3.3232,
      "step": 4296
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.878087534305936e-05,
      "loss": 3.2274,
      "step": 4297
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.875198613317926e-05,
      "loss": 3.2881,
      "step": 4298
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.872309692329915e-05,
      "loss": 3.2647,
      "step": 4299
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.869420771341904e-05,
      "loss": 3.3605,
      "step": 4300
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.866531850353893e-05,
      "loss": 3.2639,
      "step": 4301
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.863642929365882e-05,
      "loss": 3.3084,
      "step": 4302
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.860754008377871e-05,
      "loss": 3.0701,
      "step": 4303
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.85786508738986e-05,
      "loss": 3.2976,
      "step": 4304
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.85497616640185e-05,
      "loss": 3.1141,
      "step": 4305
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.852087245413838e-05,
      "loss": 3.0212,
      "step": 4306
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.849198324425827e-05,
      "loss": 3.327,
      "step": 4307
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.846309403437815e-05,
      "loss": 3.2982,
      "step": 4308
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.843420482449806e-05,
      "loss": 3.1753,
      "step": 4309
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.840531561461794e-05,
      "loss": 3.0708,
      "step": 4310
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.837642640473784e-05,
      "loss": 3.3137,
      "step": 4311
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.834753719485772e-05,
      "loss": 3.1944,
      "step": 4312
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.831864798497761e-05,
      "loss": 3.2517,
      "step": 4313
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.82897587750975e-05,
      "loss": 3.2313,
      "step": 4314
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.82608695652174e-05,
      "loss": 3.1304,
      "step": 4315
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.823198035533729e-05,
      "loss": 3.2009,
      "step": 4316
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.820309114545717e-05,
      "loss": 3.3346,
      "step": 4317
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.817420193557707e-05,
      "loss": 3.2339,
      "step": 4318
    },
    {
      "epoch": 0.61,
      "learning_rate": 7.814531272569695e-05,
      "loss": 3.3033,
      "step": 4319
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.811642351581685e-05,
      "loss": 3.2246,
      "step": 4320
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.808753430593673e-05,
      "loss": 3.3153,
      "step": 4321
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.805864509605663e-05,
      "loss": 3.2447,
      "step": 4322
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.802975588617651e-05,
      "loss": 3.2197,
      "step": 4323
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.80008666762964e-05,
      "loss": 3.257,
      "step": 4324
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.79719774664163e-05,
      "loss": 3.1468,
      "step": 4325
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.794308825653619e-05,
      "loss": 3.2364,
      "step": 4326
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.791419904665608e-05,
      "loss": 3.3317,
      "step": 4327
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.788530983677596e-05,
      "loss": 3.3097,
      "step": 4328
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.785642062689586e-05,
      "loss": 3.265,
      "step": 4329
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.782753141701575e-05,
      "loss": 3.2576,
      "step": 4330
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.779864220713564e-05,
      "loss": 3.2598,
      "step": 4331
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.776975299725553e-05,
      "loss": 3.2786,
      "step": 4332
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.774086378737542e-05,
      "loss": 3.1964,
      "step": 4333
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.77119745774953e-05,
      "loss": 3.3129,
      "step": 4334
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.76830853676152e-05,
      "loss": 3.333,
      "step": 4335
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.765419615773509e-05,
      "loss": 3.2189,
      "step": 4336
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.762530694785498e-05,
      "loss": 3.2821,
      "step": 4337
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.759641773797487e-05,
      "loss": 3.2736,
      "step": 4338
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.756752852809475e-05,
      "loss": 3.1567,
      "step": 4339
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.753863931821465e-05,
      "loss": 3.1936,
      "step": 4340
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.750975010833454e-05,
      "loss": 3.2622,
      "step": 4341
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.748086089845444e-05,
      "loss": 3.3282,
      "step": 4342
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.745197168857432e-05,
      "loss": 3.2657,
      "step": 4343
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.742308247869421e-05,
      "loss": 3.1931,
      "step": 4344
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.739419326881409e-05,
      "loss": 3.3269,
      "step": 4345
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.7365304058934e-05,
      "loss": 3.2184,
      "step": 4346
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.733641484905388e-05,
      "loss": 3.1989,
      "step": 4347
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.730752563917377e-05,
      "loss": 3.1841,
      "step": 4348
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.727863642929367e-05,
      "loss": 3.155,
      "step": 4349
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.724974721941355e-05,
      "loss": 3.2665,
      "step": 4350
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.722085800953345e-05,
      "loss": 3.2832,
      "step": 4351
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.719196879965333e-05,
      "loss": 3.127,
      "step": 4352
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.716307958977323e-05,
      "loss": 3.2496,
      "step": 4353
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.713419037989311e-05,
      "loss": 3.0619,
      "step": 4354
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.7105301170013e-05,
      "loss": 3.2216,
      "step": 4355
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.70764119601329e-05,
      "loss": 3.1777,
      "step": 4356
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.704752275025279e-05,
      "loss": 3.2618,
      "step": 4357
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.701863354037267e-05,
      "loss": 3.328,
      "step": 4358
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.698974433049256e-05,
      "loss": 3.1858,
      "step": 4359
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.696085512061246e-05,
      "loss": 3.3507,
      "step": 4360
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.693196591073234e-05,
      "loss": 3.1983,
      "step": 4361
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.690307670085225e-05,
      "loss": 3.0417,
      "step": 4362
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.687418749097213e-05,
      "loss": 3.1231,
      "step": 4363
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.684529828109202e-05,
      "loss": 3.1928,
      "step": 4364
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.68164090712119e-05,
      "loss": 3.1603,
      "step": 4365
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.67875198613318e-05,
      "loss": 3.1971,
      "step": 4366
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.675863065145169e-05,
      "loss": 3.3221,
      "step": 4367
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.672974144157158e-05,
      "loss": 3.2545,
      "step": 4368
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.670085223169146e-05,
      "loss": 3.2316,
      "step": 4369
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.667196302181136e-05,
      "loss": 3.2825,
      "step": 4370
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.664307381193125e-05,
      "loss": 3.1697,
      "step": 4371
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.661418460205114e-05,
      "loss": 3.3155,
      "step": 4372
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.658529539217104e-05,
      "loss": 3.2699,
      "step": 4373
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.655640618229092e-05,
      "loss": 3.1791,
      "step": 4374
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.652751697241081e-05,
      "loss": 3.2443,
      "step": 4375
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.649862776253069e-05,
      "loss": 3.1897,
      "step": 4376
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.64697385526506e-05,
      "loss": 3.1354,
      "step": 4377
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.644084934277048e-05,
      "loss": 3.2865,
      "step": 4378
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.641196013289037e-05,
      "loss": 3.1777,
      "step": 4379
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.638307092301025e-05,
      "loss": 3.1771,
      "step": 4380
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.635418171313015e-05,
      "loss": 3.1502,
      "step": 4381
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.632529250325004e-05,
      "loss": 3.0661,
      "step": 4382
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.629640329336993e-05,
      "loss": 3.1649,
      "step": 4383
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.626751408348983e-05,
      "loss": 3.2292,
      "step": 4384
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.623862487360971e-05,
      "loss": 3.0547,
      "step": 4385
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.62097356637296e-05,
      "loss": 3.2867,
      "step": 4386
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.618084645384948e-05,
      "loss": 3.2806,
      "step": 4387
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.615195724396939e-05,
      "loss": 3.1973,
      "step": 4388
    },
    {
      "epoch": 0.62,
      "learning_rate": 7.612306803408927e-05,
      "loss": 3.1027,
      "step": 4389
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.609417882420916e-05,
      "loss": 3.3486,
      "step": 4390
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.606528961432904e-05,
      "loss": 3.2801,
      "step": 4391
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.603640040444894e-05,
      "loss": 3.1514,
      "step": 4392
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.600751119456883e-05,
      "loss": 3.1717,
      "step": 4393
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.597862198468873e-05,
      "loss": 3.2219,
      "step": 4394
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.594973277480862e-05,
      "loss": 3.2382,
      "step": 4395
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.59208435649285e-05,
      "loss": 3.0522,
      "step": 4396
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.58919543550484e-05,
      "loss": 3.1792,
      "step": 4397
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.586306514516829e-05,
      "loss": 3.2079,
      "step": 4398
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.583417593528818e-05,
      "loss": 3.2015,
      "step": 4399
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.580528672540806e-05,
      "loss": 3.1522,
      "step": 4400
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.577639751552796e-05,
      "loss": 3.2472,
      "step": 4401
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.574750830564784e-05,
      "loss": 3.2378,
      "step": 4402
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.571861909576773e-05,
      "loss": 3.2979,
      "step": 4403
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.568972988588762e-05,
      "loss": 3.1266,
      "step": 4404
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.566084067600752e-05,
      "loss": 2.9579,
      "step": 4405
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.563195146612741e-05,
      "loss": 3.2063,
      "step": 4406
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.560306225624729e-05,
      "loss": 3.2874,
      "step": 4407
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.557417304636718e-05,
      "loss": 3.2408,
      "step": 4408
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.554528383648708e-05,
      "loss": 3.2621,
      "step": 4409
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.551639462660697e-05,
      "loss": 3.0716,
      "step": 4410
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.548750541672685e-05,
      "loss": 3.2889,
      "step": 4411
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.545861620684675e-05,
      "loss": 3.2795,
      "step": 4412
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.542972699696663e-05,
      "loss": 3.088,
      "step": 4413
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.540083778708653e-05,
      "loss": 3.2983,
      "step": 4414
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.537194857720641e-05,
      "loss": 3.19,
      "step": 4415
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.534305936732631e-05,
      "loss": 3.1916,
      "step": 4416
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.53141701574462e-05,
      "loss": 3.298,
      "step": 4417
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.528528094756608e-05,
      "loss": 3.1589,
      "step": 4418
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.525639173768598e-05,
      "loss": 3.306,
      "step": 4419
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.522750252780587e-05,
      "loss": 3.2359,
      "step": 4420
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.519861331792576e-05,
      "loss": 3.278,
      "step": 4421
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.516972410804564e-05,
      "loss": 3.3765,
      "step": 4422
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.514083489816554e-05,
      "loss": 3.2689,
      "step": 4423
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.511194568828543e-05,
      "loss": 3.3587,
      "step": 4424
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.508305647840533e-05,
      "loss": 3.3364,
      "step": 4425
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.50541672685252e-05,
      "loss": 3.2131,
      "step": 4426
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.50252780586451e-05,
      "loss": 3.2932,
      "step": 4427
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.499638884876498e-05,
      "loss": 3.2665,
      "step": 4428
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.496749963888487e-05,
      "loss": 3.381,
      "step": 4429
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.493861042900477e-05,
      "loss": 3.2373,
      "step": 4430
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.490972121912466e-05,
      "loss": 3.1545,
      "step": 4431
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.488083200924456e-05,
      "loss": 3.2665,
      "step": 4432
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.485194279936444e-05,
      "loss": 3.3637,
      "step": 4433
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.482305358948433e-05,
      "loss": 3.3338,
      "step": 4434
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.479416437960422e-05,
      "loss": 3.117,
      "step": 4435
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.476527516972412e-05,
      "loss": 2.8719,
      "step": 4436
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.4736385959844e-05,
      "loss": 3.2194,
      "step": 4437
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.470749674996389e-05,
      "loss": 3.3096,
      "step": 4438
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.467860754008377e-05,
      "loss": 3.3919,
      "step": 4439
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.464971833020368e-05,
      "loss": 3.2348,
      "step": 4440
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.462082912032356e-05,
      "loss": 3.2321,
      "step": 4441
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.459193991044345e-05,
      "loss": 3.0917,
      "step": 4442
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.456305070056335e-05,
      "loss": 3.1843,
      "step": 4443
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.453416149068323e-05,
      "loss": 3.1699,
      "step": 4444
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.450527228080313e-05,
      "loss": 3.2327,
      "step": 4445
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.447638307092301e-05,
      "loss": 3.2365,
      "step": 4446
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.444749386104291e-05,
      "loss": 3.1183,
      "step": 4447
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.441860465116279e-05,
      "loss": 3.2592,
      "step": 4448
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.438971544128268e-05,
      "loss": 3.095,
      "step": 4449
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.436082623140256e-05,
      "loss": 3.2686,
      "step": 4450
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.433193702152247e-05,
      "loss": 3.2056,
      "step": 4451
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.430304781164235e-05,
      "loss": 3.2602,
      "step": 4452
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.427415860176224e-05,
      "loss": 3.1665,
      "step": 4453
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.424526939188214e-05,
      "loss": 3.0854,
      "step": 4454
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.421638018200202e-05,
      "loss": 3.0639,
      "step": 4455
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.418749097212193e-05,
      "loss": 3.1661,
      "step": 4456
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.41586017622418e-05,
      "loss": 3.1701,
      "step": 4457
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.41297125523617e-05,
      "loss": 3.1404,
      "step": 4458
    },
    {
      "epoch": 0.63,
      "learning_rate": 7.410082334248158e-05,
      "loss": 3.1746,
      "step": 4459
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.407193413260147e-05,
      "loss": 3.2634,
      "step": 4460
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.404304492272137e-05,
      "loss": 3.058,
      "step": 4461
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.401415571284126e-05,
      "loss": 3.1933,
      "step": 4462
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.398526650296114e-05,
      "loss": 3.2946,
      "step": 4463
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.395637729308104e-05,
      "loss": 3.3623,
      "step": 4464
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.392748808320093e-05,
      "loss": 3.1468,
      "step": 4465
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.389859887332082e-05,
      "loss": 3.2427,
      "step": 4466
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.386970966344072e-05,
      "loss": 3.2678,
      "step": 4467
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.38408204535606e-05,
      "loss": 3.2054,
      "step": 4468
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.381193124368049e-05,
      "loss": 3.1882,
      "step": 4469
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.378304203380037e-05,
      "loss": 3.1352,
      "step": 4470
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.375415282392027e-05,
      "loss": 3.2657,
      "step": 4471
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.372526361404016e-05,
      "loss": 3.2018,
      "step": 4472
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.369637440416005e-05,
      "loss": 3.1582,
      "step": 4473
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.366748519427993e-05,
      "loss": 3.2181,
      "step": 4474
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.363859598439983e-05,
      "loss": 3.1828,
      "step": 4475
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.360970677451972e-05,
      "loss": 3.3757,
      "step": 4476
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.358081756463961e-05,
      "loss": 3.2835,
      "step": 4477
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.355192835475951e-05,
      "loss": 3.2192,
      "step": 4478
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.352303914487939e-05,
      "loss": 3.1814,
      "step": 4479
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.349414993499928e-05,
      "loss": 3.0554,
      "step": 4480
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.346526072511916e-05,
      "loss": 3.3356,
      "step": 4481
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.343637151523907e-05,
      "loss": 3.2975,
      "step": 4482
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.340748230535895e-05,
      "loss": 3.392,
      "step": 4483
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.337859309547884e-05,
      "loss": 3.2069,
      "step": 4484
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.334970388559872e-05,
      "loss": 3.2134,
      "step": 4485
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.332081467571862e-05,
      "loss": 3.2042,
      "step": 4486
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.329192546583851e-05,
      "loss": 3.3986,
      "step": 4487
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.32630362559584e-05,
      "loss": 3.3561,
      "step": 4488
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.32341470460783e-05,
      "loss": 3.2048,
      "step": 4489
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.320525783619818e-05,
      "loss": 3.2647,
      "step": 4490
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.317636862631807e-05,
      "loss": 3.2076,
      "step": 4491
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.314747941643795e-05,
      "loss": 3.2791,
      "step": 4492
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.311859020655786e-05,
      "loss": 3.2383,
      "step": 4493
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.308970099667774e-05,
      "loss": 3.27,
      "step": 4494
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.306081178679764e-05,
      "loss": 3.1158,
      "step": 4495
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.303192257691752e-05,
      "loss": 3.2556,
      "step": 4496
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.300303336703741e-05,
      "loss": 3.281,
      "step": 4497
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.29741441571573e-05,
      "loss": 3.1698,
      "step": 4498
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.29452549472772e-05,
      "loss": 3.2967,
      "step": 4499
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.291636573739709e-05,
      "loss": 3.2783,
      "step": 4500
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.288747652751697e-05,
      "loss": 3.315,
      "step": 4501
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.285858731763687e-05,
      "loss": 3.2573,
      "step": 4502
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.282969810775676e-05,
      "loss": 3.1854,
      "step": 4503
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.280080889787665e-05,
      "loss": 3.235,
      "step": 4504
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.277191968799653e-05,
      "loss": 3.2233,
      "step": 4505
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.274303047811643e-05,
      "loss": 3.2246,
      "step": 4506
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.271414126823631e-05,
      "loss": 3.2204,
      "step": 4507
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.268525205835622e-05,
      "loss": 3.2522,
      "step": 4508
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.26563628484761e-05,
      "loss": 3.0833,
      "step": 4509
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.262747363859599e-05,
      "loss": 3.07,
      "step": 4510
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.259858442871588e-05,
      "loss": 3.1149,
      "step": 4511
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.256969521883576e-05,
      "loss": 3.2049,
      "step": 4512
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.254080600895567e-05,
      "loss": 3.3168,
      "step": 4513
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.251191679907555e-05,
      "loss": 3.2935,
      "step": 4514
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.248302758919544e-05,
      "loss": 3.1564,
      "step": 4515
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.245413837931533e-05,
      "loss": 3.2737,
      "step": 4516
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.242524916943522e-05,
      "loss": 3.1057,
      "step": 4517
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.23963599595551e-05,
      "loss": 3.2497,
      "step": 4518
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.2367470749675e-05,
      "loss": 3.2421,
      "step": 4519
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.233858153979489e-05,
      "loss": 3.2509,
      "step": 4520
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.230969232991478e-05,
      "loss": 3.2018,
      "step": 4521
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.228080312003467e-05,
      "loss": 3.1461,
      "step": 4522
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.225191391015455e-05,
      "loss": 3.1516,
      "step": 4523
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.222302470027445e-05,
      "loss": 3.2691,
      "step": 4524
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.219413549039434e-05,
      "loss": 3.4174,
      "step": 4525
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.216524628051424e-05,
      "loss": 3.2661,
      "step": 4526
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.213635707063412e-05,
      "loss": 3.1742,
      "step": 4527
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.210746786075401e-05,
      "loss": 3.2276,
      "step": 4528
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.20785786508739e-05,
      "loss": 3.2369,
      "step": 4529
    },
    {
      "epoch": 0.64,
      "learning_rate": 7.20496894409938e-05,
      "loss": 3.2119,
      "step": 4530
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.202080023111368e-05,
      "loss": 3.2396,
      "step": 4531
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.199191102123357e-05,
      "loss": 3.1502,
      "step": 4532
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.196302181135347e-05,
      "loss": 3.1763,
      "step": 4533
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.193413260147336e-05,
      "loss": 3.1303,
      "step": 4534
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.190524339159324e-05,
      "loss": 3.0942,
      "step": 4535
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.187635418171313e-05,
      "loss": 3.1065,
      "step": 4536
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.184746497183303e-05,
      "loss": 3.2077,
      "step": 4537
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.181857576195291e-05,
      "loss": 3.2625,
      "step": 4538
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.17896865520728e-05,
      "loss": 3.237,
      "step": 4539
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.17607973421927e-05,
      "loss": 3.3133,
      "step": 4540
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.173190813231259e-05,
      "loss": 3.2684,
      "step": 4541
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.170301892243247e-05,
      "loss": 3.2643,
      "step": 4542
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.167412971255236e-05,
      "loss": 3.2464,
      "step": 4543
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.164524050267226e-05,
      "loss": 3.2603,
      "step": 4544
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.161635129279215e-05,
      "loss": 3.1629,
      "step": 4545
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.158746208291203e-05,
      "loss": 3.2386,
      "step": 4546
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.155857287303193e-05,
      "loss": 3.322,
      "step": 4547
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.152968366315182e-05,
      "loss": 2.9695,
      "step": 4548
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.15007944532717e-05,
      "loss": 3.1846,
      "step": 4549
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.14719052433916e-05,
      "loss": 3.2918,
      "step": 4550
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.144301603351149e-05,
      "loss": 3.2516,
      "step": 4551
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.141412682363138e-05,
      "loss": 3.2238,
      "step": 4552
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.138523761375126e-05,
      "loss": 3.2354,
      "step": 4553
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.135634840387115e-05,
      "loss": 3.1473,
      "step": 4554
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.132745919399105e-05,
      "loss": 3.3592,
      "step": 4555
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.129856998411094e-05,
      "loss": 3.2755,
      "step": 4556
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.126968077423082e-05,
      "loss": 3.1127,
      "step": 4557
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.124079156435072e-05,
      "loss": 3.1792,
      "step": 4558
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.121190235447061e-05,
      "loss": 3.3286,
      "step": 4559
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.118301314459049e-05,
      "loss": 3.2957,
      "step": 4560
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.11541239347104e-05,
      "loss": 3.025,
      "step": 4561
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.112523472483028e-05,
      "loss": 3.189,
      "step": 4562
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.109634551495017e-05,
      "loss": 3.1633,
      "step": 4563
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.106745630507005e-05,
      "loss": 3.0469,
      "step": 4564
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.103856709518995e-05,
      "loss": 3.3885,
      "step": 4565
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.100967788530984e-05,
      "loss": 3.293,
      "step": 4566
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.098078867542973e-05,
      "loss": 3.1842,
      "step": 4567
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.095189946554961e-05,
      "loss": 3.2436,
      "step": 4568
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.092301025566951e-05,
      "loss": 3.2228,
      "step": 4569
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.08941210457894e-05,
      "loss": 3.3918,
      "step": 4570
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.08652318359093e-05,
      "loss": 3.0825,
      "step": 4571
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.083634262602919e-05,
      "loss": 3.2249,
      "step": 4572
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.080745341614907e-05,
      "loss": 3.0199,
      "step": 4573
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.077856420626896e-05,
      "loss": 3.2107,
      "step": 4574
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.074967499638884e-05,
      "loss": 3.1974,
      "step": 4575
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.072078578650875e-05,
      "loss": 3.2753,
      "step": 4576
    },
    {
      "epoch": 0.65,
      "eval_loss": 3.4311070442199707,
      "eval_runtime": 471.4356,
      "eval_samples_per_second": 43.457,
      "eval_steps_per_second": 14.486,
      "step": 4576
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.069189657662863e-05,
      "loss": 3.0329,
      "step": 4577
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.066300736674853e-05,
      "loss": 3.3126,
      "step": 4578
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.06341181568684e-05,
      "loss": 3.1901,
      "step": 4579
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.06052289469883e-05,
      "loss": 3.163,
      "step": 4580
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.05763397371082e-05,
      "loss": 3.1173,
      "step": 4581
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.054745052722809e-05,
      "loss": 3.244,
      "step": 4582
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.051856131734798e-05,
      "loss": 3.2602,
      "step": 4583
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.048967210746786e-05,
      "loss": 2.991,
      "step": 4584
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.046078289758776e-05,
      "loss": 3.2214,
      "step": 4585
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.043189368770764e-05,
      "loss": 3.2427,
      "step": 4586
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.040300447782754e-05,
      "loss": 3.2009,
      "step": 4587
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.037411526794742e-05,
      "loss": 3.191,
      "step": 4588
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.034522605806732e-05,
      "loss": 3.2651,
      "step": 4589
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.03163368481872e-05,
      "loss": 3.3027,
      "step": 4590
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.028744763830709e-05,
      "loss": 3.2212,
      "step": 4591
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.025855842842698e-05,
      "loss": 3.2133,
      "step": 4592
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.022966921854688e-05,
      "loss": 3.1425,
      "step": 4593
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.020078000866677e-05,
      "loss": 3.2199,
      "step": 4594
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.017189079878665e-05,
      "loss": 3.2051,
      "step": 4595
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.014300158890655e-05,
      "loss": 3.1071,
      "step": 4596
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.011411237902644e-05,
      "loss": 3.2108,
      "step": 4597
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.008522316914633e-05,
      "loss": 3.1735,
      "step": 4598
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.005633395926621e-05,
      "loss": 3.1795,
      "step": 4599
    },
    {
      "epoch": 0.65,
      "learning_rate": 7.002744474938611e-05,
      "loss": 3.1864,
      "step": 4600
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.999855553950599e-05,
      "loss": 3.3092,
      "step": 4601
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.99696663296259e-05,
      "loss": 3.1383,
      "step": 4602
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.994077711974578e-05,
      "loss": 3.1475,
      "step": 4603
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.991188790986567e-05,
      "loss": 3.2344,
      "step": 4604
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.988299869998556e-05,
      "loss": 3.1652,
      "step": 4605
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.985410949010544e-05,
      "loss": 3.2111,
      "step": 4606
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.982522028022534e-05,
      "loss": 3.224,
      "step": 4607
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.979633107034523e-05,
      "loss": 3.1232,
      "step": 4608
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.976744186046513e-05,
      "loss": 3.2263,
      "step": 4609
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.9738552650585e-05,
      "loss": 3.1037,
      "step": 4610
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.97096634407049e-05,
      "loss": 3.2389,
      "step": 4611
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.968077423082478e-05,
      "loss": 3.185,
      "step": 4612
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.965188502094469e-05,
      "loss": 3.2433,
      "step": 4613
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.962299581106457e-05,
      "loss": 3.2258,
      "step": 4614
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.959410660118446e-05,
      "loss": 3.1148,
      "step": 4615
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.956521739130436e-05,
      "loss": 3.2667,
      "step": 4616
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.953632818142424e-05,
      "loss": 3.1186,
      "step": 4617
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.950743897154414e-05,
      "loss": 3.085,
      "step": 4618
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.947854976166402e-05,
      "loss": 3.1934,
      "step": 4619
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.944966055178392e-05,
      "loss": 3.0886,
      "step": 4620
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.94207713419038e-05,
      "loss": 3.1788,
      "step": 4621
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.939188213202369e-05,
      "loss": 3.2761,
      "step": 4622
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.936299292214358e-05,
      "loss": 3.237,
      "step": 4623
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.933410371226348e-05,
      "loss": 3.1859,
      "step": 4624
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.930521450238336e-05,
      "loss": 3.1281,
      "step": 4625
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.927632529250325e-05,
      "loss": 3.1964,
      "step": 4626
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.924743608262315e-05,
      "loss": 3.3033,
      "step": 4627
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.921854687274303e-05,
      "loss": 3.1437,
      "step": 4628
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.918965766286293e-05,
      "loss": 3.1905,
      "step": 4629
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.916076845298281e-05,
      "loss": 3.2168,
      "step": 4630
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.913187924310271e-05,
      "loss": 3.0697,
      "step": 4631
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.910299003322259e-05,
      "loss": 3.2662,
      "step": 4632
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.907410082334248e-05,
      "loss": 3.0232,
      "step": 4633
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.904521161346238e-05,
      "loss": 3.0126,
      "step": 4634
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.901632240358227e-05,
      "loss": 3.2893,
      "step": 4635
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.898743319370215e-05,
      "loss": 3.0733,
      "step": 4636
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.895854398382204e-05,
      "loss": 3.175,
      "step": 4637
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.892965477394194e-05,
      "loss": 3.0857,
      "step": 4638
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.890076556406183e-05,
      "loss": 3.2243,
      "step": 4639
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.887187635418173e-05,
      "loss": 3.2138,
      "step": 4640
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.88429871443016e-05,
      "loss": 3.2269,
      "step": 4641
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.88140979344215e-05,
      "loss": 3.1455,
      "step": 4642
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.878520872454138e-05,
      "loss": 3.24,
      "step": 4643
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.875631951466129e-05,
      "loss": 3.0995,
      "step": 4644
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.872743030478117e-05,
      "loss": 3.1545,
      "step": 4645
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.869854109490106e-05,
      "loss": 3.2984,
      "step": 4646
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.866965188502094e-05,
      "loss": 3.1757,
      "step": 4647
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.864076267514084e-05,
      "loss": 3.2271,
      "step": 4648
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.861187346526073e-05,
      "loss": 3.2733,
      "step": 4649
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.858298425538062e-05,
      "loss": 3.1991,
      "step": 4650
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.855409504550052e-05,
      "loss": 3.2941,
      "step": 4651
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.85252058356204e-05,
      "loss": 3.1293,
      "step": 4652
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.849631662574029e-05,
      "loss": 3.1675,
      "step": 4653
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.846742741586017e-05,
      "loss": 3.1558,
      "step": 4654
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.843853820598008e-05,
      "loss": 3.1124,
      "step": 4655
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.840964899609996e-05,
      "loss": 2.9734,
      "step": 4656
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.838075978621985e-05,
      "loss": 3.2422,
      "step": 4657
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.835187057633973e-05,
      "loss": 3.0933,
      "step": 4658
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.832298136645963e-05,
      "loss": 3.1473,
      "step": 4659
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.829409215657952e-05,
      "loss": 3.221,
      "step": 4660
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.826520294669941e-05,
      "loss": 3.2445,
      "step": 4661
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.82363137368193e-05,
      "loss": 3.1867,
      "step": 4662
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.820742452693919e-05,
      "loss": 3.2059,
      "step": 4663
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.817853531705908e-05,
      "loss": 3.1624,
      "step": 4664
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.814964610717898e-05,
      "loss": 3.0782,
      "step": 4665
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.812075689729887e-05,
      "loss": 3.2645,
      "step": 4666
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.809186768741875e-05,
      "loss": 3.1695,
      "step": 4667
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.806297847753864e-05,
      "loss": 3.2342,
      "step": 4668
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.803408926765852e-05,
      "loss": 3.159,
      "step": 4669
    },
    {
      "epoch": 0.66,
      "learning_rate": 6.800520005777842e-05,
      "loss": 3.1207,
      "step": 4670
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.797631084789831e-05,
      "loss": 3.1898,
      "step": 4671
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.79474216380182e-05,
      "loss": 3.2039,
      "step": 4672
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.791853242813809e-05,
      "loss": 3.0811,
      "step": 4673
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.788964321825798e-05,
      "loss": 3.2256,
      "step": 4674
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.786075400837787e-05,
      "loss": 3.1116,
      "step": 4675
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.783186479849777e-05,
      "loss": 3.0967,
      "step": 4676
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.780297558861766e-05,
      "loss": 3.2738,
      "step": 4677
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.777408637873754e-05,
      "loss": 3.2311,
      "step": 4678
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.774519716885744e-05,
      "loss": 3.3502,
      "step": 4679
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.771630795897732e-05,
      "loss": 3.2233,
      "step": 4680
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.768741874909722e-05,
      "loss": 3.283,
      "step": 4681
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.76585295392171e-05,
      "loss": 3.2997,
      "step": 4682
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.7629640329337e-05,
      "loss": 3.1227,
      "step": 4683
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.760075111945688e-05,
      "loss": 3.256,
      "step": 4684
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.757186190957677e-05,
      "loss": 3.1657,
      "step": 4685
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.754297269969667e-05,
      "loss": 3.2956,
      "step": 4686
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.751408348981656e-05,
      "loss": 3.1841,
      "step": 4687
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.748519427993645e-05,
      "loss": 3.2518,
      "step": 4688
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.745630507005633e-05,
      "loss": 3.0815,
      "step": 4689
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.742741586017623e-05,
      "loss": 3.1521,
      "step": 4690
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.739852665029611e-05,
      "loss": 3.2677,
      "step": 4691
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.736963744041601e-05,
      "loss": 3.1445,
      "step": 4692
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.73407482305359e-05,
      "loss": 3.217,
      "step": 4693
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.731185902065579e-05,
      "loss": 3.1792,
      "step": 4694
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.728296981077567e-05,
      "loss": 2.9674,
      "step": 4695
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.725408060089556e-05,
      "loss": 3.1301,
      "step": 4696
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.722519139101546e-05,
      "loss": 3.2711,
      "step": 4697
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.719630218113535e-05,
      "loss": 3.3011,
      "step": 4698
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.716741297125524e-05,
      "loss": 3.1289,
      "step": 4699
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.713852376137512e-05,
      "loss": 3.322,
      "step": 4700
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.710963455149502e-05,
      "loss": 3.2275,
      "step": 4701
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.708074534161491e-05,
      "loss": 3.2683,
      "step": 4702
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.70518561317348e-05,
      "loss": 3.3123,
      "step": 4703
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.702296692185469e-05,
      "loss": 3.1394,
      "step": 4704
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.699407771197458e-05,
      "loss": 3.1832,
      "step": 4705
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.696518850209446e-05,
      "loss": 3.2946,
      "step": 4706
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.693629929221437e-05,
      "loss": 3.1702,
      "step": 4707
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.690741008233425e-05,
      "loss": 3.2187,
      "step": 4708
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.687852087245414e-05,
      "loss": 3.2411,
      "step": 4709
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.684963166257404e-05,
      "loss": 3.2604,
      "step": 4710
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.682074245269392e-05,
      "loss": 3.261,
      "step": 4711
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.679185324281382e-05,
      "loss": 3.3329,
      "step": 4712
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.67629640329337e-05,
      "loss": 3.1801,
      "step": 4713
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.67340748230536e-05,
      "loss": 3.0755,
      "step": 4714
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.670518561317348e-05,
      "loss": 3.2324,
      "step": 4715
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.667629640329337e-05,
      "loss": 3.2004,
      "step": 4716
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.664740719341325e-05,
      "loss": 3.0358,
      "step": 4717
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.661851798353316e-05,
      "loss": 3.093,
      "step": 4718
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.658962877365304e-05,
      "loss": 3.2078,
      "step": 4719
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.656073956377293e-05,
      "loss": 3.3535,
      "step": 4720
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.653185035389283e-05,
      "loss": 3.2595,
      "step": 4721
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.650296114401271e-05,
      "loss": 3.1418,
      "step": 4722
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.647407193413262e-05,
      "loss": 2.9937,
      "step": 4723
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.64451827242525e-05,
      "loss": 3.2423,
      "step": 4724
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.641629351437239e-05,
      "loss": 3.3218,
      "step": 4725
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.638740430449227e-05,
      "loss": 3.1807,
      "step": 4726
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.635851509461216e-05,
      "loss": 3.1824,
      "step": 4727
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.632962588473206e-05,
      "loss": 3.1208,
      "step": 4728
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.630073667485195e-05,
      "loss": 3.2159,
      "step": 4729
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.627184746497183e-05,
      "loss": 3.1804,
      "step": 4730
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.624295825509173e-05,
      "loss": 3.1377,
      "step": 4731
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.621406904521162e-05,
      "loss": 3.2806,
      "step": 4732
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.618517983533151e-05,
      "loss": 3.2248,
      "step": 4733
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.61562906254514e-05,
      "loss": 3.1957,
      "step": 4734
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.612740141557129e-05,
      "loss": 3.1038,
      "step": 4735
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.609851220569118e-05,
      "loss": 3.2546,
      "step": 4736
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.606962299581106e-05,
      "loss": 3.1971,
      "step": 4737
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.604073378593095e-05,
      "loss": 3.2071,
      "step": 4738
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.601184457605085e-05,
      "loss": 3.3514,
      "step": 4739
    },
    {
      "epoch": 0.67,
      "learning_rate": 6.598295536617074e-05,
      "loss": 3.2246,
      "step": 4740
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.595406615629062e-05,
      "loss": 3.0967,
      "step": 4741
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.592517694641052e-05,
      "loss": 3.1957,
      "step": 4742
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.589628773653041e-05,
      "loss": 3.242,
      "step": 4743
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.58673985266503e-05,
      "loss": 3.1726,
      "step": 4744
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.58385093167702e-05,
      "loss": 3.1347,
      "step": 4745
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.580962010689008e-05,
      "loss": 3.0802,
      "step": 4746
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.578073089700997e-05,
      "loss": 3.2203,
      "step": 4747
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.575184168712985e-05,
      "loss": 3.27,
      "step": 4748
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.572295247724976e-05,
      "loss": 3.3669,
      "step": 4749
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.569406326736964e-05,
      "loss": 3.0405,
      "step": 4750
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.566517405748953e-05,
      "loss": 3.1361,
      "step": 4751
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.563628484760941e-05,
      "loss": 3.2701,
      "step": 4752
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.560739563772931e-05,
      "loss": 3.2333,
      "step": 4753
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.55785064278492e-05,
      "loss": 3.1564,
      "step": 4754
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.55496172179691e-05,
      "loss": 3.0823,
      "step": 4755
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.552072800808899e-05,
      "loss": 3.1326,
      "step": 4756
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.549183879820887e-05,
      "loss": 3.0293,
      "step": 4757
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.546294958832876e-05,
      "loss": 3.1572,
      "step": 4758
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.543406037844864e-05,
      "loss": 3.1817,
      "step": 4759
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.540517116856855e-05,
      "loss": 3.2333,
      "step": 4760
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.537628195868843e-05,
      "loss": 3.2094,
      "step": 4761
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.534739274880833e-05,
      "loss": 3.1592,
      "step": 4762
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.53185035389282e-05,
      "loss": 3.084,
      "step": 4763
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.52896143290481e-05,
      "loss": 3.2225,
      "step": 4764
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.526072511916799e-05,
      "loss": 3.0779,
      "step": 4765
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.523183590928789e-05,
      "loss": 3.1048,
      "step": 4766
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.520294669940778e-05,
      "loss": 3.2311,
      "step": 4767
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.517405748952766e-05,
      "loss": 3.187,
      "step": 4768
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.514516827964755e-05,
      "loss": 3.1343,
      "step": 4769
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.511627906976745e-05,
      "loss": 3.1523,
      "step": 4770
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.508738985988734e-05,
      "loss": 3.0041,
      "step": 4771
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.505850065000722e-05,
      "loss": 3.1949,
      "step": 4772
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.502961144012712e-05,
      "loss": 3.1414,
      "step": 4773
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.5000722230247e-05,
      "loss": 3.0958,
      "step": 4774
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.49718330203669e-05,
      "loss": 3.2459,
      "step": 4775
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.494294381048678e-05,
      "loss": 3.1617,
      "step": 4776
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.491405460060668e-05,
      "loss": 3.1454,
      "step": 4777
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.488516539072657e-05,
      "loss": 3.1938,
      "step": 4778
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.485627618084645e-05,
      "loss": 3.3808,
      "step": 4779
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.482738697096635e-05,
      "loss": 3.0696,
      "step": 4780
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.479849776108624e-05,
      "loss": 3.2487,
      "step": 4781
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.476960855120613e-05,
      "loss": 3.3306,
      "step": 4782
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.474071934132601e-05,
      "loss": 3.1838,
      "step": 4783
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.471183013144591e-05,
      "loss": 3.1287,
      "step": 4784
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.468294092156579e-05,
      "loss": 3.1744,
      "step": 4785
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.46540517116857e-05,
      "loss": 3.2149,
      "step": 4786
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.462516250180558e-05,
      "loss": 3.0864,
      "step": 4787
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.459627329192547e-05,
      "loss": 3.2708,
      "step": 4788
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.456738408204535e-05,
      "loss": 3.0833,
      "step": 4789
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.453849487216524e-05,
      "loss": 3.1123,
      "step": 4790
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.450960566228514e-05,
      "loss": 3.1003,
      "step": 4791
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.448071645240503e-05,
      "loss": 3.1161,
      "step": 4792
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.445182724252493e-05,
      "loss": 3.1413,
      "step": 4793
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.44229380326448e-05,
      "loss": 3.1732,
      "step": 4794
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.43940488227647e-05,
      "loss": 3.1983,
      "step": 4795
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.43651596128846e-05,
      "loss": 3.1057,
      "step": 4796
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.433627040300449e-05,
      "loss": 3.2081,
      "step": 4797
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.430738119312437e-05,
      "loss": 3.2113,
      "step": 4798
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.427849198324426e-05,
      "loss": 3.3364,
      "step": 4799
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.424960277336414e-05,
      "loss": 3.321,
      "step": 4800
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.422071356348405e-05,
      "loss": 3.0925,
      "step": 4801
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.419182435360393e-05,
      "loss": 3.1668,
      "step": 4802
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.416293514372382e-05,
      "loss": 3.1238,
      "step": 4803
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.413404593384372e-05,
      "loss": 3.1487,
      "step": 4804
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.41051567239636e-05,
      "loss": 2.9601,
      "step": 4805
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.407626751408349e-05,
      "loss": 3.0914,
      "step": 4806
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.404737830420338e-05,
      "loss": 3.3274,
      "step": 4807
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.401848909432328e-05,
      "loss": 3.3154,
      "step": 4808
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.398959988444316e-05,
      "loss": 3.0696,
      "step": 4809
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.396071067456305e-05,
      "loss": 3.0216,
      "step": 4810
    },
    {
      "epoch": 0.68,
      "learning_rate": 6.393182146468293e-05,
      "loss": 3.2504,
      "step": 4811
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.390293225480284e-05,
      "loss": 3.142,
      "step": 4812
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.387404304492272e-05,
      "loss": 3.1038,
      "step": 4813
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.384515383504261e-05,
      "loss": 3.1674,
      "step": 4814
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.381626462516251e-05,
      "loss": 3.2564,
      "step": 4815
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.378737541528239e-05,
      "loss": 3.144,
      "step": 4816
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.37584862054023e-05,
      "loss": 3.2916,
      "step": 4817
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.372959699552218e-05,
      "loss": 3.3567,
      "step": 4818
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.370070778564207e-05,
      "loss": 3.1593,
      "step": 4819
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.367181857576195e-05,
      "loss": 3.218,
      "step": 4820
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.364292936588184e-05,
      "loss": 3.347,
      "step": 4821
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.361404015600174e-05,
      "loss": 3.205,
      "step": 4822
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.358515094612163e-05,
      "loss": 3.1968,
      "step": 4823
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.355626173624151e-05,
      "loss": 3.1964,
      "step": 4824
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.35273725263614e-05,
      "loss": 3.2347,
      "step": 4825
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.34984833164813e-05,
      "loss": 3.3071,
      "step": 4826
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.346959410660118e-05,
      "loss": 3.2471,
      "step": 4827
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.344070489672109e-05,
      "loss": 3.2021,
      "step": 4828
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.341181568684097e-05,
      "loss": 3.1955,
      "step": 4829
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.338292647696086e-05,
      "loss": 3.0071,
      "step": 4830
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.335403726708074e-05,
      "loss": 3.2419,
      "step": 4831
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.332514805720064e-05,
      "loss": 3.0888,
      "step": 4832
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.329625884732053e-05,
      "loss": 3.1945,
      "step": 4833
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.326736963744042e-05,
      "loss": 3.0264,
      "step": 4834
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.32384804275603e-05,
      "loss": 3.1711,
      "step": 4835
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.32095912176802e-05,
      "loss": 3.1518,
      "step": 4836
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.318070200780009e-05,
      "loss": 3.0763,
      "step": 4837
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.315181279791998e-05,
      "loss": 3.1948,
      "step": 4838
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.312292358803988e-05,
      "loss": 3.2302,
      "step": 4839
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.309403437815976e-05,
      "loss": 3.1315,
      "step": 4840
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.306514516827965e-05,
      "loss": 3.0718,
      "step": 4841
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.303625595839953e-05,
      "loss": 3.2503,
      "step": 4842
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.300736674851944e-05,
      "loss": 3.1226,
      "step": 4843
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.297847753863932e-05,
      "loss": 3.2664,
      "step": 4844
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.294958832875921e-05,
      "loss": 2.9803,
      "step": 4845
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.29206991188791e-05,
      "loss": 3.2725,
      "step": 4846
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.289180990899899e-05,
      "loss": 3.1647,
      "step": 4847
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.286292069911888e-05,
      "loss": 3.164,
      "step": 4848
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.283403148923878e-05,
      "loss": 3.2146,
      "step": 4849
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.280514227935867e-05,
      "loss": 3.2211,
      "step": 4850
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.277625306947855e-05,
      "loss": 3.0367,
      "step": 4851
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.274736385959844e-05,
      "loss": 3.1545,
      "step": 4852
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.271847464971832e-05,
      "loss": 3.185,
      "step": 4853
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.268958543983823e-05,
      "loss": 3.0758,
      "step": 4854
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.266069622995811e-05,
      "loss": 3.1946,
      "step": 4855
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.2631807020078e-05,
      "loss": 3.0128,
      "step": 4856
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.260291781019789e-05,
      "loss": 3.2722,
      "step": 4857
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.257402860031778e-05,
      "loss": 3.0703,
      "step": 4858
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.254513939043767e-05,
      "loss": 3.3005,
      "step": 4859
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.251625018055757e-05,
      "loss": 3.2755,
      "step": 4860
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.248736097067746e-05,
      "loss": 3.1733,
      "step": 4861
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.245847176079734e-05,
      "loss": 3.0191,
      "step": 4862
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.242958255091724e-05,
      "loss": 3.1909,
      "step": 4863
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.240069334103713e-05,
      "loss": 3.344,
      "step": 4864
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.237180413115702e-05,
      "loss": 3.2325,
      "step": 4865
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.23429149212769e-05,
      "loss": 3.1482,
      "step": 4866
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.23140257113968e-05,
      "loss": 3.2666,
      "step": 4867
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.228513650151668e-05,
      "loss": 3.1017,
      "step": 4868
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.225624729163657e-05,
      "loss": 3.0212,
      "step": 4869
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.222735808175647e-05,
      "loss": 3.141,
      "step": 4870
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.219846887187636e-05,
      "loss": 3.1939,
      "step": 4871
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.216957966199625e-05,
      "loss": 3.1901,
      "step": 4872
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.214069045211613e-05,
      "loss": 2.9811,
      "step": 4873
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.211180124223603e-05,
      "loss": 3.1416,
      "step": 4874
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.208291203235592e-05,
      "loss": 3.2035,
      "step": 4875
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.205402282247581e-05,
      "loss": 3.3768,
      "step": 4876
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.20251336125957e-05,
      "loss": 3.1864,
      "step": 4877
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.199624440271559e-05,
      "loss": 3.2768,
      "step": 4878
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.196735519283547e-05,
      "loss": 3.2096,
      "step": 4879
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.193846598295538e-05,
      "loss": 3.2389,
      "step": 4880
    },
    {
      "epoch": 0.69,
      "learning_rate": 6.190957677307526e-05,
      "loss": 3.1871,
      "step": 4881
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.188068756319515e-05,
      "loss": 3.1195,
      "step": 4882
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.185179835331504e-05,
      "loss": 3.1172,
      "step": 4883
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.182290914343492e-05,
      "loss": 3.2083,
      "step": 4884
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.179401993355482e-05,
      "loss": 3.2845,
      "step": 4885
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.176513072367471e-05,
      "loss": 3.2313,
      "step": 4886
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.17362415137946e-05,
      "loss": 3.2976,
      "step": 4887
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.170735230391449e-05,
      "loss": 3.2979,
      "step": 4888
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.167846309403438e-05,
      "loss": 3.2395,
      "step": 4889
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.164957388415426e-05,
      "loss": 3.2283,
      "step": 4890
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.162068467427417e-05,
      "loss": 3.1972,
      "step": 4891
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.159179546439405e-05,
      "loss": 3.2107,
      "step": 4892
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.156290625451394e-05,
      "loss": 3.1887,
      "step": 4893
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.153401704463384e-05,
      "loss": 3.0887,
      "step": 4894
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.150512783475372e-05,
      "loss": 3.2432,
      "step": 4895
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.147623862487361e-05,
      "loss": 3.0324,
      "step": 4896
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.14473494149935e-05,
      "loss": 3.1536,
      "step": 4897
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.14184602051134e-05,
      "loss": 3.0445,
      "step": 4898
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.138957099523328e-05,
      "loss": 3.224,
      "step": 4899
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.136068178535317e-05,
      "loss": 3.297,
      "step": 4900
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.133179257547307e-05,
      "loss": 3.2922,
      "step": 4901
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.130290336559296e-05,
      "loss": 3.2604,
      "step": 4902
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.127401415571284e-05,
      "loss": 3.1391,
      "step": 4903
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.124512494583273e-05,
      "loss": 3.2075,
      "step": 4904
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.121623573595263e-05,
      "loss": 3.2132,
      "step": 4905
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.118734652607252e-05,
      "loss": 3.0788,
      "step": 4906
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.11584573161924e-05,
      "loss": 3.1895,
      "step": 4907
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.11295681063123e-05,
      "loss": 3.2684,
      "step": 4908
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.110067889643219e-05,
      "loss": 3.1677,
      "step": 4909
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.107178968655207e-05,
      "loss": 3.1967,
      "step": 4910
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.104290047667198e-05,
      "loss": 3.1881,
      "step": 4911
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.101401126679186e-05,
      "loss": 3.2648,
      "step": 4912
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.098512205691175e-05,
      "loss": 3.2341,
      "step": 4913
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.095623284703164e-05,
      "loss": 3.1041,
      "step": 4914
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0927343637151525e-05,
      "loss": 3.1151,
      "step": 4915
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.089845442727141e-05,
      "loss": 3.178,
      "step": 4916
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.086956521739131e-05,
      "loss": 3.0908,
      "step": 4917
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.08406760075112e-05,
      "loss": 3.1385,
      "step": 4918
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.081178679763109e-05,
      "loss": 3.1282,
      "step": 4919
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0782897587750974e-05,
      "loss": 3.2666,
      "step": 4920
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.075400837787086e-05,
      "loss": 3.1628,
      "step": 4921
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.072511916799076e-05,
      "loss": 3.1016,
      "step": 4922
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.069622995811065e-05,
      "loss": 3.1877,
      "step": 4923
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.066734074823054e-05,
      "loss": 3.241,
      "step": 4924
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.063845153835043e-05,
      "loss": 3.2473,
      "step": 4925
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0609562328470316e-05,
      "loss": 3.3245,
      "step": 4926
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.058067311859022e-05,
      "loss": 3.0849,
      "step": 4927
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0551783908710104e-05,
      "loss": 3.1867,
      "step": 4928
    },
    {
      "epoch": 0.7,
      "eval_loss": 3.405430316925049,
      "eval_runtime": 471.5274,
      "eval_samples_per_second": 43.448,
      "eval_steps_per_second": 14.483,
      "step": 4928
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.052289469882999e-05,
      "loss": 3.2446,
      "step": 4929
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.049400548894988e-05,
      "loss": 3.0791,
      "step": 4930
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0465116279069765e-05,
      "loss": 3.1938,
      "step": 4931
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0436227069189666e-05,
      "loss": 3.1848,
      "step": 4932
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.040733785930955e-05,
      "loss": 3.236,
      "step": 4933
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.037844864942944e-05,
      "loss": 3.0853,
      "step": 4934
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.034955943954933e-05,
      "loss": 3.2123,
      "step": 4935
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.032067022966922e-05,
      "loss": 3.2857,
      "step": 4936
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.029178101978911e-05,
      "loss": 3.1235,
      "step": 4937
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.026289180990901e-05,
      "loss": 2.9182,
      "step": 4938
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0234002600028895e-05,
      "loss": 3.1968,
      "step": 4939
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.020511339014878e-05,
      "loss": 3.1346,
      "step": 4940
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.017622418026867e-05,
      "loss": 3.0872,
      "step": 4941
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0147334970388557e-05,
      "loss": 3.1449,
      "step": 4942
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.011844576050846e-05,
      "loss": 3.015,
      "step": 4943
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.0089556550628344e-05,
      "loss": 3.2121,
      "step": 4944
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.006066734074823e-05,
      "loss": 3.1967,
      "step": 4945
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.003177813086812e-05,
      "loss": 3.0941,
      "step": 4946
    },
    {
      "epoch": 0.7,
      "learning_rate": 6.000288892098801e-05,
      "loss": 3.1809,
      "step": 4947
    },
    {
      "epoch": 0.7,
      "learning_rate": 5.9973999711107906e-05,
      "loss": 2.9571,
      "step": 4948
    },
    {
      "epoch": 0.7,
      "learning_rate": 5.99451105012278e-05,
      "loss": 3.1754,
      "step": 4949
    },
    {
      "epoch": 0.7,
      "learning_rate": 5.991622129134769e-05,
      "loss": 3.143,
      "step": 4950
    },
    {
      "epoch": 0.7,
      "learning_rate": 5.9887332081467574e-05,
      "loss": 3.0815,
      "step": 4951
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.985844287158746e-05,
      "loss": 3.2158,
      "step": 4952
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.982955366170736e-05,
      "loss": 3.1671,
      "step": 4953
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.980066445182725e-05,
      "loss": 3.2071,
      "step": 4954
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.9771775241947136e-05,
      "loss": 3.1664,
      "step": 4955
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.974288603206702e-05,
      "loss": 3.1808,
      "step": 4956
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.971399682218691e-05,
      "loss": 3.1447,
      "step": 4957
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.9685107612306804e-05,
      "loss": 3.2338,
      "step": 4958
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.96562184024267e-05,
      "loss": 3.191,
      "step": 4959
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.962732919254659e-05,
      "loss": 3.1847,
      "step": 4960
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.959843998266648e-05,
      "loss": 3.018,
      "step": 4961
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.9569550772786365e-05,
      "loss": 3.0103,
      "step": 4962
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.954066156290625e-05,
      "loss": 3.1028,
      "step": 4963
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.951177235302615e-05,
      "loss": 3.2361,
      "step": 4964
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.948288314314604e-05,
      "loss": 3.2101,
      "step": 4965
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.945399393326593e-05,
      "loss": 3.3008,
      "step": 4966
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.9425104723385814e-05,
      "loss": 3.0315,
      "step": 4967
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.93962155135057e-05,
      "loss": 3.2395,
      "step": 4968
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.93673263036256e-05,
      "loss": 3.1923,
      "step": 4969
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.933843709374549e-05,
      "loss": 3.185,
      "step": 4970
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.930954788386538e-05,
      "loss": 2.9979,
      "step": 4971
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.928065867398527e-05,
      "loss": 3.169,
      "step": 4972
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.925176946410516e-05,
      "loss": 3.1904,
      "step": 4973
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.922288025422506e-05,
      "loss": 3.1637,
      "step": 4974
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.9193991044344944e-05,
      "loss": 2.9797,
      "step": 4975
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.916510183446483e-05,
      "loss": 3.1532,
      "step": 4976
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.913621262458472e-05,
      "loss": 3.1193,
      "step": 4977
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.9107323414704606e-05,
      "loss": 3.1299,
      "step": 4978
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.907843420482449e-05,
      "loss": 3.0867,
      "step": 4979
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.904954499494439e-05,
      "loss": 3.1889,
      "step": 4980
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.902065578506428e-05,
      "loss": 3.192,
      "step": 4981
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8991766575184174e-05,
      "loss": 3.2168,
      "step": 4982
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.896287736530406e-05,
      "loss": 3.2185,
      "step": 4983
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.893398815542395e-05,
      "loss": 3.176,
      "step": 4984
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.890509894554385e-05,
      "loss": 3.1294,
      "step": 4985
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8876209735663736e-05,
      "loss": 3.1345,
      "step": 4986
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.884732052578362e-05,
      "loss": 3.3109,
      "step": 4987
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.881843131590351e-05,
      "loss": 3.0086,
      "step": 4988
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.87895421060234e-05,
      "loss": 3.2247,
      "step": 4989
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.87606528961433e-05,
      "loss": 3.1655,
      "step": 4990
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8731763686263185e-05,
      "loss": 3.1094,
      "step": 4991
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.870287447638307e-05,
      "loss": 3.1659,
      "step": 4992
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8673985266502965e-05,
      "loss": 3.1532,
      "step": 4993
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.864509605662285e-05,
      "loss": 3.2451,
      "step": 4994
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8616206846742746e-05,
      "loss": 3.1007,
      "step": 4995
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.858731763686264e-05,
      "loss": 3.1713,
      "step": 4996
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.855842842698253e-05,
      "loss": 3.1186,
      "step": 4997
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8529539217102414e-05,
      "loss": 3.1076,
      "step": 4998
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.85006500072223e-05,
      "loss": 3.2508,
      "step": 4999
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.84717607973422e-05,
      "loss": 3.0121,
      "step": 5000
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.844287158746209e-05,
      "loss": 3.1222,
      "step": 5001
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8413982377581976e-05,
      "loss": 3.19,
      "step": 5002
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.838509316770186e-05,
      "loss": 3.1305,
      "step": 5003
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.835620395782175e-05,
      "loss": 3.1201,
      "step": 5004
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8327314747941644e-05,
      "loss": 3.2659,
      "step": 5005
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.829842553806154e-05,
      "loss": 3.1431,
      "step": 5006
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.826953632818143e-05,
      "loss": 3.2168,
      "step": 5007
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.824064711830132e-05,
      "loss": 3.1828,
      "step": 5008
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8211757908421206e-05,
      "loss": 3.1976,
      "step": 5009
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.818286869854109e-05,
      "loss": 3.0626,
      "step": 5010
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.815397948866099e-05,
      "loss": 3.1023,
      "step": 5011
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.812509027878088e-05,
      "loss": 3.1348,
      "step": 5012
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.809620106890077e-05,
      "loss": 3.2077,
      "step": 5013
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.8067311859020655e-05,
      "loss": 3.312,
      "step": 5014
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.803842264914054e-05,
      "loss": 3.3103,
      "step": 5015
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.800953343926044e-05,
      "loss": 3.1398,
      "step": 5016
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.798064422938033e-05,
      "loss": 3.1509,
      "step": 5017
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.795175501950022e-05,
      "loss": 3.2007,
      "step": 5018
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.792286580962011e-05,
      "loss": 3.2066,
      "step": 5019
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.789397659974e-05,
      "loss": 3.1849,
      "step": 5020
    },
    {
      "epoch": 0.71,
      "learning_rate": 5.78650873898599e-05,
      "loss": 3.2473,
      "step": 5021
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7836198179979785e-05,
      "loss": 3.1957,
      "step": 5022
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.780730897009967e-05,
      "loss": 3.1588,
      "step": 5023
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.777841976021956e-05,
      "loss": 3.1026,
      "step": 5024
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7749530550339446e-05,
      "loss": 3.0227,
      "step": 5025
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.772064134045933e-05,
      "loss": 3.1725,
      "step": 5026
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7691752130579234e-05,
      "loss": 3.3349,
      "step": 5027
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.766286292069912e-05,
      "loss": 3.179,
      "step": 5028
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7633973710819014e-05,
      "loss": 3.2301,
      "step": 5029
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.76050845009389e-05,
      "loss": 3.1841,
      "step": 5030
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.757619529105879e-05,
      "loss": 3.1111,
      "step": 5031
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.754730608117869e-05,
      "loss": 3.1174,
      "step": 5032
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7518416871298576e-05,
      "loss": 3.1633,
      "step": 5033
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.748952766141846e-05,
      "loss": 2.9734,
      "step": 5034
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.746063845153835e-05,
      "loss": 3.1083,
      "step": 5035
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.743174924165824e-05,
      "loss": 3.211,
      "step": 5036
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.740286003177814e-05,
      "loss": 3.1702,
      "step": 5037
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7373970821898025e-05,
      "loss": 3.2022,
      "step": 5038
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.734508161201791e-05,
      "loss": 3.1054,
      "step": 5039
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7316192402137806e-05,
      "loss": 3.1327,
      "step": 5040
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.728730319225769e-05,
      "loss": 3.0072,
      "step": 5041
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.725841398237759e-05,
      "loss": 3.1727,
      "step": 5042
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.722952477249748e-05,
      "loss": 3.247,
      "step": 5043
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.720063556261737e-05,
      "loss": 3.1123,
      "step": 5044
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7171746352737255e-05,
      "loss": 3.1567,
      "step": 5045
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.714285714285714e-05,
      "loss": 3.1468,
      "step": 5046
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.711396793297703e-05,
      "loss": 3.1196,
      "step": 5047
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.708507872309693e-05,
      "loss": 3.193,
      "step": 5048
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7056189513216816e-05,
      "loss": 3.0999,
      "step": 5049
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.7027300303336703e-05,
      "loss": 3.2171,
      "step": 5050
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.69984110934566e-05,
      "loss": 3.2542,
      "step": 5051
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6969521883576484e-05,
      "loss": 3.2912,
      "step": 5052
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.694063267369638e-05,
      "loss": 3.1588,
      "step": 5053
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.691174346381627e-05,
      "loss": 3.2516,
      "step": 5054
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.688285425393616e-05,
      "loss": 3.1515,
      "step": 5055
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6853965044056046e-05,
      "loss": 3.2265,
      "step": 5056
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.682507583417593e-05,
      "loss": 3.1729,
      "step": 5057
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6796186624295834e-05,
      "loss": 3.2611,
      "step": 5058
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.676729741441572e-05,
      "loss": 3.2122,
      "step": 5059
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.673840820453561e-05,
      "loss": 3.3132,
      "step": 5060
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6709518994655495e-05,
      "loss": 3.195,
      "step": 5061
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.668062978477538e-05,
      "loss": 3.208,
      "step": 5062
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.665174057489528e-05,
      "loss": 3.2916,
      "step": 5063
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.662285136501517e-05,
      "loss": 3.1674,
      "step": 5064
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6593962155135063e-05,
      "loss": 3.1717,
      "step": 5065
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.656507294525495e-05,
      "loss": 3.1739,
      "step": 5066
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.653618373537484e-05,
      "loss": 3.0539,
      "step": 5067
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6507294525494725e-05,
      "loss": 3.1643,
      "step": 5068
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6478405315614625e-05,
      "loss": 3.1571,
      "step": 5069
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.644951610573451e-05,
      "loss": 3.1703,
      "step": 5070
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.64206268958544e-05,
      "loss": 3.1374,
      "step": 5071
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6391737685974286e-05,
      "loss": 3.1227,
      "step": 5072
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.636284847609417e-05,
      "loss": 3.1776,
      "step": 5073
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6333959266214074e-05,
      "loss": 3.2057,
      "step": 5074
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.630507005633396e-05,
      "loss": 3.2371,
      "step": 5075
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6276180846453855e-05,
      "loss": 3.2429,
      "step": 5076
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.624729163657374e-05,
      "loss": 3.1389,
      "step": 5077
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.621840242669363e-05,
      "loss": 3.2158,
      "step": 5078
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.618951321681353e-05,
      "loss": 3.2406,
      "step": 5079
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6160624006933417e-05,
      "loss": 2.9815,
      "step": 5080
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6131734797053304e-05,
      "loss": 3.1771,
      "step": 5081
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.610284558717319e-05,
      "loss": 3.0717,
      "step": 5082
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.607395637729308e-05,
      "loss": 3.1711,
      "step": 5083
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.604506716741298e-05,
      "loss": 3.1081,
      "step": 5084
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.6016177957532865e-05,
      "loss": 3.1248,
      "step": 5085
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.598728874765275e-05,
      "loss": 3.0333,
      "step": 5086
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.5958399537772646e-05,
      "loss": 3.2139,
      "step": 5087
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.592951032789253e-05,
      "loss": 3.2611,
      "step": 5088
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.590062111801242e-05,
      "loss": 3.1232,
      "step": 5089
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.587173190813232e-05,
      "loss": 3.109,
      "step": 5090
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.584284269825221e-05,
      "loss": 3.1966,
      "step": 5091
    },
    {
      "epoch": 0.72,
      "learning_rate": 5.5813953488372095e-05,
      "loss": 3.1803,
      "step": 5092
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.578506427849198e-05,
      "loss": 3.0442,
      "step": 5093
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.575617506861187e-05,
      "loss": 3.1491,
      "step": 5094
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.572728585873177e-05,
      "loss": 3.3084,
      "step": 5095
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.569839664885166e-05,
      "loss": 3.2134,
      "step": 5096
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5669507438971544e-05,
      "loss": 3.219,
      "step": 5097
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.564061822909144e-05,
      "loss": 2.9762,
      "step": 5098
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5611729019211325e-05,
      "loss": 2.9179,
      "step": 5099
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.558283980933122e-05,
      "loss": 3.2013,
      "step": 5100
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.555395059945111e-05,
      "loss": 2.9588,
      "step": 5101
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5525061389571e-05,
      "loss": 3.0973,
      "step": 5102
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5496172179690886e-05,
      "loss": 3.1533,
      "step": 5103
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5467282969810774e-05,
      "loss": 3.1741,
      "step": 5104
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5438393759930674e-05,
      "loss": 3.0876,
      "step": 5105
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.540950455005056e-05,
      "loss": 3.2014,
      "step": 5106
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.538061534017045e-05,
      "loss": 3.1576,
      "step": 5107
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5351726130290335e-05,
      "loss": 3.1465,
      "step": 5108
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.532283692041023e-05,
      "loss": 3.2683,
      "step": 5109
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.529394771053012e-05,
      "loss": 3.266,
      "step": 5110
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.526505850065001e-05,
      "loss": 3.1291,
      "step": 5111
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5236169290769904e-05,
      "loss": 3.2675,
      "step": 5112
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.520728008088979e-05,
      "loss": 3.2307,
      "step": 5113
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.517839087100968e-05,
      "loss": 2.9991,
      "step": 5114
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5149501661129565e-05,
      "loss": 3.1703,
      "step": 5115
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.5120612451249466e-05,
      "loss": 3.059,
      "step": 5116
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.509172324136935e-05,
      "loss": 3.1885,
      "step": 5117
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.506283403148924e-05,
      "loss": 3.2284,
      "step": 5118
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.503394482160913e-05,
      "loss": 3.1545,
      "step": 5119
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.500505561172902e-05,
      "loss": 3.1512,
      "step": 5120
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4976166401848914e-05,
      "loss": 3.2021,
      "step": 5121
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.49472771919688e-05,
      "loss": 3.1952,
      "step": 5122
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4918387982088695e-05,
      "loss": 3.1533,
      "step": 5123
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.488949877220858e-05,
      "loss": 3.2113,
      "step": 5124
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.486060956232847e-05,
      "loss": 3.1937,
      "step": 5125
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.483172035244837e-05,
      "loss": 3.1545,
      "step": 5126
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.480283114256826e-05,
      "loss": 3.2336,
      "step": 5127
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4773941932688144e-05,
      "loss": 3.1997,
      "step": 5128
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.474505272280803e-05,
      "loss": 3.0685,
      "step": 5129
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.471616351292792e-05,
      "loss": 3.0653,
      "step": 5130
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.468727430304782e-05,
      "loss": 3.2056,
      "step": 5131
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4658385093167706e-05,
      "loss": 3.2199,
      "step": 5132
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.462949588328759e-05,
      "loss": 3.2525,
      "step": 5133
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.460060667340749e-05,
      "loss": 3.2594,
      "step": 5134
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4571717463527374e-05,
      "loss": 3.2341,
      "step": 5135
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.454282825364726e-05,
      "loss": 3.2136,
      "step": 5136
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.451393904376716e-05,
      "loss": 3.1541,
      "step": 5137
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.448504983388705e-05,
      "loss": 3.1913,
      "step": 5138
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4456160624006935e-05,
      "loss": 3.0299,
      "step": 5139
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.442727141412682e-05,
      "loss": 3.1656,
      "step": 5140
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.439838220424671e-05,
      "loss": 3.237,
      "step": 5141
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.436949299436661e-05,
      "loss": 3.2268,
      "step": 5142
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.43406037844865e-05,
      "loss": 3.1631,
      "step": 5143
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4311714574606384e-05,
      "loss": 3.1169,
      "step": 5144
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.428282536472628e-05,
      "loss": 3.0628,
      "step": 5145
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4253936154846165e-05,
      "loss": 3.2385,
      "step": 5146
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4225046944966066e-05,
      "loss": 3.1183,
      "step": 5147
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.419615773508595e-05,
      "loss": 3.0587,
      "step": 5148
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.416726852520584e-05,
      "loss": 3.1037,
      "step": 5149
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.413837931532573e-05,
      "loss": 3.0909,
      "step": 5150
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4109490105445614e-05,
      "loss": 2.9987,
      "step": 5151
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.4080600895565515e-05,
      "loss": 3.2461,
      "step": 5152
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.40517116856854e-05,
      "loss": 3.1846,
      "step": 5153
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.402282247580529e-05,
      "loss": 3.1396,
      "step": 5154
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.3993933265925176e-05,
      "loss": 3.2256,
      "step": 5155
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.396504405604507e-05,
      "loss": 3.1149,
      "step": 5156
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.3936154846164957e-05,
      "loss": 3.1215,
      "step": 5157
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.390726563628486e-05,
      "loss": 3.1063,
      "step": 5158
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.3878376426404744e-05,
      "loss": 3.1478,
      "step": 5159
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.384948721652463e-05,
      "loss": 3.1244,
      "step": 5160
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.382059800664452e-05,
      "loss": 3.1537,
      "step": 5161
    },
    {
      "epoch": 0.73,
      "learning_rate": 5.3791708796764405e-05,
      "loss": 3.1722,
      "step": 5162
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3762819586884306e-05,
      "loss": 3.1656,
      "step": 5163
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.373393037700419e-05,
      "loss": 3.2235,
      "step": 5164
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.370504116712408e-05,
      "loss": 3.2493,
      "step": 5165
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.367615195724397e-05,
      "loss": 3.0734,
      "step": 5166
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.364726274736386e-05,
      "loss": 3.2295,
      "step": 5167
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3618373537483755e-05,
      "loss": 3.1616,
      "step": 5168
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.358948432760364e-05,
      "loss": 3.2689,
      "step": 5169
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3560595117723536e-05,
      "loss": 3.252,
      "step": 5170
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.353170590784342e-05,
      "loss": 3.1139,
      "step": 5171
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.350281669796331e-05,
      "loss": 3.2654,
      "step": 5172
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.347392748808321e-05,
      "loss": 3.2094,
      "step": 5173
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.34450382782031e-05,
      "loss": 3.0202,
      "step": 5174
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3416149068322984e-05,
      "loss": 3.2164,
      "step": 5175
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.338725985844287e-05,
      "loss": 3.143,
      "step": 5176
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.335837064856276e-05,
      "loss": 3.2617,
      "step": 5177
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.332948143868265e-05,
      "loss": 3.0766,
      "step": 5178
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3300592228802546e-05,
      "loss": 3.2323,
      "step": 5179
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.327170301892243e-05,
      "loss": 3.1394,
      "step": 5180
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.324281380904233e-05,
      "loss": 3.1591,
      "step": 5181
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3213924599162214e-05,
      "loss": 3.0493,
      "step": 5182
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.31850353892821e-05,
      "loss": 3.2181,
      "step": 5183
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3156146179402e-05,
      "loss": 3.153,
      "step": 5184
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.312725696952189e-05,
      "loss": 3.163,
      "step": 5185
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.3098367759641776e-05,
      "loss": 3.1937,
      "step": 5186
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.306947854976166e-05,
      "loss": 3.1086,
      "step": 5187
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.304058933988155e-05,
      "loss": 3.1899,
      "step": 5188
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.301170013000145e-05,
      "loss": 2.8025,
      "step": 5189
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.298281092012134e-05,
      "loss": 3.2051,
      "step": 5190
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2953921710241225e-05,
      "loss": 3.1696,
      "step": 5191
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.292503250036112e-05,
      "loss": 2.8641,
      "step": 5192
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2896143290481006e-05,
      "loss": 3.0356,
      "step": 5193
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2867254080600906e-05,
      "loss": 2.9993,
      "step": 5194
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.283836487072079e-05,
      "loss": 3.1589,
      "step": 5195
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.280947566084068e-05,
      "loss": 2.8759,
      "step": 5196
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.278058645096057e-05,
      "loss": 3.2366,
      "step": 5197
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2751697241080454e-05,
      "loss": 3.2196,
      "step": 5198
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2722808031200355e-05,
      "loss": 3.126,
      "step": 5199
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.269391882132024e-05,
      "loss": 3.0893,
      "step": 5200
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.266502961144013e-05,
      "loss": 3.2265,
      "step": 5201
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2636140401560016e-05,
      "loss": 3.0793,
      "step": 5202
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.260725119167991e-05,
      "loss": 3.1028,
      "step": 5203
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.25783619817998e-05,
      "loss": 3.12,
      "step": 5204
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.25494727719197e-05,
      "loss": 3.0613,
      "step": 5205
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2520583562039585e-05,
      "loss": 3.0577,
      "step": 5206
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.249169435215947e-05,
      "loss": 3.079,
      "step": 5207
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.246280514227936e-05,
      "loss": 3.2207,
      "step": 5208
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2433915932399246e-05,
      "loss": 3.1076,
      "step": 5209
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2405026722519146e-05,
      "loss": 3.1991,
      "step": 5210
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2376137512639033e-05,
      "loss": 3.2086,
      "step": 5211
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.234724830275892e-05,
      "loss": 2.8687,
      "step": 5212
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.231835909287881e-05,
      "loss": 3.1321,
      "step": 5213
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.22894698829987e-05,
      "loss": 3.2268,
      "step": 5214
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2260580673118595e-05,
      "loss": 3.2203,
      "step": 5215
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.223169146323849e-05,
      "loss": 3.1536,
      "step": 5216
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2202802253358376e-05,
      "loss": 3.2663,
      "step": 5217
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.217391304347826e-05,
      "loss": 3.2019,
      "step": 5218
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.214502383359815e-05,
      "loss": 3.0641,
      "step": 5219
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.211613462371805e-05,
      "loss": 3.0838,
      "step": 5220
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.208724541383794e-05,
      "loss": 3.1684,
      "step": 5221
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.2058356203957825e-05,
      "loss": 3.149,
      "step": 5222
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.202946699407771e-05,
      "loss": 3.0806,
      "step": 5223
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.20005777841976e-05,
      "loss": 3.2016,
      "step": 5224
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.197168857431749e-05,
      "loss": 3.1248,
      "step": 5225
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.1942799364437387e-05,
      "loss": 3.2296,
      "step": 5226
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.191391015455728e-05,
      "loss": 3.1643,
      "step": 5227
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.188502094467717e-05,
      "loss": 3.119,
      "step": 5228
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.1856131734797054e-05,
      "loss": 3.0145,
      "step": 5229
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.182724252491694e-05,
      "loss": 3.1054,
      "step": 5230
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.179835331503684e-05,
      "loss": 2.9761,
      "step": 5231
    },
    {
      "epoch": 0.74,
      "learning_rate": 5.176946410515673e-05,
      "loss": 3.1658,
      "step": 5232
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1740574895276616e-05,
      "loss": 3.1663,
      "step": 5233
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.17116856853965e-05,
      "loss": 3.1418,
      "step": 5234
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.168279647551639e-05,
      "loss": 3.1757,
      "step": 5235
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.165390726563629e-05,
      "loss": 3.2377,
      "step": 5236
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.162501805575618e-05,
      "loss": 3.1921,
      "step": 5237
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1596128845876065e-05,
      "loss": 3.0378,
      "step": 5238
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.156723963599596e-05,
      "loss": 3.1379,
      "step": 5239
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1538350426115846e-05,
      "loss": 3.1644,
      "step": 5240
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1509461216235747e-05,
      "loss": 3.212,
      "step": 5241
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1480572006355634e-05,
      "loss": 2.8968,
      "step": 5242
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.145168279647552e-05,
      "loss": 3.2158,
      "step": 5243
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.142279358659541e-05,
      "loss": 2.9899,
      "step": 5244
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1393904376715295e-05,
      "loss": 2.9403,
      "step": 5245
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.136501516683518e-05,
      "loss": 3.2206,
      "step": 5246
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.133612595695508e-05,
      "loss": 3.1934,
      "step": 5247
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.130723674707497e-05,
      "loss": 3.1608,
      "step": 5248
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1278347537194856e-05,
      "loss": 3.0498,
      "step": 5249
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.124945832731475e-05,
      "loss": 3.1471,
      "step": 5250
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.122056911743464e-05,
      "loss": 3.0589,
      "step": 5251
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.119167990755454e-05,
      "loss": 3.2602,
      "step": 5252
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1162790697674425e-05,
      "loss": 3.211,
      "step": 5253
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.113390148779431e-05,
      "loss": 3.0975,
      "step": 5254
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.11050122779142e-05,
      "loss": 3.1378,
      "step": 5255
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1076123068034086e-05,
      "loss": 3.0698,
      "step": 5256
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.104723385815399e-05,
      "loss": 3.2347,
      "step": 5257
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.1018344648273874e-05,
      "loss": 3.2977,
      "step": 5258
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.098945543839376e-05,
      "loss": 3.1856,
      "step": 5259
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.096056622851365e-05,
      "loss": 3.0422,
      "step": 5260
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.093167701863354e-05,
      "loss": 3.1303,
      "step": 5261
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0902787808753436e-05,
      "loss": 3.1463,
      "step": 5262
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.087389859887333e-05,
      "loss": 3.1041,
      "step": 5263
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0845009388993216e-05,
      "loss": 3.029,
      "step": 5264
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0816120179113103e-05,
      "loss": 3.1371,
      "step": 5265
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.078723096923299e-05,
      "loss": 3.002,
      "step": 5266
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.075834175935288e-05,
      "loss": 3.1604,
      "step": 5267
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.072945254947278e-05,
      "loss": 3.1258,
      "step": 5268
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0700563339592665e-05,
      "loss": 3.0734,
      "step": 5269
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.067167412971255e-05,
      "loss": 3.2201,
      "step": 5270
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.064278491983244e-05,
      "loss": 3.1679,
      "step": 5271
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.061389570995233e-05,
      "loss": 3.1449,
      "step": 5272
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.058500650007223e-05,
      "loss": 3.2038,
      "step": 5273
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.055611729019212e-05,
      "loss": 3.28,
      "step": 5274
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.052722808031201e-05,
      "loss": 3.1376,
      "step": 5275
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0498338870431895e-05,
      "loss": 3.0618,
      "step": 5276
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.046944966055178e-05,
      "loss": 3.1943,
      "step": 5277
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.044056045067168e-05,
      "loss": 3.162,
      "step": 5278
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.041167124079157e-05,
      "loss": 3.1661,
      "step": 5279
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.038278203091146e-05,
      "loss": 3.2009,
      "step": 5280
    },
    {
      "epoch": 0.75,
      "eval_loss": 3.3782970905303955,
      "eval_runtime": 472.5543,
      "eval_samples_per_second": 43.354,
      "eval_steps_per_second": 14.451,
      "step": 5280
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0353892821031344e-05,
      "loss": 2.8974,
      "step": 5281
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.032500361115123e-05,
      "loss": 3.1747,
      "step": 5282
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.029611440127113e-05,
      "loss": 3.2308,
      "step": 5283
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.026722519139102e-05,
      "loss": 3.224,
      "step": 5284
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.023833598151091e-05,
      "loss": 3.0434,
      "step": 5285
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.02094467716308e-05,
      "loss": 3.265,
      "step": 5286
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0180557561750686e-05,
      "loss": 3.176,
      "step": 5287
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.015166835187057e-05,
      "loss": 3.1993,
      "step": 5288
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0122779141990474e-05,
      "loss": 3.1893,
      "step": 5289
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.009388993211036e-05,
      "loss": 3.1407,
      "step": 5290
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.006500072223025e-05,
      "loss": 3.1743,
      "step": 5291
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.0036111512350135e-05,
      "loss": 3.0759,
      "step": 5292
    },
    {
      "epoch": 0.75,
      "learning_rate": 5.000722230247002e-05,
      "loss": 3.1738,
      "step": 5293
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.9978333092589916e-05,
      "loss": 3.016,
      "step": 5294
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.994944388270981e-05,
      "loss": 3.1617,
      "step": 5295
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.99205546728297e-05,
      "loss": 3.0395,
      "step": 5296
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.989166546294959e-05,
      "loss": 3.0232,
      "step": 5297
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.9862776253069485e-05,
      "loss": 3.1016,
      "step": 5298
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.983388704318937e-05,
      "loss": 3.1152,
      "step": 5299
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.9804997833309265e-05,
      "loss": 3.0738,
      "step": 5300
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.977610862342915e-05,
      "loss": 3.1971,
      "step": 5301
    },
    {
      "epoch": 0.75,
      "learning_rate": 4.974721941354904e-05,
      "loss": 3.1543,
      "step": 5302
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.971833020366893e-05,
      "loss": 3.2111,
      "step": 5303
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.968944099378882e-05,
      "loss": 3.056,
      "step": 5304
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.966055178390871e-05,
      "loss": 3.0425,
      "step": 5305
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.96316625740286e-05,
      "loss": 3.1703,
      "step": 5306
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.960277336414849e-05,
      "loss": 3.1008,
      "step": 5307
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.957388415426838e-05,
      "loss": 3.2465,
      "step": 5308
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.9544994944388276e-05,
      "loss": 3.1674,
      "step": 5309
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.951610573450816e-05,
      "loss": 3.0894,
      "step": 5310
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.948721652462806e-05,
      "loss": 3.2196,
      "step": 5311
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.9458327314747944e-05,
      "loss": 3.2325,
      "step": 5312
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.942943810486784e-05,
      "loss": 3.2285,
      "step": 5313
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.9400548894987725e-05,
      "loss": 3.1028,
      "step": 5314
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.937165968510761e-05,
      "loss": 3.1758,
      "step": 5315
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.9342770475227506e-05,
      "loss": 3.1665,
      "step": 5316
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.931388126534739e-05,
      "loss": 3.0878,
      "step": 5317
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.928499205546728e-05,
      "loss": 3.2068,
      "step": 5318
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.9256102845587174e-05,
      "loss": 3.0782,
      "step": 5319
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.922721363570707e-05,
      "loss": 3.2516,
      "step": 5320
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.919832442582696e-05,
      "loss": 3.1966,
      "step": 5321
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.916943521594685e-05,
      "loss": 3.1208,
      "step": 5322
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.9140546006066735e-05,
      "loss": 2.9893,
      "step": 5323
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.911165679618663e-05,
      "loss": 2.9886,
      "step": 5324
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.9082767586306516e-05,
      "loss": 3.2782,
      "step": 5325
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.905387837642641e-05,
      "loss": 2.8643,
      "step": 5326
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.90249891665463e-05,
      "loss": 3.2197,
      "step": 5327
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8996099956666184e-05,
      "loss": 3.1602,
      "step": 5328
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.896721074678608e-05,
      "loss": 3.2031,
      "step": 5329
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8938321536905965e-05,
      "loss": 3.152,
      "step": 5330
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.890943232702586e-05,
      "loss": 3.1258,
      "step": 5331
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.888054311714575e-05,
      "loss": 3.1495,
      "step": 5332
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.885165390726564e-05,
      "loss": 3.0992,
      "step": 5333
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8822764697385533e-05,
      "loss": 3.0927,
      "step": 5334
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.879387548750542e-05,
      "loss": 3.1619,
      "step": 5335
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.876498627762531e-05,
      "loss": 3.1023,
      "step": 5336
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.87360970677452e-05,
      "loss": 3.1128,
      "step": 5337
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.870720785786509e-05,
      "loss": 3.104,
      "step": 5338
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8678318647984976e-05,
      "loss": 2.8925,
      "step": 5339
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.864942943810487e-05,
      "loss": 3.1436,
      "step": 5340
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8620540228224756e-05,
      "loss": 3.2427,
      "step": 5341
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.859165101834465e-05,
      "loss": 3.2214,
      "step": 5342
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8562761808464544e-05,
      "loss": 3.1747,
      "step": 5343
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.853387259858443e-05,
      "loss": 3.2018,
      "step": 5344
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8504983388704325e-05,
      "loss": 3.1354,
      "step": 5345
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.847609417882421e-05,
      "loss": 3.0884,
      "step": 5346
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8447204968944106e-05,
      "loss": 3.1204,
      "step": 5347
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.841831575906399e-05,
      "loss": 3.0706,
      "step": 5348
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.838942654918388e-05,
      "loss": 3.154,
      "step": 5349
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8360537339303774e-05,
      "loss": 3.1685,
      "step": 5350
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.833164812942366e-05,
      "loss": 3.2801,
      "step": 5351
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.830275891954355e-05,
      "loss": 3.0305,
      "step": 5352
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.827386970966344e-05,
      "loss": 2.9677,
      "step": 5353
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8244980499783335e-05,
      "loss": 3.1084,
      "step": 5354
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.821609128990322e-05,
      "loss": 2.9526,
      "step": 5355
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8187202080023116e-05,
      "loss": 3.0777,
      "step": 5356
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8158312870143e-05,
      "loss": 3.1239,
      "step": 5357
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.81294236602629e-05,
      "loss": 3.1601,
      "step": 5358
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8100534450382784e-05,
      "loss": 3.2667,
      "step": 5359
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.807164524050267e-05,
      "loss": 3.0284,
      "step": 5360
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.8042756030622565e-05,
      "loss": 3.1923,
      "step": 5361
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.801386682074245e-05,
      "loss": 3.1982,
      "step": 5362
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.7984977610862346e-05,
      "loss": 3.2425,
      "step": 5363
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.795608840098223e-05,
      "loss": 3.1461,
      "step": 5364
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.792719919110212e-05,
      "loss": 3.1147,
      "step": 5365
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.7898309981222014e-05,
      "loss": 3.2444,
      "step": 5366
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.786942077134191e-05,
      "loss": 3.2765,
      "step": 5367
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.78405315614618e-05,
      "loss": 3.2053,
      "step": 5368
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.781164235158169e-05,
      "loss": 3.1532,
      "step": 5369
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.7782753141701576e-05,
      "loss": 3.1784,
      "step": 5370
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.775386393182147e-05,
      "loss": 3.062,
      "step": 5371
    },
    {
      "epoch": 0.76,
      "learning_rate": 4.7724974721941357e-05,
      "loss": 3.0357,
      "step": 5372
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7696085512061244e-05,
      "loss": 3.2979,
      "step": 5373
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.766719630218114e-05,
      "loss": 3.0243,
      "step": 5374
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7638307092301024e-05,
      "loss": 3.2944,
      "step": 5375
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.760941788242092e-05,
      "loss": 3.2016,
      "step": 5376
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7580528672540805e-05,
      "loss": 3.1149,
      "step": 5377
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.75516394626607e-05,
      "loss": 3.2387,
      "step": 5378
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.752275025278059e-05,
      "loss": 3.1331,
      "step": 5379
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.749386104290048e-05,
      "loss": 3.1386,
      "step": 5380
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7464971833020374e-05,
      "loss": 3.1489,
      "step": 5381
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.743608262314026e-05,
      "loss": 2.882,
      "step": 5382
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.740719341326015e-05,
      "loss": 3.078,
      "step": 5383
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.737830420338004e-05,
      "loss": 3.0792,
      "step": 5384
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.734941499349993e-05,
      "loss": 3.0568,
      "step": 5385
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7320525783619816e-05,
      "loss": 3.1153,
      "step": 5386
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.729163657373971e-05,
      "loss": 3.2472,
      "step": 5387
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.72627473638596e-05,
      "loss": 3.1513,
      "step": 5388
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.723385815397949e-05,
      "loss": 3.273,
      "step": 5389
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7204968944099384e-05,
      "loss": 3.1339,
      "step": 5390
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.717607973421927e-05,
      "loss": 3.2117,
      "step": 5391
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7147190524339165e-05,
      "loss": 3.0859,
      "step": 5392
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.711830131445905e-05,
      "loss": 3.0746,
      "step": 5393
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.708941210457894e-05,
      "loss": 3.1658,
      "step": 5394
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.706052289469883e-05,
      "loss": 3.1445,
      "step": 5395
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.703163368481872e-05,
      "loss": 3.0719,
      "step": 5396
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.7002744474938614e-05,
      "loss": 3.182,
      "step": 5397
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.69738552650585e-05,
      "loss": 3.207,
      "step": 5398
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.694496605517839e-05,
      "loss": 3.1019,
      "step": 5399
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.691607684529828e-05,
      "loss": 3.1441,
      "step": 5400
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6887187635418176e-05,
      "loss": 3.1312,
      "step": 5401
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.685829842553807e-05,
      "loss": 3.0884,
      "step": 5402
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.682940921565796e-05,
      "loss": 3.0809,
      "step": 5403
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6800520005777844e-05,
      "loss": 2.9426,
      "step": 5404
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.677163079589774e-05,
      "loss": 3.1012,
      "step": 5405
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6742741586017625e-05,
      "loss": 3.2305,
      "step": 5406
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.671385237613751e-05,
      "loss": 3.169,
      "step": 5407
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6684963166257406e-05,
      "loss": 3.237,
      "step": 5408
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.665607395637729e-05,
      "loss": 3.2113,
      "step": 5409
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6627184746497186e-05,
      "loss": 3.3055,
      "step": 5410
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6598295536617073e-05,
      "loss": 2.98,
      "step": 5411
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.656940632673697e-05,
      "loss": 3.0582,
      "step": 5412
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6540517116856854e-05,
      "loss": 3.2416,
      "step": 5413
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.651162790697675e-05,
      "loss": 3.1169,
      "step": 5414
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6482738697096635e-05,
      "loss": 3.0852,
      "step": 5415
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.645384948721653e-05,
      "loss": 3.0634,
      "step": 5416
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6424960277336416e-05,
      "loss": 3.0914,
      "step": 5417
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.639607106745631e-05,
      "loss": 3.1037,
      "step": 5418
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.63671818575762e-05,
      "loss": 3.0749,
      "step": 5419
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6338292647696084e-05,
      "loss": 2.9571,
      "step": 5420
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.630940343781598e-05,
      "loss": 3.0772,
      "step": 5421
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6280514227935865e-05,
      "loss": 3.1799,
      "step": 5422
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.625162501805576e-05,
      "loss": 3.3737,
      "step": 5423
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6222735808175646e-05,
      "loss": 2.9617,
      "step": 5424
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.619384659829554e-05,
      "loss": 3.2483,
      "step": 5425
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.6164957388415433e-05,
      "loss": 3.2585,
      "step": 5426
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.613606817853532e-05,
      "loss": 3.2683,
      "step": 5427
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.610717896865521e-05,
      "loss": 3.1022,
      "step": 5428
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.60782897587751e-05,
      "loss": 3.2188,
      "step": 5429
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.604940054889499e-05,
      "loss": 3.1245,
      "step": 5430
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.602051133901488e-05,
      "loss": 3.2285,
      "step": 5431
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.599162212913477e-05,
      "loss": 3.1879,
      "step": 5432
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5962732919254656e-05,
      "loss": 3.0554,
      "step": 5433
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.593384370937455e-05,
      "loss": 3.1055,
      "step": 5434
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.590495449949444e-05,
      "loss": 3.2009,
      "step": 5435
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.587606528961433e-05,
      "loss": 3.1747,
      "step": 5436
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5847176079734225e-05,
      "loss": 3.0973,
      "step": 5437
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.581828686985411e-05,
      "loss": 3.0826,
      "step": 5438
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5789397659974006e-05,
      "loss": 3.0979,
      "step": 5439
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.576050845009389e-05,
      "loss": 3.0841,
      "step": 5440
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.573161924021378e-05,
      "loss": 2.9547,
      "step": 5441
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.5702730030333674e-05,
      "loss": 3.1232,
      "step": 5442
    },
    {
      "epoch": 0.77,
      "learning_rate": 4.567384082045356e-05,
      "loss": 3.2677,
      "step": 5443
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5644951610573454e-05,
      "loss": 3.2299,
      "step": 5444
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.561606240069334e-05,
      "loss": 3.1185,
      "step": 5445
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.558717319081323e-05,
      "loss": 3.0334,
      "step": 5446
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.555828398093312e-05,
      "loss": 3.1425,
      "step": 5447
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5529394771053016e-05,
      "loss": 3.058,
      "step": 5448
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.55005055611729e-05,
      "loss": 3.0042,
      "step": 5449
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.54716163512928e-05,
      "loss": 3.1667,
      "step": 5450
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5442727141412684e-05,
      "loss": 3.1671,
      "step": 5451
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.541383793153258e-05,
      "loss": 3.0781,
      "step": 5452
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5384948721652465e-05,
      "loss": 3.0669,
      "step": 5453
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.535605951177235e-05,
      "loss": 3.0942,
      "step": 5454
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5327170301892246e-05,
      "loss": 3.0422,
      "step": 5455
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.529828109201213e-05,
      "loss": 3.1276,
      "step": 5456
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.526939188213203e-05,
      "loss": 2.9173,
      "step": 5457
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5240502672251914e-05,
      "loss": 3.0792,
      "step": 5458
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.521161346237181e-05,
      "loss": 3.1322,
      "step": 5459
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.51827242524917e-05,
      "loss": 3.112,
      "step": 5460
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.515383504261159e-05,
      "loss": 3.1221,
      "step": 5461
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5124945832731476e-05,
      "loss": 3.1167,
      "step": 5462
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.509605662285137e-05,
      "loss": 3.2434,
      "step": 5463
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.5067167412971256e-05,
      "loss": 3.0045,
      "step": 5464
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.503827820309115e-05,
      "loss": 3.0463,
      "step": 5465
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.500938899321104e-05,
      "loss": 2.8559,
      "step": 5466
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4980499783330924e-05,
      "loss": 3.0624,
      "step": 5467
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.495161057345082e-05,
      "loss": 3.1129,
      "step": 5468
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4922721363570705e-05,
      "loss": 3.1786,
      "step": 5469
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.48938321536906e-05,
      "loss": 2.9897,
      "step": 5470
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.486494294381049e-05,
      "loss": 3.013,
      "step": 5471
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.483605373393038e-05,
      "loss": 3.2182,
      "step": 5472
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4807164524050274e-05,
      "loss": 3.0045,
      "step": 5473
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.477827531417016e-05,
      "loss": 2.9859,
      "step": 5474
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.474938610429005e-05,
      "loss": 3.1631,
      "step": 5475
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.472049689440994e-05,
      "loss": 3.3028,
      "step": 5476
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.469160768452983e-05,
      "loss": 3.1968,
      "step": 5477
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.466271847464972e-05,
      "loss": 3.1583,
      "step": 5478
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.463382926476961e-05,
      "loss": 3.1995,
      "step": 5479
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.46049400548895e-05,
      "loss": 3.1569,
      "step": 5480
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.457605084500939e-05,
      "loss": 3.0926,
      "step": 5481
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.454716163512928e-05,
      "loss": 3.1807,
      "step": 5482
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.451827242524917e-05,
      "loss": 3.1208,
      "step": 5483
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4489383215369065e-05,
      "loss": 3.1641,
      "step": 5484
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.446049400548895e-05,
      "loss": 3.1255,
      "step": 5485
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4431604795608846e-05,
      "loss": 3.2746,
      "step": 5486
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.440271558572873e-05,
      "loss": 3.1542,
      "step": 5487
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.437382637584862e-05,
      "loss": 2.9932,
      "step": 5488
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4344937165968514e-05,
      "loss": 3.2184,
      "step": 5489
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.43160479560884e-05,
      "loss": 3.1889,
      "step": 5490
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4287158746208295e-05,
      "loss": 3.1113,
      "step": 5491
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.425826953632818e-05,
      "loss": 3.135,
      "step": 5492
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.422938032644807e-05,
      "loss": 3.2097,
      "step": 5493
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.420049111656796e-05,
      "loss": 3.2467,
      "step": 5494
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.417160190668786e-05,
      "loss": 3.2019,
      "step": 5495
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4142712696807744e-05,
      "loss": 3.1264,
      "step": 5496
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.411382348692764e-05,
      "loss": 3.0359,
      "step": 5497
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4084934277047525e-05,
      "loss": 3.1811,
      "step": 5498
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.405604506716742e-05,
      "loss": 3.1627,
      "step": 5499
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.4027155857287305e-05,
      "loss": 3.0806,
      "step": 5500
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.399826664740719e-05,
      "loss": 3.117,
      "step": 5501
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.3969377437527086e-05,
      "loss": 3.1178,
      "step": 5502
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.394048822764697e-05,
      "loss": 3.1311,
      "step": 5503
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.391159901776686e-05,
      "loss": 3.1437,
      "step": 5504
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.3882709807886754e-05,
      "loss": 3.1687,
      "step": 5505
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.385382059800665e-05,
      "loss": 3.1019,
      "step": 5506
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.382493138812654e-05,
      "loss": 3.0737,
      "step": 5507
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.379604217824643e-05,
      "loss": 3.1824,
      "step": 5508
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.3767152968366316e-05,
      "loss": 3.1009,
      "step": 5509
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.373826375848621e-05,
      "loss": 3.096,
      "step": 5510
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.37093745486061e-05,
      "loss": 3.1401,
      "step": 5511
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.368048533872599e-05,
      "loss": 3.092,
      "step": 5512
    },
    {
      "epoch": 0.78,
      "learning_rate": 4.365159612884588e-05,
      "loss": 3.1106,
      "step": 5513
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3622706918965765e-05,
      "loss": 3.0776,
      "step": 5514
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.359381770908566e-05,
      "loss": 2.9404,
      "step": 5515
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3564928499205546e-05,
      "loss": 3.2286,
      "step": 5516
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.353603928932544e-05,
      "loss": 3.0828,
      "step": 5517
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.350715007944533e-05,
      "loss": 2.9596,
      "step": 5518
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.347826086956522e-05,
      "loss": 3.0728,
      "step": 5519
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3449371659685114e-05,
      "loss": 3.1399,
      "step": 5520
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3420482449805e-05,
      "loss": 3.0716,
      "step": 5521
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.339159323992489e-05,
      "loss": 3.1167,
      "step": 5522
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.336270403004478e-05,
      "loss": 3.0735,
      "step": 5523
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.333381482016467e-05,
      "loss": 3.0712,
      "step": 5524
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.330492561028456e-05,
      "loss": 3.264,
      "step": 5525
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.327603640040445e-05,
      "loss": 3.0417,
      "step": 5526
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.324714719052434e-05,
      "loss": 3.154,
      "step": 5527
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.321825798064423e-05,
      "loss": 3.085,
      "step": 5528
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3189368770764125e-05,
      "loss": 3.2064,
      "step": 5529
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.316047956088401e-05,
      "loss": 3.1593,
      "step": 5530
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3131590351003906e-05,
      "loss": 3.197,
      "step": 5531
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.310270114112379e-05,
      "loss": 2.9919,
      "step": 5532
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3073811931243686e-05,
      "loss": 3.201,
      "step": 5533
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.3044922721363574e-05,
      "loss": 2.9411,
      "step": 5534
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.301603351148346e-05,
      "loss": 3.051,
      "step": 5535
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2987144301603354e-05,
      "loss": 3.1746,
      "step": 5536
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.295825509172324e-05,
      "loss": 3.1132,
      "step": 5537
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.292936588184313e-05,
      "loss": 3.1978,
      "step": 5538
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.290047667196302e-05,
      "loss": 2.9316,
      "step": 5539
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.287158746208291e-05,
      "loss": 3.1084,
      "step": 5540
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.28426982522028e-05,
      "loss": 3.1946,
      "step": 5541
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.28138090423227e-05,
      "loss": 3.1314,
      "step": 5542
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2784919832442584e-05,
      "loss": 2.9638,
      "step": 5543
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.275603062256248e-05,
      "loss": 3.1317,
      "step": 5544
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2727141412682365e-05,
      "loss": 3.0273,
      "step": 5545
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.269825220280226e-05,
      "loss": 3.1777,
      "step": 5546
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2669362992922146e-05,
      "loss": 3.1611,
      "step": 5547
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.264047378304203e-05,
      "loss": 3.0826,
      "step": 5548
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.261158457316193e-05,
      "loss": 3.1074,
      "step": 5549
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2582695363281814e-05,
      "loss": 3.1507,
      "step": 5550
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.25538061534017e-05,
      "loss": 2.9495,
      "step": 5551
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2524916943521595e-05,
      "loss": 3.155,
      "step": 5552
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.249602773364149e-05,
      "loss": 3.105,
      "step": 5553
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.246713852376138e-05,
      "loss": 3.1968,
      "step": 5554
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.243824931388127e-05,
      "loss": 3.0571,
      "step": 5555
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2409360104001156e-05,
      "loss": 3.086,
      "step": 5556
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.238047089412105e-05,
      "loss": 3.1495,
      "step": 5557
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.235158168424094e-05,
      "loss": 3.0802,
      "step": 5558
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2322692474360824e-05,
      "loss": 3.2062,
      "step": 5559
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.229380326448072e-05,
      "loss": 2.9959,
      "step": 5560
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2264914054600605e-05,
      "loss": 3.0546,
      "step": 5561
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.22360248447205e-05,
      "loss": 3.0392,
      "step": 5562
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2207135634840386e-05,
      "loss": 3.0652,
      "step": 5563
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.217824642496028e-05,
      "loss": 2.9144,
      "step": 5564
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2149357215080174e-05,
      "loss": 2.9547,
      "step": 5565
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.212046800520006e-05,
      "loss": 3.2341,
      "step": 5566
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.2091578795319955e-05,
      "loss": 3.0326,
      "step": 5567
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.206268958543984e-05,
      "loss": 3.1595,
      "step": 5568
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.203380037555973e-05,
      "loss": 3.2042,
      "step": 5569
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.200491116567962e-05,
      "loss": 3.252,
      "step": 5570
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.197602195579951e-05,
      "loss": 3.0884,
      "step": 5571
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.1947132745919397e-05,
      "loss": 3.1607,
      "step": 5572
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.191824353603929e-05,
      "loss": 3.0982,
      "step": 5573
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.188935432615918e-05,
      "loss": 3.1095,
      "step": 5574
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.186046511627907e-05,
      "loss": 3.0324,
      "step": 5575
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.1831575906398965e-05,
      "loss": 3.0945,
      "step": 5576
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.180268669651885e-05,
      "loss": 3.096,
      "step": 5577
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.1773797486638746e-05,
      "loss": 3.102,
      "step": 5578
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.174490827675863e-05,
      "loss": 3.1228,
      "step": 5579
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.171601906687853e-05,
      "loss": 3.1557,
      "step": 5580
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.1687129856998414e-05,
      "loss": 3.12,
      "step": 5581
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.16582406471183e-05,
      "loss": 3.3135,
      "step": 5582
    },
    {
      "epoch": 0.79,
      "learning_rate": 4.1629351437238195e-05,
      "loss": 3.0905,
      "step": 5583
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.160046222735808e-05,
      "loss": 3.2271,
      "step": 5584
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.157157301747797e-05,
      "loss": 3.1322,
      "step": 5585
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.154268380759786e-05,
      "loss": 3.0577,
      "step": 5586
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.1513794597717757e-05,
      "loss": 2.9445,
      "step": 5587
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.148490538783765e-05,
      "loss": 3.1296,
      "step": 5588
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.145601617795754e-05,
      "loss": 3.1376,
      "step": 5589
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.1427126968077424e-05,
      "loss": 3.1108,
      "step": 5590
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.139823775819732e-05,
      "loss": 3.1172,
      "step": 5591
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.1369348548317205e-05,
      "loss": 3.1765,
      "step": 5592
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.134045933843709e-05,
      "loss": 3.1219,
      "step": 5593
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.1311570128556986e-05,
      "loss": 3.1936,
      "step": 5594
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.128268091867687e-05,
      "loss": 3.1505,
      "step": 5595
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.125379170879677e-05,
      "loss": 3.1478,
      "step": 5596
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.1224902498916654e-05,
      "loss": 3.1023,
      "step": 5597
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.119601328903655e-05,
      "loss": 3.0069,
      "step": 5598
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.1167124079156435e-05,
      "loss": 3.054,
      "step": 5599
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.113823486927633e-05,
      "loss": 2.9706,
      "step": 5600
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.110934565939622e-05,
      "loss": 3.152,
      "step": 5601
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.108045644951611e-05,
      "loss": 3.076,
      "step": 5602
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.1051567239636e-05,
      "loss": 3.2783,
      "step": 5603
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.102267802975589e-05,
      "loss": 3.0151,
      "step": 5604
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.099378881987578e-05,
      "loss": 3.1927,
      "step": 5605
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0964899609995665e-05,
      "loss": 3.0818,
      "step": 5606
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.093601040011556e-05,
      "loss": 3.0521,
      "step": 5607
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0907121190235446e-05,
      "loss": 3.1366,
      "step": 5608
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.087823198035534e-05,
      "loss": 3.1397,
      "step": 5609
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0849342770475226e-05,
      "loss": 3.1935,
      "step": 5610
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.082045356059512e-05,
      "loss": 3.1474,
      "step": 5611
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0791564350715014e-05,
      "loss": 3.0972,
      "step": 5612
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.07626751408349e-05,
      "loss": 3.0839,
      "step": 5613
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.073378593095479e-05,
      "loss": 3.0802,
      "step": 5614
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.070489672107468e-05,
      "loss": 3.189,
      "step": 5615
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.067600751119457e-05,
      "loss": 3.0959,
      "step": 5616
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.064711830131446e-05,
      "loss": 3.2138,
      "step": 5617
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.061822909143435e-05,
      "loss": 2.9459,
      "step": 5618
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.058933988155424e-05,
      "loss": 3.188,
      "step": 5619
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.056045067167413e-05,
      "loss": 3.2137,
      "step": 5620
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.053156146179402e-05,
      "loss": 3.0826,
      "step": 5621
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.050267225191391e-05,
      "loss": 3.2074,
      "step": 5622
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0473783042033806e-05,
      "loss": 3.1752,
      "step": 5623
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.044489383215369e-05,
      "loss": 3.0614,
      "step": 5624
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0416004622273586e-05,
      "loss": 3.1443,
      "step": 5625
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0387115412393473e-05,
      "loss": 3.1138,
      "step": 5626
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.035822620251336e-05,
      "loss": 3.0843,
      "step": 5627
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0329336992633254e-05,
      "loss": 3.1314,
      "step": 5628
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.030044778275314e-05,
      "loss": 3.0496,
      "step": 5629
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0271558572873035e-05,
      "loss": 3.2638,
      "step": 5630
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.024266936299292e-05,
      "loss": 3.176,
      "step": 5631
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.021378015311281e-05,
      "loss": 3.1037,
      "step": 5632
    },
    {
      "epoch": 0.8,
      "eval_loss": 3.35094952583313,
      "eval_runtime": 472.6533,
      "eval_samples_per_second": 43.345,
      "eval_steps_per_second": 14.448,
      "step": 5632
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.01848909432327e-05,
      "loss": 3.1948,
      "step": 5633
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.01560017333526e-05,
      "loss": 3.1198,
      "step": 5634
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.012711252347249e-05,
      "loss": 3.2102,
      "step": 5635
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.009822331359238e-05,
      "loss": 3.1267,
      "step": 5636
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0069334103712265e-05,
      "loss": 3.1205,
      "step": 5637
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.004044489383216e-05,
      "loss": 3.1644,
      "step": 5638
    },
    {
      "epoch": 0.8,
      "learning_rate": 4.0011555683952046e-05,
      "loss": 3.211,
      "step": 5639
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.998266647407193e-05,
      "loss": 3.1573,
      "step": 5640
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.995377726419183e-05,
      "loss": 3.0154,
      "step": 5641
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.9924888054311714e-05,
      "loss": 2.8478,
      "step": 5642
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.989599884443161e-05,
      "loss": 3.2054,
      "step": 5643
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.9867109634551495e-05,
      "loss": 3.1459,
      "step": 5644
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.983822042467139e-05,
      "loss": 3.2107,
      "step": 5645
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.980933121479128e-05,
      "loss": 3.0071,
      "step": 5646
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.978044200491117e-05,
      "loss": 3.0957,
      "step": 5647
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.9751552795031056e-05,
      "loss": 3.1346,
      "step": 5648
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.972266358515095e-05,
      "loss": 3.0385,
      "step": 5649
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.969377437527084e-05,
      "loss": 2.986,
      "step": 5650
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.966488516539073e-05,
      "loss": 3.1062,
      "step": 5651
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.963599595551062e-05,
      "loss": 3.1744,
      "step": 5652
    },
    {
      "epoch": 0.8,
      "learning_rate": 3.9607106745630505e-05,
      "loss": 3.0897,
      "step": 5653
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.95782175357504e-05,
      "loss": 3.2293,
      "step": 5654
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.9549328325870286e-05,
      "loss": 3.1602,
      "step": 5655
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.952043911599018e-05,
      "loss": 3.1055,
      "step": 5656
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.949154990611007e-05,
      "loss": 2.9947,
      "step": 5657
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.946266069622996e-05,
      "loss": 3.0883,
      "step": 5658
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.9433771486349854e-05,
      "loss": 3.0944,
      "step": 5659
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.940488227646974e-05,
      "loss": 3.1408,
      "step": 5660
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.937599306658963e-05,
      "loss": 3.1542,
      "step": 5661
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.934710385670952e-05,
      "loss": 3.1758,
      "step": 5662
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.931821464682941e-05,
      "loss": 3.0643,
      "step": 5663
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.92893254369493e-05,
      "loss": 3.2559,
      "step": 5664
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.926043622706919e-05,
      "loss": 3.1238,
      "step": 5665
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.923154701718908e-05,
      "loss": 3.1534,
      "step": 5666
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.920265780730897e-05,
      "loss": 3.1156,
      "step": 5667
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.917376859742886e-05,
      "loss": 3.1027,
      "step": 5668
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.914487938754875e-05,
      "loss": 3.1216,
      "step": 5669
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.9115990177668646e-05,
      "loss": 3.0997,
      "step": 5670
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.908710096778853e-05,
      "loss": 3.183,
      "step": 5671
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.905821175790843e-05,
      "loss": 3.1182,
      "step": 5672
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.9029322548028314e-05,
      "loss": 3.1436,
      "step": 5673
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.90004333381482e-05,
      "loss": 3.1885,
      "step": 5674
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8971544128268095e-05,
      "loss": 3.1091,
      "step": 5675
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.894265491838798e-05,
      "loss": 3.0086,
      "step": 5676
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8913765708507876e-05,
      "loss": 3.1322,
      "step": 5677
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.888487649862776e-05,
      "loss": 3.086,
      "step": 5678
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.885598728874765e-05,
      "loss": 3.1215,
      "step": 5679
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8827098078867544e-05,
      "loss": 3.0616,
      "step": 5680
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.879820886898744e-05,
      "loss": 3.3159,
      "step": 5681
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8769319659107324e-05,
      "loss": 3.1251,
      "step": 5682
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.874043044922722e-05,
      "loss": 3.1339,
      "step": 5683
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8711541239347105e-05,
      "loss": 3.2388,
      "step": 5684
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8682652029467e-05,
      "loss": 3.132,
      "step": 5685
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8653762819586886e-05,
      "loss": 3.1474,
      "step": 5686
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.862487360970677e-05,
      "loss": 3.1231,
      "step": 5687
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.859598439982667e-05,
      "loss": 3.1913,
      "step": 5688
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8567095189946554e-05,
      "loss": 3.2096,
      "step": 5689
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.853820598006645e-05,
      "loss": 3.144,
      "step": 5690
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8509316770186335e-05,
      "loss": 3.079,
      "step": 5691
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.848042756030623e-05,
      "loss": 3.1755,
      "step": 5692
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.845153835042612e-05,
      "loss": 3.1755,
      "step": 5693
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.842264914054601e-05,
      "loss": 3.0532,
      "step": 5694
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.83937599306659e-05,
      "loss": 3.1325,
      "step": 5695
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.836487072078579e-05,
      "loss": 3.0894,
      "step": 5696
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.833598151090568e-05,
      "loss": 2.9988,
      "step": 5697
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.830709230102557e-05,
      "loss": 2.995,
      "step": 5698
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.827820309114546e-05,
      "loss": 3.0924,
      "step": 5699
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8249313881265345e-05,
      "loss": 3.1804,
      "step": 5700
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.822042467138524e-05,
      "loss": 3.1738,
      "step": 5701
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8191535461505126e-05,
      "loss": 2.9256,
      "step": 5702
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.816264625162502e-05,
      "loss": 3.1161,
      "step": 5703
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8133757041744914e-05,
      "loss": 3.2177,
      "step": 5704
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.81048678318648e-05,
      "loss": 2.9096,
      "step": 5705
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.8075978621984695e-05,
      "loss": 3.1869,
      "step": 5706
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.804708941210458e-05,
      "loss": 2.9959,
      "step": 5707
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.801820020222447e-05,
      "loss": 3.15,
      "step": 5708
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.798931099234436e-05,
      "loss": 3.203,
      "step": 5709
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.796042178246425e-05,
      "loss": 3.1453,
      "step": 5710
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.7931532572584144e-05,
      "loss": 3.0416,
      "step": 5711
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.790264336270403e-05,
      "loss": 3.1197,
      "step": 5712
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.787375415282392e-05,
      "loss": 3.0855,
      "step": 5713
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.784486494294381e-05,
      "loss": 3.2529,
      "step": 5714
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.7815975733063705e-05,
      "loss": 3.1082,
      "step": 5715
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.778708652318359e-05,
      "loss": 3.0494,
      "step": 5716
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.7758197313303486e-05,
      "loss": 3.1625,
      "step": 5717
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.772930810342337e-05,
      "loss": 3.1332,
      "step": 5718
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.770041889354327e-05,
      "loss": 2.9911,
      "step": 5719
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.7671529683663154e-05,
      "loss": 3.1648,
      "step": 5720
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.764264047378304e-05,
      "loss": 3.1478,
      "step": 5721
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.7613751263902935e-05,
      "loss": 3.2288,
      "step": 5722
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.758486205402282e-05,
      "loss": 3.0045,
      "step": 5723
    },
    {
      "epoch": 0.81,
      "learning_rate": 3.7555972844142716e-05,
      "loss": 3.1102,
      "step": 5724
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.75270836342626e-05,
      "loss": 3.0512,
      "step": 5725
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.749819442438249e-05,
      "loss": 3.0181,
      "step": 5726
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.7469305214502384e-05,
      "loss": 3.0271,
      "step": 5727
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.744041600462228e-05,
      "loss": 3.019,
      "step": 5728
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.7411526794742165e-05,
      "loss": 3.1775,
      "step": 5729
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.738263758486206e-05,
      "loss": 3.1779,
      "step": 5730
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.7353748374981946e-05,
      "loss": 2.8991,
      "step": 5731
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.732485916510184e-05,
      "loss": 3.1068,
      "step": 5732
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.7295969955221727e-05,
      "loss": 2.9594,
      "step": 5733
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.7267080745341614e-05,
      "loss": 3.0418,
      "step": 5734
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.723819153546151e-05,
      "loss": 3.0018,
      "step": 5735
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.7209302325581394e-05,
      "loss": 3.1529,
      "step": 5736
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.718041311570128e-05,
      "loss": 3.0914,
      "step": 5737
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.7151523905821175e-05,
      "loss": 3.1752,
      "step": 5738
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.712263469594107e-05,
      "loss": 3.0554,
      "step": 5739
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.709374548606096e-05,
      "loss": 3.147,
      "step": 5740
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.706485627618085e-05,
      "loss": 3.1708,
      "step": 5741
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.703596706630074e-05,
      "loss": 3.1171,
      "step": 5742
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.700707785642063e-05,
      "loss": 3.1213,
      "step": 5743
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.697818864654052e-05,
      "loss": 3.188,
      "step": 5744
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.694929943666041e-05,
      "loss": 2.901,
      "step": 5745
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.69204102267803e-05,
      "loss": 3.0635,
      "step": 5746
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6891521016900186e-05,
      "loss": 3.1305,
      "step": 5747
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.686263180702008e-05,
      "loss": 3.0389,
      "step": 5748
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.683374259713997e-05,
      "loss": 3.125,
      "step": 5749
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.680485338725986e-05,
      "loss": 3.0956,
      "step": 5750
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6775964177379754e-05,
      "loss": 3.1115,
      "step": 5751
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.674707496749964e-05,
      "loss": 3.0745,
      "step": 5752
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6718185757619535e-05,
      "loss": 3.2283,
      "step": 5753
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.668929654773942e-05,
      "loss": 3.2078,
      "step": 5754
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.666040733785931e-05,
      "loss": 3.0762,
      "step": 5755
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.66315181279792e-05,
      "loss": 3.127,
      "step": 5756
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.660262891809909e-05,
      "loss": 3.0701,
      "step": 5757
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.657373970821898e-05,
      "loss": 3.0519,
      "step": 5758
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.654485049833887e-05,
      "loss": 3.0686,
      "step": 5759
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.651596128845876e-05,
      "loss": 3.1107,
      "step": 5760
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.648707207857865e-05,
      "loss": 3.004,
      "step": 5761
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6458182868698546e-05,
      "loss": 2.9062,
      "step": 5762
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.642929365881843e-05,
      "loss": 2.9686,
      "step": 5763
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.640040444893833e-05,
      "loss": 3.1951,
      "step": 5764
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6371515239058214e-05,
      "loss": 3.1462,
      "step": 5765
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.634262602917811e-05,
      "loss": 3.2414,
      "step": 5766
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6313736819297995e-05,
      "loss": 3.0106,
      "step": 5767
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.628484760941788e-05,
      "loss": 3.2559,
      "step": 5768
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6255958399537775e-05,
      "loss": 3.1398,
      "step": 5769
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.622706918965766e-05,
      "loss": 3.1913,
      "step": 5770
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.619817997977755e-05,
      "loss": 3.045,
      "step": 5771
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6169290769897443e-05,
      "loss": 3.1046,
      "step": 5772
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.614040156001734e-05,
      "loss": 3.1989,
      "step": 5773
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6111512350137224e-05,
      "loss": 3.2277,
      "step": 5774
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.608262314025712e-05,
      "loss": 3.0579,
      "step": 5775
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.6053733930377005e-05,
      "loss": 3.0565,
      "step": 5776
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.60248447204969e-05,
      "loss": 3.2018,
      "step": 5777
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.5995955510616786e-05,
      "loss": 3.2889,
      "step": 5778
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.596706630073668e-05,
      "loss": 3.0778,
      "step": 5779
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.593817709085657e-05,
      "loss": 3.1019,
      "step": 5780
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.5909287880976454e-05,
      "loss": 3.241,
      "step": 5781
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.588039867109635e-05,
      "loss": 3.1829,
      "step": 5782
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.5851509461216235e-05,
      "loss": 3.1955,
      "step": 5783
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.582262025133613e-05,
      "loss": 3.2168,
      "step": 5784
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.5793731041456016e-05,
      "loss": 3.1748,
      "step": 5785
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.576484183157591e-05,
      "loss": 3.2433,
      "step": 5786
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.57359526216958e-05,
      "loss": 3.0951,
      "step": 5787
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.570706341181569e-05,
      "loss": 3.1137,
      "step": 5788
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.567817420193558e-05,
      "loss": 3.1623,
      "step": 5789
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.564928499205547e-05,
      "loss": 3.0718,
      "step": 5790
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.562039578217536e-05,
      "loss": 3.1426,
      "step": 5791
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.5591506572295245e-05,
      "loss": 3.1258,
      "step": 5792
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.556261736241514e-05,
      "loss": 3.1174,
      "step": 5793
    },
    {
      "epoch": 0.82,
      "learning_rate": 3.5533728152535026e-05,
      "loss": 3.0803,
      "step": 5794
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.550483894265492e-05,
      "loss": 3.174,
      "step": 5795
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.547594973277481e-05,
      "loss": 3.209,
      "step": 5796
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.54470605228947e-05,
      "loss": 3.1218,
      "step": 5797
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.5418171313014595e-05,
      "loss": 3.1554,
      "step": 5798
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.538928210313448e-05,
      "loss": 3.1106,
      "step": 5799
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.5360392893254376e-05,
      "loss": 3.1266,
      "step": 5800
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.533150368337426e-05,
      "loss": 3.1263,
      "step": 5801
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.530261447349415e-05,
      "loss": 3.1582,
      "step": 5802
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.5273725263614044e-05,
      "loss": 3.1887,
      "step": 5803
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.524483605373393e-05,
      "loss": 3.0272,
      "step": 5804
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.521594684385382e-05,
      "loss": 3.1066,
      "step": 5805
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.518705763397371e-05,
      "loss": 3.1363,
      "step": 5806
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.51581684240936e-05,
      "loss": 3.0005,
      "step": 5807
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.512927921421349e-05,
      "loss": 3.1468,
      "step": 5808
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.5100390004333386e-05,
      "loss": 3.137,
      "step": 5809
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.507150079445327e-05,
      "loss": 2.8747,
      "step": 5810
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.504261158457317e-05,
      "loss": 3.1055,
      "step": 5811
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.5013722374693054e-05,
      "loss": 3.0413,
      "step": 5812
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.498483316481295e-05,
      "loss": 3.0186,
      "step": 5813
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4955943954932835e-05,
      "loss": 3.0483,
      "step": 5814
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.492705474505272e-05,
      "loss": 2.9472,
      "step": 5815
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4898165535172616e-05,
      "loss": 3.1869,
      "step": 5816
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.48692763252925e-05,
      "loss": 3.091,
      "step": 5817
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.484038711541239e-05,
      "loss": 3.0505,
      "step": 5818
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4811497905532284e-05,
      "loss": 3.0532,
      "step": 5819
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.478260869565218e-05,
      "loss": 2.9293,
      "step": 5820
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.475371948577207e-05,
      "loss": 3.1341,
      "step": 5821
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.472483027589196e-05,
      "loss": 2.8325,
      "step": 5822
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4695941066011846e-05,
      "loss": 3.2622,
      "step": 5823
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.466705185613174e-05,
      "loss": 3.1963,
      "step": 5824
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4638162646251626e-05,
      "loss": 3.0659,
      "step": 5825
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4609273436371513e-05,
      "loss": 3.1355,
      "step": 5826
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.458038422649141e-05,
      "loss": 3.0923,
      "step": 5827
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4551495016611294e-05,
      "loss": 3.1681,
      "step": 5828
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.452260580673119e-05,
      "loss": 3.0811,
      "step": 5829
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4493716596851075e-05,
      "loss": 3.1874,
      "step": 5830
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.446482738697097e-05,
      "loss": 3.1023,
      "step": 5831
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.443593817709086e-05,
      "loss": 3.1716,
      "step": 5832
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.440704896721075e-05,
      "loss": 2.9498,
      "step": 5833
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4378159757330644e-05,
      "loss": 3.0727,
      "step": 5834
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.434927054745053e-05,
      "loss": 3.0121,
      "step": 5835
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.432038133757042e-05,
      "loss": 3.1766,
      "step": 5836
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.429149212769031e-05,
      "loss": 3.0586,
      "step": 5837
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.42626029178102e-05,
      "loss": 3.2201,
      "step": 5838
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4233713707930086e-05,
      "loss": 3.0646,
      "step": 5839
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.420482449804998e-05,
      "loss": 3.1721,
      "step": 5840
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.417593528816987e-05,
      "loss": 3.0589,
      "step": 5841
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.414704607828976e-05,
      "loss": 3.1793,
      "step": 5842
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.411815686840965e-05,
      "loss": 3.114,
      "step": 5843
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.408926765852954e-05,
      "loss": 3.0718,
      "step": 5844
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.4060378448649435e-05,
      "loss": 3.2071,
      "step": 5845
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.403148923876932e-05,
      "loss": 3.1263,
      "step": 5846
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.400260002888921e-05,
      "loss": 3.1385,
      "step": 5847
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.39737108190091e-05,
      "loss": 3.032,
      "step": 5848
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.394482160912899e-05,
      "loss": 3.0534,
      "step": 5849
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.3915932399248884e-05,
      "loss": 3.0448,
      "step": 5850
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.388704318936877e-05,
      "loss": 3.0444,
      "step": 5851
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.385815397948866e-05,
      "loss": 3.1049,
      "step": 5852
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.382926476960855e-05,
      "loss": 3.0882,
      "step": 5853
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.380037555972844e-05,
      "loss": 3.1759,
      "step": 5854
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.377148634984833e-05,
      "loss": 3.0127,
      "step": 5855
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.374259713996823e-05,
      "loss": 3.1281,
      "step": 5856
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.3713707930088114e-05,
      "loss": 3.1889,
      "step": 5857
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.368481872020801e-05,
      "loss": 2.9567,
      "step": 5858
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.3655929510327895e-05,
      "loss": 3.0444,
      "step": 5859
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.362704030044778e-05,
      "loss": 3.0166,
      "step": 5860
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.3598151090567675e-05,
      "loss": 3.1204,
      "step": 5861
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.356926188068756e-05,
      "loss": 3.0538,
      "step": 5862
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.3540372670807456e-05,
      "loss": 3.235,
      "step": 5863
    },
    {
      "epoch": 0.83,
      "learning_rate": 3.351148346092734e-05,
      "loss": 3.2553,
      "step": 5864
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.348259425104723e-05,
      "loss": 2.9994,
      "step": 5865
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.3453705041167124e-05,
      "loss": 3.061,
      "step": 5866
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.342481583128702e-05,
      "loss": 3.1033,
      "step": 5867
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.339592662140691e-05,
      "loss": 3.0325,
      "step": 5868
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.33670374115268e-05,
      "loss": 3.1356,
      "step": 5869
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.3338148201646686e-05,
      "loss": 3.1409,
      "step": 5870
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.330925899176658e-05,
      "loss": 2.8731,
      "step": 5871
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.328036978188647e-05,
      "loss": 2.917,
      "step": 5872
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.3251480572006354e-05,
      "loss": 3.1651,
      "step": 5873
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.322259136212625e-05,
      "loss": 3.0817,
      "step": 5874
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.3193702152246135e-05,
      "loss": 3.1778,
      "step": 5875
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.316481294236603e-05,
      "loss": 3.1177,
      "step": 5876
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.3135923732485916e-05,
      "loss": 2.9929,
      "step": 5877
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.310703452260581e-05,
      "loss": 3.0402,
      "step": 5878
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.30781453127257e-05,
      "loss": 3.0111,
      "step": 5879
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.304925610284559e-05,
      "loss": 3.1887,
      "step": 5880
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.302036689296548e-05,
      "loss": 2.7138,
      "step": 5881
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.299147768308537e-05,
      "loss": 3.0009,
      "step": 5882
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.296258847320526e-05,
      "loss": 2.9829,
      "step": 5883
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.293369926332515e-05,
      "loss": 3.099,
      "step": 5884
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.290481005344504e-05,
      "loss": 3.1086,
      "step": 5885
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2875920843564926e-05,
      "loss": 3.0638,
      "step": 5886
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.284703163368482e-05,
      "loss": 3.0933,
      "step": 5887
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.281814242380471e-05,
      "loss": 3.1918,
      "step": 5888
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.27892532139246e-05,
      "loss": 3.117,
      "step": 5889
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2760364004044495e-05,
      "loss": 3.009,
      "step": 5890
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.273147479416438e-05,
      "loss": 3.1886,
      "step": 5891
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2702585584284276e-05,
      "loss": 2.9678,
      "step": 5892
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.267369637440416e-05,
      "loss": 3.154,
      "step": 5893
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.264480716452405e-05,
      "loss": 3.207,
      "step": 5894
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2615917954643944e-05,
      "loss": 3.0728,
      "step": 5895
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.258702874476383e-05,
      "loss": 3.0138,
      "step": 5896
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2558139534883724e-05,
      "loss": 3.0363,
      "step": 5897
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.252925032500361e-05,
      "loss": 3.1933,
      "step": 5898
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.25003611151235e-05,
      "loss": 3.0942,
      "step": 5899
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.247147190524339e-05,
      "loss": 3.1072,
      "step": 5900
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2442582695363286e-05,
      "loss": 3.1678,
      "step": 5901
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.241369348548317e-05,
      "loss": 3.0824,
      "step": 5902
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.238480427560307e-05,
      "loss": 3.0526,
      "step": 5903
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2355915065722954e-05,
      "loss": 3.0983,
      "step": 5904
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.232702585584285e-05,
      "loss": 3.2043,
      "step": 5905
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2298136645962735e-05,
      "loss": 3.1118,
      "step": 5906
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.226924743608262e-05,
      "loss": 3.1107,
      "step": 5907
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2240358226202516e-05,
      "loss": 3.1229,
      "step": 5908
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.22114690163224e-05,
      "loss": 3.1159,
      "step": 5909
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.21825798064423e-05,
      "loss": 3.0419,
      "step": 5910
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2153690596562184e-05,
      "loss": 3.2061,
      "step": 5911
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.212480138668207e-05,
      "loss": 3.106,
      "step": 5912
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2095912176801965e-05,
      "loss": 3.0385,
      "step": 5913
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.206702296692186e-05,
      "loss": 3.1538,
      "step": 5914
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.2038133757041745e-05,
      "loss": 3.0929,
      "step": 5915
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.200924454716164e-05,
      "loss": 3.0517,
      "step": 5916
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.1980355337281526e-05,
      "loss": 2.9422,
      "step": 5917
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.195146612740142e-05,
      "loss": 3.1486,
      "step": 5918
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.192257691752131e-05,
      "loss": 3.1113,
      "step": 5919
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.1893687707641194e-05,
      "loss": 3.1429,
      "step": 5920
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.186479849776109e-05,
      "loss": 3.0844,
      "step": 5921
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.1835909287880975e-05,
      "loss": 3.0432,
      "step": 5922
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.180702007800087e-05,
      "loss": 2.9963,
      "step": 5923
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.1778130868120756e-05,
      "loss": 2.9608,
      "step": 5924
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.174924165824065e-05,
      "loss": 3.1549,
      "step": 5925
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.1720352448360544e-05,
      "loss": 2.9382,
      "step": 5926
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.169146323848043e-05,
      "loss": 3.1066,
      "step": 5927
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.166257402860032e-05,
      "loss": 3.0371,
      "step": 5928
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.163368481872021e-05,
      "loss": 3.04,
      "step": 5929
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.16047956088401e-05,
      "loss": 3.1512,
      "step": 5930
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.157590639895999e-05,
      "loss": 3.1722,
      "step": 5931
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.154701718907988e-05,
      "loss": 2.9079,
      "step": 5932
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.1518127979199767e-05,
      "loss": 2.8963,
      "step": 5933
    },
    {
      "epoch": 0.84,
      "learning_rate": 3.148923876931966e-05,
      "loss": 3.0942,
      "step": 5934
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.146034955943955e-05,
      "loss": 3.1648,
      "step": 5935
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.143146034955944e-05,
      "loss": 3.0909,
      "step": 5936
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.1402571139679335e-05,
      "loss": 3.0367,
      "step": 5937
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.137368192979922e-05,
      "loss": 2.9949,
      "step": 5938
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.1344792719919116e-05,
      "loss": 3.0726,
      "step": 5939
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.1315903510039e-05,
      "loss": 3.2591,
      "step": 5940
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.128701430015889e-05,
      "loss": 2.9792,
      "step": 5941
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.1258125090278784e-05,
      "loss": 3.0628,
      "step": 5942
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.122923588039867e-05,
      "loss": 2.9426,
      "step": 5943
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.1200346670518565e-05,
      "loss": 3.031,
      "step": 5944
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.117145746063845e-05,
      "loss": 3.1403,
      "step": 5945
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.114256825075834e-05,
      "loss": 3.1605,
      "step": 5946
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.111367904087823e-05,
      "loss": 2.9976,
      "step": 5947
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.1084789830998127e-05,
      "loss": 3.139,
      "step": 5948
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.1055900621118014e-05,
      "loss": 3.1015,
      "step": 5949
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.102701141123791e-05,
      "loss": 2.9959,
      "step": 5950
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0998122201357794e-05,
      "loss": 3.1794,
      "step": 5951
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.096923299147769e-05,
      "loss": 3.0914,
      "step": 5952
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0940343781597575e-05,
      "loss": 3.0307,
      "step": 5953
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.091145457171746e-05,
      "loss": 2.9845,
      "step": 5954
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0882565361837356e-05,
      "loss": 3.0399,
      "step": 5955
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.085367615195724e-05,
      "loss": 3.2198,
      "step": 5956
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.082478694207713e-05,
      "loss": 3.1431,
      "step": 5957
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0795897732197024e-05,
      "loss": 2.9562,
      "step": 5958
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.076700852231692e-05,
      "loss": 3.0612,
      "step": 5959
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0738119312436805e-05,
      "loss": 2.9404,
      "step": 5960
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.07092301025567e-05,
      "loss": 3.0973,
      "step": 5961
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0680340892676586e-05,
      "loss": 3.1981,
      "step": 5962
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.065145168279648e-05,
      "loss": 3.0318,
      "step": 5963
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.062256247291637e-05,
      "loss": 3.0276,
      "step": 5964
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.059367326303626e-05,
      "loss": 2.9113,
      "step": 5965
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.056478405315615e-05,
      "loss": 3.1369,
      "step": 5966
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0535894843276035e-05,
      "loss": 3.108,
      "step": 5967
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.050700563339593e-05,
      "loss": 3.0283,
      "step": 5968
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.047811642351582e-05,
      "loss": 3.0095,
      "step": 5969
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0449227213635706e-05,
      "loss": 3.19,
      "step": 5970
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.04203380037556e-05,
      "loss": 3.0185,
      "step": 5971
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0391448793875487e-05,
      "loss": 3.0372,
      "step": 5972
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.036255958399538e-05,
      "loss": 2.9137,
      "step": 5973
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.033367037411527e-05,
      "loss": 3.1789,
      "step": 5974
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0304781164235158e-05,
      "loss": 3.2761,
      "step": 5975
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0275891954355052e-05,
      "loss": 2.9749,
      "step": 5976
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.024700274447494e-05,
      "loss": 2.9962,
      "step": 5977
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0218113534594833e-05,
      "loss": 3.0592,
      "step": 5978
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.018922432471472e-05,
      "loss": 3.1442,
      "step": 5979
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.016033511483461e-05,
      "loss": 3.0294,
      "step": 5980
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0131445904954504e-05,
      "loss": 3.1141,
      "step": 5981
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.010255669507439e-05,
      "loss": 2.9618,
      "step": 5982
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0073667485194278e-05,
      "loss": 2.9459,
      "step": 5983
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.0044778275314172e-05,
      "loss": 3.0394,
      "step": 5984
    },
    {
      "epoch": 0.85,
      "eval_loss": 3.3314239978790283,
      "eval_runtime": 472.5204,
      "eval_samples_per_second": 43.357,
      "eval_steps_per_second": 14.452,
      "step": 5984
    },
    {
      "epoch": 0.85,
      "learning_rate": 3.001588906543406e-05,
      "loss": 3.1192,
      "step": 5985
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9986999855553953e-05,
      "loss": 2.9188,
      "step": 5986
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9958110645673843e-05,
      "loss": 2.913,
      "step": 5987
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.992922143579373e-05,
      "loss": 3.1198,
      "step": 5988
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9900332225913624e-05,
      "loss": 2.9629,
      "step": 5989
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.987144301603351e-05,
      "loss": 3.0012,
      "step": 5990
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9842553806153402e-05,
      "loss": 3.0383,
      "step": 5991
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9813664596273296e-05,
      "loss": 3.0143,
      "step": 5992
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9784775386393183e-05,
      "loss": 3.0603,
      "step": 5993
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9755886176513076e-05,
      "loss": 2.8586,
      "step": 5994
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9726996966632964e-05,
      "loss": 3.0506,
      "step": 5995
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.969810775675285e-05,
      "loss": 3.1137,
      "step": 5996
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9669218546872744e-05,
      "loss": 2.989,
      "step": 5997
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9640329336992635e-05,
      "loss": 3.1343,
      "step": 5998
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.961144012711253e-05,
      "loss": 2.9752,
      "step": 5999
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9582550917232416e-05,
      "loss": 3.0701,
      "step": 6000
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9553661707352303e-05,
      "loss": 3.1102,
      "step": 6001
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9524772497472197e-05,
      "loss": 3.1009,
      "step": 6002
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9495883287592087e-05,
      "loss": 3.004,
      "step": 6003
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9466994077711974e-05,
      "loss": 3.1166,
      "step": 6004
    },
    {
      "epoch": 0.85,
      "learning_rate": 2.9438104867831868e-05,
      "loss": 3.0657,
      "step": 6005
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9409215657951755e-05,
      "loss": 2.9392,
      "step": 6006
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.938032644807165e-05,
      "loss": 3.1314,
      "step": 6007
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9351437238191536e-05,
      "loss": 3.0163,
      "step": 6008
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9322548028311426e-05,
      "loss": 3.2141,
      "step": 6009
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.929365881843132e-05,
      "loss": 2.911,
      "step": 6010
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9264769608551207e-05,
      "loss": 2.9646,
      "step": 6011
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.92358803986711e-05,
      "loss": 3.0786,
      "step": 6012
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9206991188790988e-05,
      "loss": 3.0391,
      "step": 6013
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9178101978910875e-05,
      "loss": 3.1584,
      "step": 6014
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.914921276903077e-05,
      "loss": 3.0752,
      "step": 6015
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.912032355915066e-05,
      "loss": 3.1557,
      "step": 6016
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9091434349270546e-05,
      "loss": 3.0763,
      "step": 6017
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.906254513939044e-05,
      "loss": 3.075,
      "step": 6018
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.9033655929510327e-05,
      "loss": 3.0939,
      "step": 6019
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.900476671963022e-05,
      "loss": 3.0377,
      "step": 6020
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.897587750975011e-05,
      "loss": 3.0022,
      "step": 6021
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.894698829987e-05,
      "loss": 3.0906,
      "step": 6022
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8918099089989892e-05,
      "loss": 3.1968,
      "step": 6023
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.888920988010978e-05,
      "loss": 3.0314,
      "step": 6024
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8860320670229666e-05,
      "loss": 3.1483,
      "step": 6025
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.883143146034956e-05,
      "loss": 3.1315,
      "step": 6026
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.880254225046945e-05,
      "loss": 3.1747,
      "step": 6027
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8773653040589345e-05,
      "loss": 3.1125,
      "step": 6028
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.874476383070923e-05,
      "loss": 3.2254,
      "step": 6029
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.871587462082912e-05,
      "loss": 3.0684,
      "step": 6030
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8686985410949013e-05,
      "loss": 3.0463,
      "step": 6031
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8658096201068903e-05,
      "loss": 2.9878,
      "step": 6032
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8629206991188793e-05,
      "loss": 3.1846,
      "step": 6033
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8600317781308684e-05,
      "loss": 3.2068,
      "step": 6034
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.857142857142857e-05,
      "loss": 2.9923,
      "step": 6035
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8542539361548465e-05,
      "loss": 2.9978,
      "step": 6036
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8513650151668352e-05,
      "loss": 3.0313,
      "step": 6037
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8484760941788242e-05,
      "loss": 3.124,
      "step": 6038
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8455871731908136e-05,
      "loss": 2.8379,
      "step": 6039
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8426982522028023e-05,
      "loss": 3.0884,
      "step": 6040
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8398093312147917e-05,
      "loss": 2.9867,
      "step": 6041
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8369204102267804e-05,
      "loss": 3.1433,
      "step": 6042
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.834031489238769e-05,
      "loss": 3.1496,
      "step": 6043
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8311425682507585e-05,
      "loss": 3.0847,
      "step": 6044
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8282536472627475e-05,
      "loss": 3.1325,
      "step": 6045
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8253647262747362e-05,
      "loss": 3.072,
      "step": 6046
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8224758052867256e-05,
      "loss": 3.1329,
      "step": 6047
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8195868842987143e-05,
      "loss": 2.9888,
      "step": 6048
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8166979633107037e-05,
      "loss": 3.1742,
      "step": 6049
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8138090423226927e-05,
      "loss": 3.077,
      "step": 6050
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8109201213346814e-05,
      "loss": 2.8865,
      "step": 6051
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8080312003466708e-05,
      "loss": 3.0586,
      "step": 6052
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.8051422793586595e-05,
      "loss": 2.852,
      "step": 6053
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.802253358370649e-05,
      "loss": 3.1931,
      "step": 6054
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7993644373826376e-05,
      "loss": 3.0282,
      "step": 6055
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7964755163946267e-05,
      "loss": 3.0781,
      "step": 6056
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.793586595406616e-05,
      "loss": 3.1172,
      "step": 6057
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7906976744186048e-05,
      "loss": 3.1747,
      "step": 6058
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7878087534305935e-05,
      "loss": 3.1268,
      "step": 6059
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.784919832442583e-05,
      "loss": 3.1027,
      "step": 6060
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.782030911454572e-05,
      "loss": 3.1948,
      "step": 6061
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.779141990466561e-05,
      "loss": 2.917,
      "step": 6062
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.77625306947855e-05,
      "loss": 3.135,
      "step": 6063
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7733641484905387e-05,
      "loss": 3.2288,
      "step": 6064
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.770475227502528e-05,
      "loss": 2.9871,
      "step": 6065
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7675863065145168e-05,
      "loss": 3.0479,
      "step": 6066
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.764697385526506e-05,
      "loss": 3.2121,
      "step": 6067
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7618084645384952e-05,
      "loss": 2.9775,
      "step": 6068
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.758919543550484e-05,
      "loss": 3.032,
      "step": 6069
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7560306225624733e-05,
      "loss": 2.9528,
      "step": 6070
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.753141701574462e-05,
      "loss": 2.944,
      "step": 6071
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.750252780586451e-05,
      "loss": 3.0658,
      "step": 6072
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.74736385959844e-05,
      "loss": 3.151,
      "step": 6073
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.744474938610429e-05,
      "loss": 3.1508,
      "step": 6074
    },
    {
      "epoch": 0.86,
      "learning_rate": 2.7415860176224185e-05,
      "loss": 3.1112,
      "step": 6075
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7386970966344072e-05,
      "loss": 3.1306,
      "step": 6076
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.735808175646396e-05,
      "loss": 3.0435,
      "step": 6077
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7329192546583853e-05,
      "loss": 3.1812,
      "step": 6078
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7300303336703743e-05,
      "loss": 3.0725,
      "step": 6079
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.727141412682363e-05,
      "loss": 3.0793,
      "step": 6080
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7242524916943524e-05,
      "loss": 3.0949,
      "step": 6081
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.721363570706341e-05,
      "loss": 3.0673,
      "step": 6082
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7184746497183305e-05,
      "loss": 3.1137,
      "step": 6083
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7155857287303192e-05,
      "loss": 3.1059,
      "step": 6084
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7126968077423083e-05,
      "loss": 3.1905,
      "step": 6085
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7098078867542976e-05,
      "loss": 3.1762,
      "step": 6086
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7069189657662863e-05,
      "loss": 3.1323,
      "step": 6087
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7040300447782757e-05,
      "loss": 2.8869,
      "step": 6088
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.7011411237902644e-05,
      "loss": 3.0922,
      "step": 6089
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6982522028022535e-05,
      "loss": 3.1239,
      "step": 6090
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.695363281814243e-05,
      "loss": 3.0454,
      "step": 6091
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6924743608262316e-05,
      "loss": 3.1074,
      "step": 6092
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6895854398382203e-05,
      "loss": 3.1191,
      "step": 6093
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6866965188502097e-05,
      "loss": 3.0851,
      "step": 6094
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6838075978621984e-05,
      "loss": 2.9811,
      "step": 6095
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6809186768741877e-05,
      "loss": 3.1226,
      "step": 6096
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6780297558861768e-05,
      "loss": 3.1572,
      "step": 6097
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6751408348981655e-05,
      "loss": 3.0967,
      "step": 6098
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.672251913910155e-05,
      "loss": 3.071,
      "step": 6099
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6693629929221436e-05,
      "loss": 3.1452,
      "step": 6100
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6664740719341326e-05,
      "loss": 3.0124,
      "step": 6101
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6635851509461217e-05,
      "loss": 3.021,
      "step": 6102
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6606962299581107e-05,
      "loss": 3.0597,
      "step": 6103
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6578073089701e-05,
      "loss": 3.1272,
      "step": 6104
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6549183879820888e-05,
      "loss": 3.0598,
      "step": 6105
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6520294669940775e-05,
      "loss": 3.0873,
      "step": 6106
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.649140546006067e-05,
      "loss": 3.3011,
      "step": 6107
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.646251625018056e-05,
      "loss": 2.9501,
      "step": 6108
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6433627040300453e-05,
      "loss": 3.1859,
      "step": 6109
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.640473783042034e-05,
      "loss": 3.1604,
      "step": 6110
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6375848620540227e-05,
      "loss": 3.0502,
      "step": 6111
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.634695941066012e-05,
      "loss": 3.1266,
      "step": 6112
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6318070200780008e-05,
      "loss": 3.1003,
      "step": 6113
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.62891809908999e-05,
      "loss": 2.9669,
      "step": 6114
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6260291781019792e-05,
      "loss": 3.0027,
      "step": 6115
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.623140257113968e-05,
      "loss": 3.2412,
      "step": 6116
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6202513361259573e-05,
      "loss": 3.1539,
      "step": 6117
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.617362415137946e-05,
      "loss": 3.1696,
      "step": 6118
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.614473494149935e-05,
      "loss": 3.1418,
      "step": 6119
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6115845731619244e-05,
      "loss": 2.9554,
      "step": 6120
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.608695652173913e-05,
      "loss": 2.9984,
      "step": 6121
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6058067311859025e-05,
      "loss": 3.2487,
      "step": 6122
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.6029178101978912e-05,
      "loss": 3.0161,
      "step": 6123
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.60002888920988e-05,
      "loss": 3.0282,
      "step": 6124
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5971399682218693e-05,
      "loss": 3.081,
      "step": 6125
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5942510472338584e-05,
      "loss": 3.0999,
      "step": 6126
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.591362126245847e-05,
      "loss": 3.1143,
      "step": 6127
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5884732052578365e-05,
      "loss": 3.0133,
      "step": 6128
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.585584284269825e-05,
      "loss": 3.0934,
      "step": 6129
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5826953632818145e-05,
      "loss": 2.9122,
      "step": 6130
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5798064422938033e-05,
      "loss": 3.0558,
      "step": 6131
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5769175213057923e-05,
      "loss": 3.1037,
      "step": 6132
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5740286003177817e-05,
      "loss": 3.083,
      "step": 6133
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5711396793297704e-05,
      "loss": 3.2274,
      "step": 6134
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.568250758341759e-05,
      "loss": 3.0305,
      "step": 6135
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5653618373537485e-05,
      "loss": 3.1308,
      "step": 6136
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5624729163657375e-05,
      "loss": 3.1218,
      "step": 6137
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.559583995377727e-05,
      "loss": 3.1344,
      "step": 6138
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5566950743897156e-05,
      "loss": 3.2072,
      "step": 6139
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5538061534017043e-05,
      "loss": 3.0067,
      "step": 6140
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5509172324136937e-05,
      "loss": 3.1007,
      "step": 6141
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5480283114256824e-05,
      "loss": 3.1694,
      "step": 6142
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5451393904376718e-05,
      "loss": 3.1094,
      "step": 6143
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5422504694496608e-05,
      "loss": 3.0832,
      "step": 6144
    },
    {
      "epoch": 0.87,
      "learning_rate": 2.5393615484616495e-05,
      "loss": 3.2072,
      "step": 6145
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.536472627473639e-05,
      "loss": 3.1934,
      "step": 6146
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.5335837064856276e-05,
      "loss": 3.0024,
      "step": 6147
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.5306947854976167e-05,
      "loss": 3.0558,
      "step": 6148
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.527805864509606e-05,
      "loss": 3.0448,
      "step": 6149
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.5249169435215947e-05,
      "loss": 3.2103,
      "step": 6150
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.522028022533584e-05,
      "loss": 3.0402,
      "step": 6151
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.519139101545573e-05,
      "loss": 3.0199,
      "step": 6152
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.5162501805575615e-05,
      "loss": 2.9997,
      "step": 6153
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.513361259569551e-05,
      "loss": 3.0992,
      "step": 6154
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.51047233858154e-05,
      "loss": 3.0769,
      "step": 6155
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.5075834175935287e-05,
      "loss": 3.0348,
      "step": 6156
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.504694496605518e-05,
      "loss": 3.0612,
      "step": 6157
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.5018055756175068e-05,
      "loss": 3.1116,
      "step": 6158
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4989166546294958e-05,
      "loss": 3.1809,
      "step": 6159
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.496027733641485e-05,
      "loss": 2.9785,
      "step": 6160
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4931388126534742e-05,
      "loss": 3.1658,
      "step": 6161
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4902498916654633e-05,
      "loss": 3.0982,
      "step": 6162
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.487360970677452e-05,
      "loss": 3.153,
      "step": 6163
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.484472049689441e-05,
      "loss": 3.1239,
      "step": 6164
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.48158312870143e-05,
      "loss": 3.0978,
      "step": 6165
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.478694207713419e-05,
      "loss": 3.008,
      "step": 6166
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.475805286725408e-05,
      "loss": 3.0674,
      "step": 6167
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4729163657373972e-05,
      "loss": 2.8771,
      "step": 6168
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4700274447493862e-05,
      "loss": 3.0125,
      "step": 6169
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4671385237613753e-05,
      "loss": 3.1738,
      "step": 6170
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.464249602773364e-05,
      "loss": 3.0806,
      "step": 6171
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4613606817853534e-05,
      "loss": 2.9466,
      "step": 6172
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4584717607973424e-05,
      "loss": 3.0043,
      "step": 6173
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4555828398093315e-05,
      "loss": 3.0152,
      "step": 6174
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4526939188213205e-05,
      "loss": 3.0443,
      "step": 6175
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4498049978333092e-05,
      "loss": 3.0692,
      "step": 6176
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4469160768452982e-05,
      "loss": 3.1542,
      "step": 6177
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4440271558572876e-05,
      "loss": 3.1115,
      "step": 6178
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4411382348692767e-05,
      "loss": 3.1023,
      "step": 6179
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4382493138812654e-05,
      "loss": 3.1114,
      "step": 6180
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4353603928932544e-05,
      "loss": 3.1446,
      "step": 6181
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4324714719052435e-05,
      "loss": 3.0185,
      "step": 6182
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4295825509172325e-05,
      "loss": 2.9706,
      "step": 6183
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4266936299292216e-05,
      "loss": 3.0785,
      "step": 6184
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4238047089412106e-05,
      "loss": 2.9412,
      "step": 6185
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4209157879531996e-05,
      "loss": 3.1385,
      "step": 6186
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4180268669651887e-05,
      "loss": 3.2132,
      "step": 6187
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4151379459771774e-05,
      "loss": 3.1793,
      "step": 6188
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4122490249891668e-05,
      "loss": 3.1368,
      "step": 6189
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4093601040011558e-05,
      "loss": 2.9862,
      "step": 6190
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.406471183013145e-05,
      "loss": 3.1003,
      "step": 6191
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4035822620251336e-05,
      "loss": 3.0735,
      "step": 6192
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.4006933410371226e-05,
      "loss": 2.8659,
      "step": 6193
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3978044200491117e-05,
      "loss": 3.0916,
      "step": 6194
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3949154990611007e-05,
      "loss": 3.1532,
      "step": 6195
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.39202657807309e-05,
      "loss": 3.1385,
      "step": 6196
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3891376570850788e-05,
      "loss": 2.9107,
      "step": 6197
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3862487360970678e-05,
      "loss": 3.0304,
      "step": 6198
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.383359815109057e-05,
      "loss": 3.1211,
      "step": 6199
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.380470894121046e-05,
      "loss": 3.0548,
      "step": 6200
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.377581973133035e-05,
      "loss": 3.1243,
      "step": 6201
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.374693052145024e-05,
      "loss": 3.1833,
      "step": 6202
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.371804131157013e-05,
      "loss": 2.9298,
      "step": 6203
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.368915210169002e-05,
      "loss": 3.1059,
      "step": 6204
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3660262891809908e-05,
      "loss": 3.1305,
      "step": 6205
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.36313736819298e-05,
      "loss": 3.0229,
      "step": 6206
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3602484472049692e-05,
      "loss": 3.1692,
      "step": 6207
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3573595262169583e-05,
      "loss": 3.1101,
      "step": 6208
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.354470605228947e-05,
      "loss": 2.9789,
      "step": 6209
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.351581684240936e-05,
      "loss": 3.1158,
      "step": 6210
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.348692763252925e-05,
      "loss": 3.0718,
      "step": 6211
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.345803842264914e-05,
      "loss": 3.157,
      "step": 6212
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3429149212769035e-05,
      "loss": 3.1713,
      "step": 6213
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3400260002888922e-05,
      "loss": 2.9399,
      "step": 6214
    },
    {
      "epoch": 0.88,
      "learning_rate": 2.3371370793008812e-05,
      "loss": 3.1105,
      "step": 6215
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3342481583128703e-05,
      "loss": 3.0805,
      "step": 6216
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3313592373248593e-05,
      "loss": 3.1651,
      "step": 6217
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3284703163368484e-05,
      "loss": 2.9673,
      "step": 6218
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3255813953488374e-05,
      "loss": 3.0281,
      "step": 6219
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3226924743608265e-05,
      "loss": 3.0074,
      "step": 6220
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3198035533728155e-05,
      "loss": 3.175,
      "step": 6221
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3169146323848042e-05,
      "loss": 3.1241,
      "step": 6222
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3140257113967932e-05,
      "loss": 2.9221,
      "step": 6223
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3111367904087823e-05,
      "loss": 3.089,
      "step": 6224
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3082478694207717e-05,
      "loss": 3.1303,
      "step": 6225
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3053589484327604e-05,
      "loss": 3.0978,
      "step": 6226
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.3024700274447494e-05,
      "loss": 2.8762,
      "step": 6227
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2995811064567385e-05,
      "loss": 2.7943,
      "step": 6228
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2966921854687275e-05,
      "loss": 3.1311,
      "step": 6229
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2938032644807165e-05,
      "loss": 3.0437,
      "step": 6230
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2909143434927056e-05,
      "loss": 3.0388,
      "step": 6231
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2880254225046946e-05,
      "loss": 3.1408,
      "step": 6232
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2851365015166837e-05,
      "loss": 3.0528,
      "step": 6233
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2822475805286727e-05,
      "loss": 3.1269,
      "step": 6234
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2793586595406614e-05,
      "loss": 3.1778,
      "step": 6235
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2764697385526508e-05,
      "loss": 3.1633,
      "step": 6236
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.27358081756464e-05,
      "loss": 3.0909,
      "step": 6237
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.270691896576629e-05,
      "loss": 3.0567,
      "step": 6238
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2678029755886176e-05,
      "loss": 2.9684,
      "step": 6239
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2649140546006066e-05,
      "loss": 3.0475,
      "step": 6240
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2620251336125957e-05,
      "loss": 3.0568,
      "step": 6241
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.259136212624585e-05,
      "loss": 3.1187,
      "step": 6242
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2562472916365738e-05,
      "loss": 3.0751,
      "step": 6243
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2533583706485628e-05,
      "loss": 3.0804,
      "step": 6244
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.250469449660552e-05,
      "loss": 3.1316,
      "step": 6245
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.247580528672541e-05,
      "loss": 2.9817,
      "step": 6246
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.24469160768453e-05,
      "loss": 3.0022,
      "step": 6247
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.241802686696519e-05,
      "loss": 3.0777,
      "step": 6248
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.238913765708508e-05,
      "loss": 3.0788,
      "step": 6249
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.236024844720497e-05,
      "loss": 3.1099,
      "step": 6250
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.233135923732486e-05,
      "loss": 3.0913,
      "step": 6251
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.230247002744475e-05,
      "loss": 3.0541,
      "step": 6252
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.227358081756464e-05,
      "loss": 2.9763,
      "step": 6253
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2244691607684533e-05,
      "loss": 3.0568,
      "step": 6254
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2215802397804423e-05,
      "loss": 3.2268,
      "step": 6255
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.218691318792431e-05,
      "loss": 3.0721,
      "step": 6256
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.21580239780442e-05,
      "loss": 3.1068,
      "step": 6257
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.212913476816409e-05,
      "loss": 3.0956,
      "step": 6258
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.210024555828398e-05,
      "loss": 2.968,
      "step": 6259
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2071356348403872e-05,
      "loss": 3.0614,
      "step": 6260
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2042467138523762e-05,
      "loss": 2.9459,
      "step": 6261
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.2013577928643653e-05,
      "loss": 3.0805,
      "step": 6262
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1984688718763543e-05,
      "loss": 2.9983,
      "step": 6263
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.195579950888343e-05,
      "loss": 3.1816,
      "step": 6264
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1926910299003324e-05,
      "loss": 3.1293,
      "step": 6265
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1898021089123214e-05,
      "loss": 3.0368,
      "step": 6266
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1869131879243105e-05,
      "loss": 3.0531,
      "step": 6267
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1840242669362995e-05,
      "loss": 3.1396,
      "step": 6268
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1811353459482882e-05,
      "loss": 2.9849,
      "step": 6269
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1782464249602773e-05,
      "loss": 3.1006,
      "step": 6270
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1753575039722667e-05,
      "loss": 3.1267,
      "step": 6271
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1724685829842557e-05,
      "loss": 3.1102,
      "step": 6272
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1695796619962444e-05,
      "loss": 3.1123,
      "step": 6273
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1666907410082335e-05,
      "loss": 3.0632,
      "step": 6274
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1638018200202225e-05,
      "loss": 3.1324,
      "step": 6275
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1609128990322115e-05,
      "loss": 2.9165,
      "step": 6276
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1580239780442006e-05,
      "loss": 3.1479,
      "step": 6277
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1551350570561896e-05,
      "loss": 3.0407,
      "step": 6278
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1522461360681787e-05,
      "loss": 3.1366,
      "step": 6279
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1493572150801677e-05,
      "loss": 3.132,
      "step": 6280
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1464682940921564e-05,
      "loss": 3.1849,
      "step": 6281
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1435793731041455e-05,
      "loss": 3.0809,
      "step": 6282
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.140690452116135e-05,
      "loss": 3.1511,
      "step": 6283
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.137801531128124e-05,
      "loss": 3.1877,
      "step": 6284
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.134912610140113e-05,
      "loss": 3.0118,
      "step": 6285
    },
    {
      "epoch": 0.89,
      "learning_rate": 2.1320236891521016e-05,
      "loss": 3.0608,
      "step": 6286
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.1291347681640907e-05,
      "loss": 2.9224,
      "step": 6287
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.1262458471760797e-05,
      "loss": 3.0317,
      "step": 6288
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.123356926188069e-05,
      "loss": 3.1024,
      "step": 6289
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.1204680052000578e-05,
      "loss": 2.9715,
      "step": 6290
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.117579084212047e-05,
      "loss": 3.0075,
      "step": 6291
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.114690163224036e-05,
      "loss": 3.0364,
      "step": 6292
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.111801242236025e-05,
      "loss": 3.0769,
      "step": 6293
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.108912321248014e-05,
      "loss": 2.9219,
      "step": 6294
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.106023400260003e-05,
      "loss": 3.1058,
      "step": 6295
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.103134479271992e-05,
      "loss": 3.0322,
      "step": 6296
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.100245558283981e-05,
      "loss": 2.9246,
      "step": 6297
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0973566372959698e-05,
      "loss": 3.0454,
      "step": 6298
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.094467716307959e-05,
      "loss": 3.1417,
      "step": 6299
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0915787953199483e-05,
      "loss": 2.9522,
      "step": 6300
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0886898743319373e-05,
      "loss": 3.0798,
      "step": 6301
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0858009533439263e-05,
      "loss": 3.0087,
      "step": 6302
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.082912032355915e-05,
      "loss": 3.03,
      "step": 6303
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.080023111367904e-05,
      "loss": 3.1378,
      "step": 6304
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.077134190379893e-05,
      "loss": 3.0412,
      "step": 6305
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0742452693918825e-05,
      "loss": 3.1259,
      "step": 6306
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0713563484038712e-05,
      "loss": 2.9775,
      "step": 6307
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0684674274158603e-05,
      "loss": 2.9517,
      "step": 6308
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0655785064278493e-05,
      "loss": 3.0625,
      "step": 6309
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0626895854398384e-05,
      "loss": 2.9316,
      "step": 6310
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0598006644518274e-05,
      "loss": 3.1149,
      "step": 6311
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0569117434638164e-05,
      "loss": 3.2036,
      "step": 6312
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0540228224758055e-05,
      "loss": 3.1569,
      "step": 6313
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0511339014877945e-05,
      "loss": 3.1524,
      "step": 6314
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0482449804997832e-05,
      "loss": 3.0428,
      "step": 6315
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0453560595117723e-05,
      "loss": 3.0148,
      "step": 6316
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0424671385237613e-05,
      "loss": 3.051,
      "step": 6317
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0395782175357507e-05,
      "loss": 3.0329,
      "step": 6318
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0366892965477394e-05,
      "loss": 2.9549,
      "step": 6319
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0338003755597285e-05,
      "loss": 3.1519,
      "step": 6320
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0309114545717175e-05,
      "loss": 2.9162,
      "step": 6321
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0280225335837065e-05,
      "loss": 3.012,
      "step": 6322
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0251336125956956e-05,
      "loss": 2.8915,
      "step": 6323
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0222446916076846e-05,
      "loss": 3.0236,
      "step": 6324
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0193557706196737e-05,
      "loss": 3.0933,
      "step": 6325
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0164668496316627e-05,
      "loss": 3.1941,
      "step": 6326
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0135779286436518e-05,
      "loss": 3.1418,
      "step": 6327
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.0106890076556405e-05,
      "loss": 3.0804,
      "step": 6328
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.00780008666763e-05,
      "loss": 3.0721,
      "step": 6329
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.004911165679619e-05,
      "loss": 3.2047,
      "step": 6330
    },
    {
      "epoch": 0.9,
      "learning_rate": 2.002022244691608e-05,
      "loss": 3.0389,
      "step": 6331
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9991333237035966e-05,
      "loss": 3.1679,
      "step": 6332
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9962444027155857e-05,
      "loss": 3.0885,
      "step": 6333
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9933554817275747e-05,
      "loss": 3.0729,
      "step": 6334
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.990466560739564e-05,
      "loss": 2.9343,
      "step": 6335
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9875776397515528e-05,
      "loss": 3.0759,
      "step": 6336
    },
    {
      "epoch": 0.9,
      "eval_loss": 3.3094747066497803,
      "eval_runtime": 472.097,
      "eval_samples_per_second": 43.396,
      "eval_steps_per_second": 14.465,
      "step": 6336
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.984688718763542e-05,
      "loss": 3.1271,
      "step": 6337
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.981799797775531e-05,
      "loss": 2.8808,
      "step": 6338
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.97891087678752e-05,
      "loss": 2.9369,
      "step": 6339
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.976021955799509e-05,
      "loss": 3.0485,
      "step": 6340
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.973133034811498e-05,
      "loss": 3.0099,
      "step": 6341
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.970244113823487e-05,
      "loss": 3.0134,
      "step": 6342
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.967355192835476e-05,
      "loss": 3.1073,
      "step": 6343
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.964466271847465e-05,
      "loss": 3.1248,
      "step": 6344
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.961577350859454e-05,
      "loss": 2.9222,
      "step": 6345
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.958688429871443e-05,
      "loss": 3.0365,
      "step": 6346
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9557995088834323e-05,
      "loss": 3.0688,
      "step": 6347
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9529105878954213e-05,
      "loss": 3.073,
      "step": 6348
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.95002166690741e-05,
      "loss": 2.8182,
      "step": 6349
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.947132745919399e-05,
      "loss": 3.1043,
      "step": 6350
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.944243824931388e-05,
      "loss": 3.1836,
      "step": 6351
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9413549039433772e-05,
      "loss": 3.1243,
      "step": 6352
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9384659829553662e-05,
      "loss": 3.1342,
      "step": 6353
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9355770619673553e-05,
      "loss": 3.1523,
      "step": 6354
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9326881409793443e-05,
      "loss": 3.1038,
      "step": 6355
    },
    {
      "epoch": 0.9,
      "learning_rate": 1.9297992199913334e-05,
      "loss": 2.985,
      "step": 6356
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9269102990033224e-05,
      "loss": 2.9722,
      "step": 6357
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9240213780153114e-05,
      "loss": 3.0341,
      "step": 6358
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9211324570273005e-05,
      "loss": 3.0669,
      "step": 6359
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9182435360392895e-05,
      "loss": 2.9711,
      "step": 6360
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9153546150512786e-05,
      "loss": 3.0699,
      "step": 6361
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9124656940632673e-05,
      "loss": 2.9977,
      "step": 6362
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9095767730752563e-05,
      "loss": 3.0904,
      "step": 6363
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9066878520872457e-05,
      "loss": 3.2007,
      "step": 6364
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9037989310992347e-05,
      "loss": 3.0675,
      "step": 6365
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.9009100101112234e-05,
      "loss": 2.9933,
      "step": 6366
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8980210891232125e-05,
      "loss": 3.0551,
      "step": 6367
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8951321681352015e-05,
      "loss": 3.1746,
      "step": 6368
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8922432471471906e-05,
      "loss": 2.9598,
      "step": 6369
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8893543261591796e-05,
      "loss": 2.9074,
      "step": 6370
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8864654051711687e-05,
      "loss": 3.0811,
      "step": 6371
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8835764841831577e-05,
      "loss": 2.9267,
      "step": 6372
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8806875631951468e-05,
      "loss": 2.8797,
      "step": 6373
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8777986422071358e-05,
      "loss": 2.917,
      "step": 6374
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8749097212191245e-05,
      "loss": 3.0835,
      "step": 6375
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.872020800231114e-05,
      "loss": 2.9676,
      "step": 6376
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.869131879243103e-05,
      "loss": 3.1162,
      "step": 6377
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.866242958255092e-05,
      "loss": 3.0682,
      "step": 6378
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8633540372670807e-05,
      "loss": 3.166,
      "step": 6379
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8604651162790697e-05,
      "loss": 3.1365,
      "step": 6380
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8575761952910588e-05,
      "loss": 3.1739,
      "step": 6381
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.854687274303048e-05,
      "loss": 3.0823,
      "step": 6382
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.851798353315037e-05,
      "loss": 3.1357,
      "step": 6383
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.848909432327026e-05,
      "loss": 3.0598,
      "step": 6384
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.846020511339015e-05,
      "loss": 2.9985,
      "step": 6385
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.843131590351004e-05,
      "loss": 3.013,
      "step": 6386
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.840242669362993e-05,
      "loss": 3.0844,
      "step": 6387
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.837353748374982e-05,
      "loss": 2.892,
      "step": 6388
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.834464827386971e-05,
      "loss": 2.9809,
      "step": 6389
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.83157590639896e-05,
      "loss": 3.1594,
      "step": 6390
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.828686985410949e-05,
      "loss": 3.0109,
      "step": 6391
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.825798064422938e-05,
      "loss": 3.0524,
      "step": 6392
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8229091434349273e-05,
      "loss": 3.11,
      "step": 6393
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8200202224469163e-05,
      "loss": 3.0801,
      "step": 6394
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8171313014589054e-05,
      "loss": 3.0253,
      "step": 6395
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.814242380470894e-05,
      "loss": 3.1307,
      "step": 6396
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.811353459482883e-05,
      "loss": 3.0491,
      "step": 6397
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8084645384948722e-05,
      "loss": 2.8271,
      "step": 6398
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8055756175068612e-05,
      "loss": 3.1495,
      "step": 6399
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.8026866965188503e-05,
      "loss": 3.1997,
      "step": 6400
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7997977755308393e-05,
      "loss": 3.0777,
      "step": 6401
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7969088545428283e-05,
      "loss": 3.0433,
      "step": 6402
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7940199335548174e-05,
      "loss": 2.9814,
      "step": 6403
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7911310125668064e-05,
      "loss": 3.0668,
      "step": 6404
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7882420915787955e-05,
      "loss": 3.0224,
      "step": 6405
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7853531705907845e-05,
      "loss": 2.9634,
      "step": 6406
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7824642496027736e-05,
      "loss": 3.064,
      "step": 6407
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7795753286147623e-05,
      "loss": 3.1138,
      "step": 6408
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7766864076267513e-05,
      "loss": 2.8335,
      "step": 6409
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7737974866387404e-05,
      "loss": 3.0095,
      "step": 6410
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7709085656507297e-05,
      "loss": 3.1095,
      "step": 6411
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7680196446627188e-05,
      "loss": 2.9441,
      "step": 6412
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7651307236747075e-05,
      "loss": 3.0799,
      "step": 6413
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7622418026866965e-05,
      "loss": 3.0277,
      "step": 6414
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7593528816986856e-05,
      "loss": 2.9749,
      "step": 6415
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7564639607106746e-05,
      "loss": 3.1017,
      "step": 6416
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7535750397226637e-05,
      "loss": 3.1977,
      "step": 6417
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7506861187346527e-05,
      "loss": 3.0746,
      "step": 6418
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7477971977466418e-05,
      "loss": 3.0608,
      "step": 6419
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7449082767586308e-05,
      "loss": 3.1119,
      "step": 6420
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7420193557706195e-05,
      "loss": 2.9672,
      "step": 6421
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.739130434782609e-05,
      "loss": 3.1024,
      "step": 6422
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.736241513794598e-05,
      "loss": 3.1323,
      "step": 6423
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.733352592806587e-05,
      "loss": 3.093,
      "step": 6424
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7304636718185757e-05,
      "loss": 3.0176,
      "step": 6425
    },
    {
      "epoch": 0.91,
      "learning_rate": 1.7275747508305647e-05,
      "loss": 3.1097,
      "step": 6426
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.7246858298425538e-05,
      "loss": 2.881,
      "step": 6427
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.721796908854543e-05,
      "loss": 3.1614,
      "step": 6428
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.7189079878665322e-05,
      "loss": 3.0169,
      "step": 6429
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.716019066878521e-05,
      "loss": 3.0446,
      "step": 6430
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.71313014589051e-05,
      "loss": 2.9543,
      "step": 6431
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.710241224902499e-05,
      "loss": 3.0359,
      "step": 6432
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.707352303914488e-05,
      "loss": 3.0817,
      "step": 6433
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.704463382926477e-05,
      "loss": 3.0606,
      "step": 6434
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.701574461938466e-05,
      "loss": 3.0164,
      "step": 6435
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.698685540950455e-05,
      "loss": 3.1532,
      "step": 6436
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6957966199624442e-05,
      "loss": 2.9962,
      "step": 6437
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.692907698974433e-05,
      "loss": 3.1381,
      "step": 6438
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.690018777986422e-05,
      "loss": 3.0888,
      "step": 6439
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6871298569984113e-05,
      "loss": 3.0824,
      "step": 6440
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6842409360104004e-05,
      "loss": 2.7761,
      "step": 6441
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.681352015022389e-05,
      "loss": 3.1271,
      "step": 6442
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.678463094034378e-05,
      "loss": 3.1044,
      "step": 6443
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.675574173046367e-05,
      "loss": 3.0952,
      "step": 6444
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6726852520583562e-05,
      "loss": 2.8992,
      "step": 6445
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6697963310703456e-05,
      "loss": 3.1165,
      "step": 6446
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6669074100823343e-05,
      "loss": 3.2024,
      "step": 6447
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6640184890943233e-05,
      "loss": 2.7159,
      "step": 6448
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6611295681063124e-05,
      "loss": 3.1082,
      "step": 6449
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6582406471183014e-05,
      "loss": 3.1386,
      "step": 6450
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6553517261302905e-05,
      "loss": 3.0407,
      "step": 6451
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6524628051422795e-05,
      "loss": 3.0281,
      "step": 6452
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6495738841542686e-05,
      "loss": 3.0393,
      "step": 6453
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6466849631662576e-05,
      "loss": 3.0324,
      "step": 6454
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6437960421782463e-05,
      "loss": 2.9705,
      "step": 6455
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6409071211902354e-05,
      "loss": 2.9463,
      "step": 6456
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6380182002022247e-05,
      "loss": 3.0211,
      "step": 6457
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6351292792142138e-05,
      "loss": 2.9502,
      "step": 6458
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6322403582262025e-05,
      "loss": 2.9689,
      "step": 6459
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6293514372381915e-05,
      "loss": 3.1397,
      "step": 6460
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6264625162501806e-05,
      "loss": 3.0759,
      "step": 6461
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6235735952621696e-05,
      "loss": 2.9888,
      "step": 6462
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6206846742741587e-05,
      "loss": 3.0348,
      "step": 6463
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6177957532861477e-05,
      "loss": 3.0546,
      "step": 6464
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6149068322981367e-05,
      "loss": 3.0652,
      "step": 6465
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6120179113101258e-05,
      "loss": 2.9981,
      "step": 6466
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.609128990322115e-05,
      "loss": 3.0046,
      "step": 6467
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.6062400693341035e-05,
      "loss": 2.9306,
      "step": 6468
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.603351148346093e-05,
      "loss": 3.0131,
      "step": 6469
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.600462227358082e-05,
      "loss": 3.0597,
      "step": 6470
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.597573306370071e-05,
      "loss": 2.9273,
      "step": 6471
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5946843853820597e-05,
      "loss": 3.0676,
      "step": 6472
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5917954643940488e-05,
      "loss": 3.0043,
      "step": 6473
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5889065434060378e-05,
      "loss": 3.0176,
      "step": 6474
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5860176224180272e-05,
      "loss": 3.1659,
      "step": 6475
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.583128701430016e-05,
      "loss": 3.2128,
      "step": 6476
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.580239780442005e-05,
      "loss": 3.0655,
      "step": 6477
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.577350859453994e-05,
      "loss": 2.962,
      "step": 6478
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.574461938465983e-05,
      "loss": 3.2248,
      "step": 6479
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.571573017477972e-05,
      "loss": 3.1467,
      "step": 6480
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.568684096489961e-05,
      "loss": 2.7129,
      "step": 6481
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.56579517550195e-05,
      "loss": 3.1727,
      "step": 6482
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5629062545139392e-05,
      "loss": 2.9672,
      "step": 6483
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5600173335259282e-05,
      "loss": 3.0074,
      "step": 6484
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.557128412537917e-05,
      "loss": 3.0807,
      "step": 6485
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5542394915499063e-05,
      "loss": 3.1327,
      "step": 6486
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5513505705618954e-05,
      "loss": 2.9074,
      "step": 6487
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5484616495738844e-05,
      "loss": 3.0862,
      "step": 6488
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.545572728585873e-05,
      "loss": 3.0671,
      "step": 6489
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.542683807597862e-05,
      "loss": 3.0044,
      "step": 6490
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5397948866098512e-05,
      "loss": 3.1161,
      "step": 6491
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5369059656218402e-05,
      "loss": 2.9369,
      "step": 6492
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5340170446338293e-05,
      "loss": 3.1136,
      "step": 6493
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5311281236458183e-05,
      "loss": 3.0749,
      "step": 6494
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5282392026578074e-05,
      "loss": 3.1212,
      "step": 6495
    },
    {
      "epoch": 0.92,
      "learning_rate": 1.5253502816697964e-05,
      "loss": 3.0919,
      "step": 6496
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5224613606817853e-05,
      "loss": 3.0911,
      "step": 6497
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5195724396937743e-05,
      "loss": 3.2119,
      "step": 6498
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5166835187057636e-05,
      "loss": 3.1722,
      "step": 6499
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5137945977177526e-05,
      "loss": 3.1116,
      "step": 6500
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5109056767297416e-05,
      "loss": 2.9683,
      "step": 6501
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5080167557417305e-05,
      "loss": 3.0567,
      "step": 6502
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5051278347537196e-05,
      "loss": 3.0653,
      "step": 6503
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.5022389137657086e-05,
      "loss": 2.9544,
      "step": 6504
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4993499927776976e-05,
      "loss": 3.0834,
      "step": 6505
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4964610717896865e-05,
      "loss": 3.1934,
      "step": 6506
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4935721508016756e-05,
      "loss": 3.0452,
      "step": 6507
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4906832298136648e-05,
      "loss": 3.1946,
      "step": 6508
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4877943088256538e-05,
      "loss": 2.9808,
      "step": 6509
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4849053878376425e-05,
      "loss": 3.0162,
      "step": 6510
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4820164668496317e-05,
      "loss": 3.0212,
      "step": 6511
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4791275458616208e-05,
      "loss": 2.8753,
      "step": 6512
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4762386248736098e-05,
      "loss": 2.995,
      "step": 6513
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4733497038855987e-05,
      "loss": 2.9755,
      "step": 6514
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4704607828975877e-05,
      "loss": 2.9062,
      "step": 6515
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4675718619095768e-05,
      "loss": 3.1349,
      "step": 6516
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.464682940921566e-05,
      "loss": 2.9564,
      "step": 6517
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.461794019933555e-05,
      "loss": 3.0887,
      "step": 6518
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4589050989455438e-05,
      "loss": 3.1461,
      "step": 6519
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.456016177957533e-05,
      "loss": 2.9868,
      "step": 6520
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.453127256969522e-05,
      "loss": 2.8988,
      "step": 6521
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.450238335981511e-05,
      "loss": 2.965,
      "step": 6522
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4473494149935e-05,
      "loss": 2.9583,
      "step": 6523
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.444460494005489e-05,
      "loss": 3.0323,
      "step": 6524
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.441571573017478e-05,
      "loss": 3.0841,
      "step": 6525
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4386826520294672e-05,
      "loss": 3.0853,
      "step": 6526
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.435793731041456e-05,
      "loss": 3.1281,
      "step": 6527
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4329048100534451e-05,
      "loss": 2.9758,
      "step": 6528
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4300158890654342e-05,
      "loss": 3.1012,
      "step": 6529
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4271269680774232e-05,
      "loss": 3.1736,
      "step": 6530
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4242380470894121e-05,
      "loss": 2.9168,
      "step": 6531
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4213491261014012e-05,
      "loss": 2.9395,
      "step": 6532
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4184602051133902e-05,
      "loss": 3.179,
      "step": 6533
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4155712841253792e-05,
      "loss": 3.0333,
      "step": 6534
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4126823631373681e-05,
      "loss": 3.0926,
      "step": 6535
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4097934421493572e-05,
      "loss": 3.0646,
      "step": 6536
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4069045211613464e-05,
      "loss": 3.0286,
      "step": 6537
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4040156001733354e-05,
      "loss": 3.095,
      "step": 6538
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.4011266791853245e-05,
      "loss": 3.0268,
      "step": 6539
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3982377581973133e-05,
      "loss": 3.0744,
      "step": 6540
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3953488372093024e-05,
      "loss": 3.1404,
      "step": 6541
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3924599162212914e-05,
      "loss": 3.1033,
      "step": 6542
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3895709952332805e-05,
      "loss": 3.0797,
      "step": 6543
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3866820742452693e-05,
      "loss": 3.1705,
      "step": 6544
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3837931532572584e-05,
      "loss": 3.0224,
      "step": 6545
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3809042322692476e-05,
      "loss": 2.8371,
      "step": 6546
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3780153112812366e-05,
      "loss": 2.915,
      "step": 6547
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3751263902932255e-05,
      "loss": 2.9611,
      "step": 6548
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3722374693052146e-05,
      "loss": 3.1396,
      "step": 6549
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3693485483172036e-05,
      "loss": 2.9345,
      "step": 6550
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3664596273291926e-05,
      "loss": 3.0005,
      "step": 6551
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3635707063411815e-05,
      "loss": 2.9932,
      "step": 6552
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3606817853531706e-05,
      "loss": 3.0044,
      "step": 6553
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3577928643651596e-05,
      "loss": 2.9093,
      "step": 6554
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3549039433771488e-05,
      "loss": 3.0998,
      "step": 6555
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3520150223891379e-05,
      "loss": 3.0678,
      "step": 6556
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3491261014011267e-05,
      "loss": 3.1757,
      "step": 6557
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3462371804131158e-05,
      "loss": 2.8707,
      "step": 6558
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3433482594251048e-05,
      "loss": 3.0569,
      "step": 6559
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3404593384370939e-05,
      "loss": 3.0525,
      "step": 6560
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3375704174490827e-05,
      "loss": 3.019,
      "step": 6561
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3346814964610718e-05,
      "loss": 3.0802,
      "step": 6562
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3317925754730608e-05,
      "loss": 3.1213,
      "step": 6563
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.32890365448505e-05,
      "loss": 3.1184,
      "step": 6564
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.3260147334970387e-05,
      "loss": 3.0327,
      "step": 6565
    },
    {
      "epoch": 0.93,
      "learning_rate": 1.323125812509028e-05,
      "loss": 3.0315,
      "step": 6566
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.320236891521017e-05,
      "loss": 3.0842,
      "step": 6567
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.317347970533006e-05,
      "loss": 3.0889,
      "step": 6568
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.314459049544995e-05,
      "loss": 3.0679,
      "step": 6569
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.311570128556984e-05,
      "loss": 3.0597,
      "step": 6570
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.308681207568973e-05,
      "loss": 2.9619,
      "step": 6571
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.3057922865809622e-05,
      "loss": 3.0097,
      "step": 6572
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.3029033655929513e-05,
      "loss": 3.1552,
      "step": 6573
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.30001444460494e-05,
      "loss": 3.0397,
      "step": 6574
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2971255236169292e-05,
      "loss": 3.1342,
      "step": 6575
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2942366026289182e-05,
      "loss": 3.0991,
      "step": 6576
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2913476816409073e-05,
      "loss": 3.0415,
      "step": 6577
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2884587606528961e-05,
      "loss": 3.0041,
      "step": 6578
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2855698396648852e-05,
      "loss": 2.9867,
      "step": 6579
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2826809186768742e-05,
      "loss": 2.9954,
      "step": 6580
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2797919976888634e-05,
      "loss": 3.0415,
      "step": 6581
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2769030767008522e-05,
      "loss": 3.0023,
      "step": 6582
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2740141557128412e-05,
      "loss": 2.8956,
      "step": 6583
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2711252347248304e-05,
      "loss": 2.8938,
      "step": 6584
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2682363137368195e-05,
      "loss": 3.0022,
      "step": 6585
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2653473927488083e-05,
      "loss": 3.0178,
      "step": 6586
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2624584717607974e-05,
      "loss": 3.0648,
      "step": 6587
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2595695507727864e-05,
      "loss": 3.0671,
      "step": 6588
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2566806297847755e-05,
      "loss": 3.0802,
      "step": 6589
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2537917087967643e-05,
      "loss": 3.0391,
      "step": 6590
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2509027878087534e-05,
      "loss": 3.1359,
      "step": 6591
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2480138668207424e-05,
      "loss": 2.9749,
      "step": 6592
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2451249458327316e-05,
      "loss": 3.0734,
      "step": 6593
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2422360248447205e-05,
      "loss": 3.0312,
      "step": 6594
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2393471038567096e-05,
      "loss": 3.1918,
      "step": 6595
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2364581828686986e-05,
      "loss": 2.9507,
      "step": 6596
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2335692618806876e-05,
      "loss": 3.1098,
      "step": 6597
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2306803408926767e-05,
      "loss": 2.9422,
      "step": 6598
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2277914199046657e-05,
      "loss": 3.2444,
      "step": 6599
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2249024989166546e-05,
      "loss": 3.0561,
      "step": 6600
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2220135779286438e-05,
      "loss": 3.0919,
      "step": 6601
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2191246569406327e-05,
      "loss": 2.9444,
      "step": 6602
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2162357359526217e-05,
      "loss": 3.1651,
      "step": 6603
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2133468149646108e-05,
      "loss": 3.0717,
      "step": 6604
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2104578939765998e-05,
      "loss": 2.861,
      "step": 6605
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2075689729885887e-05,
      "loss": 3.1206,
      "step": 6606
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2046800520005779e-05,
      "loss": 3.0615,
      "step": 6607
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.2017911310125668e-05,
      "loss": 3.1829,
      "step": 6608
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1989022100245558e-05,
      "loss": 3.0526,
      "step": 6609
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.196013289036545e-05,
      "loss": 3.0154,
      "step": 6610
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1931243680485339e-05,
      "loss": 3.0023,
      "step": 6611
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.190235447060523e-05,
      "loss": 2.9746,
      "step": 6612
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.187346526072512e-05,
      "loss": 3.0954,
      "step": 6613
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.184457605084501e-05,
      "loss": 3.0068,
      "step": 6614
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.18156868409649e-05,
      "loss": 3.1252,
      "step": 6615
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1786797631084791e-05,
      "loss": 3.038,
      "step": 6616
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.175790842120468e-05,
      "loss": 3.0386,
      "step": 6617
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.172901921132457e-05,
      "loss": 3.0446,
      "step": 6618
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1700130001444461e-05,
      "loss": 3.0548,
      "step": 6619
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1671240791564351e-05,
      "loss": 3.117,
      "step": 6620
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1642351581684242e-05,
      "loss": 2.9915,
      "step": 6621
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1613462371804132e-05,
      "loss": 3.1946,
      "step": 6622
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1584573161924021e-05,
      "loss": 2.9643,
      "step": 6623
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1555683952043911e-05,
      "loss": 3.0892,
      "step": 6624
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1526794742163802e-05,
      "loss": 2.959,
      "step": 6625
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1497905532283692e-05,
      "loss": 2.9867,
      "step": 6626
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1469016322403583e-05,
      "loss": 3.0531,
      "step": 6627
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1440127112523473e-05,
      "loss": 3.1347,
      "step": 6628
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1411237902643364e-05,
      "loss": 3.1322,
      "step": 6629
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1382348692763254e-05,
      "loss": 2.8694,
      "step": 6630
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1353459482883144e-05,
      "loss": 3.0304,
      "step": 6631
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1324570273003033e-05,
      "loss": 3.0735,
      "step": 6632
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1295681063122925e-05,
      "loss": 3.0781,
      "step": 6633
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1266791853242814e-05,
      "loss": 3.2113,
      "step": 6634
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1237902643362705e-05,
      "loss": 3.1034,
      "step": 6635
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1209013433482595e-05,
      "loss": 3.0906,
      "step": 6636
    },
    {
      "epoch": 0.94,
      "learning_rate": 1.1180124223602485e-05,
      "loss": 3.0979,
      "step": 6637
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1151235013722374e-05,
      "loss": 3.0023,
      "step": 6638
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1122345803842266e-05,
      "loss": 3.1516,
      "step": 6639
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1093456593962155e-05,
      "loss": 3.141,
      "step": 6640
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1064567384082045e-05,
      "loss": 3.0904,
      "step": 6641
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1035678174201936e-05,
      "loss": 3.1283,
      "step": 6642
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.1006788964321826e-05,
      "loss": 2.9994,
      "step": 6643
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0977899754441715e-05,
      "loss": 3.1035,
      "step": 6644
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0949010544561607e-05,
      "loss": 3.1685,
      "step": 6645
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0920121334681498e-05,
      "loss": 2.9159,
      "step": 6646
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0891232124801386e-05,
      "loss": 3.0004,
      "step": 6647
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0862342914921279e-05,
      "loss": 3.0733,
      "step": 6648
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0833453705041167e-05,
      "loss": 2.9666,
      "step": 6649
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0804564495161058e-05,
      "loss": 3.0647,
      "step": 6650
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0775675285280948e-05,
      "loss": 3.0639,
      "step": 6651
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0746786075400839e-05,
      "loss": 3.1523,
      "step": 6652
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0717896865520727e-05,
      "loss": 3.0446,
      "step": 6653
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.068900765564062e-05,
      "loss": 2.9252,
      "step": 6654
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0660118445760508e-05,
      "loss": 3.0308,
      "step": 6655
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0631229235880399e-05,
      "loss": 3.2152,
      "step": 6656
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0602340026000289e-05,
      "loss": 3.0464,
      "step": 6657
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.057345081612018e-05,
      "loss": 3.028,
      "step": 6658
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.054456160624007e-05,
      "loss": 3.1068,
      "step": 6659
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.051567239635996e-05,
      "loss": 2.9705,
      "step": 6660
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0486783186479849e-05,
      "loss": 2.9226,
      "step": 6661
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0457893976599741e-05,
      "loss": 2.8683,
      "step": 6662
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0429004766719632e-05,
      "loss": 3.086,
      "step": 6663
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.040011555683952e-05,
      "loss": 2.85,
      "step": 6664
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0371226346959413e-05,
      "loss": 3.0683,
      "step": 6665
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0342337137079301e-05,
      "loss": 3.0026,
      "step": 6666
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0313447927199192e-05,
      "loss": 3.0649,
      "step": 6667
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0284558717319082e-05,
      "loss": 3.0641,
      "step": 6668
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0255669507438973e-05,
      "loss": 3.0668,
      "step": 6669
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0226780297558861e-05,
      "loss": 3.0345,
      "step": 6670
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0197891087678754e-05,
      "loss": 3.0709,
      "step": 6671
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0169001877798642e-05,
      "loss": 2.8739,
      "step": 6672
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0140112667918533e-05,
      "loss": 3.0179,
      "step": 6673
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0111223458038423e-05,
      "loss": 3.0285,
      "step": 6674
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0082334248158314e-05,
      "loss": 3.0476,
      "step": 6675
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0053445038278202e-05,
      "loss": 3.0647,
      "step": 6676
    },
    {
      "epoch": 0.95,
      "learning_rate": 1.0024555828398094e-05,
      "loss": 3.0383,
      "step": 6677
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.995666618517983e-06,
      "loss": 3.0514,
      "step": 6678
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.966777408637874e-06,
      "loss": 3.198,
      "step": 6679
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.937888198757764e-06,
      "loss": 3.1583,
      "step": 6680
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.908998988877655e-06,
      "loss": 2.9471,
      "step": 6681
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.880109778997545e-06,
      "loss": 3.0897,
      "step": 6682
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.851220569117435e-06,
      "loss": 3.1209,
      "step": 6683
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.822331359237326e-06,
      "loss": 2.9804,
      "step": 6684
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.793442149357215e-06,
      "loss": 2.9855,
      "step": 6685
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.764552939477107e-06,
      "loss": 3.0023,
      "step": 6686
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.735663729596995e-06,
      "loss": 3.104,
      "step": 6687
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.706774519716886e-06,
      "loss": 3.1989,
      "step": 6688
    },
    {
      "epoch": 0.95,
      "eval_loss": 3.2940800189971924,
      "eval_runtime": 471.7893,
      "eval_samples_per_second": 43.424,
      "eval_steps_per_second": 14.475,
      "step": 6688
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.677885309836776e-06,
      "loss": 3.0624,
      "step": 6689
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.648996099956667e-06,
      "loss": 3.0859,
      "step": 6690
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.620106890076557e-06,
      "loss": 3.1474,
      "step": 6691
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.591217680196448e-06,
      "loss": 2.9431,
      "step": 6692
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.562328470316336e-06,
      "loss": 3.0516,
      "step": 6693
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.533439260436229e-06,
      "loss": 3.0875,
      "step": 6694
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.504550050556117e-06,
      "loss": 3.0427,
      "step": 6695
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.475660840676008e-06,
      "loss": 2.9797,
      "step": 6696
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.446771630795898e-06,
      "loss": 3.033,
      "step": 6697
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.417882420915789e-06,
      "loss": 2.9754,
      "step": 6698
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.388993211035679e-06,
      "loss": 2.9346,
      "step": 6699
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.36010400115557e-06,
      "loss": 3.1591,
      "step": 6700
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.33121479127546e-06,
      "loss": 3.022,
      "step": 6701
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.302325581395349e-06,
      "loss": 3.0103,
      "step": 6702
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.27343637151524e-06,
      "loss": 3.0677,
      "step": 6703
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.24454716163513e-06,
      "loss": 3.1395,
      "step": 6704
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.21565795175502e-06,
      "loss": 2.9199,
      "step": 6705
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.18676874187491e-06,
      "loss": 3.1351,
      "step": 6706
    },
    {
      "epoch": 0.95,
      "learning_rate": 9.1578795319948e-06,
      "loss": 3.0665,
      "step": 6707
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.12899032211469e-06,
      "loss": 3.0258,
      "step": 6708
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.100101112234582e-06,
      "loss": 3.0946,
      "step": 6709
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.07121190235447e-06,
      "loss": 3.0898,
      "step": 6710
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.042322692474361e-06,
      "loss": 2.9603,
      "step": 6711
    },
    {
      "epoch": 0.96,
      "learning_rate": 9.013433482594251e-06,
      "loss": 3.1076,
      "step": 6712
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.984544272714142e-06,
      "loss": 3.1297,
      "step": 6713
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.955655062834032e-06,
      "loss": 3.1439,
      "step": 6714
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.926765852953923e-06,
      "loss": 2.9976,
      "step": 6715
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.897876643073811e-06,
      "loss": 3.1051,
      "step": 6716
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.868987433193702e-06,
      "loss": 2.979,
      "step": 6717
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.840098223313594e-06,
      "loss": 3.129,
      "step": 6718
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.811209013433483e-06,
      "loss": 2.9428,
      "step": 6719
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.782319803553373e-06,
      "loss": 3.0957,
      "step": 6720
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.753430593673264e-06,
      "loss": 3.0369,
      "step": 6721
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.724541383793154e-06,
      "loss": 3.0829,
      "step": 6722
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.695652173913044e-06,
      "loss": 3.1139,
      "step": 6723
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.666762964032935e-06,
      "loss": 2.9817,
      "step": 6724
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.637873754152824e-06,
      "loss": 3.0779,
      "step": 6725
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.608984544272716e-06,
      "loss": 3.1538,
      "step": 6726
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.580095334392604e-06,
      "loss": 3.0275,
      "step": 6727
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.551206124512495e-06,
      "loss": 2.9255,
      "step": 6728
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.522316914632385e-06,
      "loss": 2.8025,
      "step": 6729
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.493427704752276e-06,
      "loss": 3.1091,
      "step": 6730
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.464538494872165e-06,
      "loss": 3.1142,
      "step": 6731
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.435649284992057e-06,
      "loss": 3.0966,
      "step": 6732
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.406760075111945e-06,
      "loss": 3.0464,
      "step": 6733
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.377870865231836e-06,
      "loss": 3.0345,
      "step": 6734
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.348981655351728e-06,
      "loss": 3.0332,
      "step": 6735
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.320092445471617e-06,
      "loss": 2.8829,
      "step": 6736
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.291203235591507e-06,
      "loss": 3.0757,
      "step": 6737
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.262314025711398e-06,
      "loss": 3.1809,
      "step": 6738
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.233424815831288e-06,
      "loss": 3.0978,
      "step": 6739
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.204535605951177e-06,
      "loss": 3.0285,
      "step": 6740
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.175646396071069e-06,
      "loss": 3.2434,
      "step": 6741
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.146757186190958e-06,
      "loss": 2.9622,
      "step": 6742
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.117867976310848e-06,
      "loss": 2.9389,
      "step": 6743
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.088978766430739e-06,
      "loss": 3.2191,
      "step": 6744
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.060089556550629e-06,
      "loss": 3.0754,
      "step": 6745
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.031200346670518e-06,
      "loss": 3.0331,
      "step": 6746
    },
    {
      "epoch": 0.96,
      "learning_rate": 8.00231113679041e-06,
      "loss": 3.0158,
      "step": 6747
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.973421926910299e-06,
      "loss": 3.0167,
      "step": 6748
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.944532717030189e-06,
      "loss": 3.1016,
      "step": 6749
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.91564350715008e-06,
      "loss": 2.97,
      "step": 6750
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.88675429726997e-06,
      "loss": 3.0349,
      "step": 6751
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.85786508738986e-06,
      "loss": 3.0124,
      "step": 6752
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.82897587750975e-06,
      "loss": 3.0379,
      "step": 6753
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.800086667629641e-06,
      "loss": 3.1907,
      "step": 6754
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.771197457749532e-06,
      "loss": 3.0362,
      "step": 6755
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.742308247869422e-06,
      "loss": 3.059,
      "step": 6756
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.71341903798931e-06,
      "loss": 3.0152,
      "step": 6757
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.684529828109201e-06,
      "loss": 3.0157,
      "step": 6758
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.655640618229092e-06,
      "loss": 3.0653,
      "step": 6759
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.626751408348982e-06,
      "loss": 3.1621,
      "step": 6760
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.597862198468872e-06,
      "loss": 3.0576,
      "step": 6761
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.568972988588763e-06,
      "loss": 3.1017,
      "step": 6762
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.540083778708653e-06,
      "loss": 3.0208,
      "step": 6763
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.511194568828543e-06,
      "loss": 3.0832,
      "step": 6764
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.482305358948433e-06,
      "loss": 3.0535,
      "step": 6765
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.453416149068324e-06,
      "loss": 2.9272,
      "step": 6766
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.424526939188213e-06,
      "loss": 3.0445,
      "step": 6767
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.395637729308104e-06,
      "loss": 3.1716,
      "step": 6768
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.3667485194279935e-06,
      "loss": 2.9934,
      "step": 6769
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.337859309547884e-06,
      "loss": 3.1079,
      "step": 6770
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.308970099667775e-06,
      "loss": 2.8938,
      "step": 6771
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.280080889787665e-06,
      "loss": 2.8871,
      "step": 6772
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.251191679907555e-06,
      "loss": 2.9292,
      "step": 6773
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.222302470027445e-06,
      "loss": 3.0288,
      "step": 6774
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.193413260147336e-06,
      "loss": 3.0358,
      "step": 6775
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.164524050267226e-06,
      "loss": 3.0653,
      "step": 6776
    },
    {
      "epoch": 0.96,
      "learning_rate": 7.135634840387116e-06,
      "loss": 3.0944,
      "step": 6777
    },
    {
      "epoch": 0.97,
      "learning_rate": 7.106745630507006e-06,
      "loss": 3.1736,
      "step": 6778
    },
    {
      "epoch": 0.97,
      "learning_rate": 7.077856420626896e-06,
      "loss": 2.9908,
      "step": 6779
    },
    {
      "epoch": 0.97,
      "learning_rate": 7.048967210746786e-06,
      "loss": 2.9514,
      "step": 6780
    },
    {
      "epoch": 0.97,
      "learning_rate": 7.020078000866677e-06,
      "loss": 2.9189,
      "step": 6781
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.991188790986567e-06,
      "loss": 3.1035,
      "step": 6782
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.962299581106457e-06,
      "loss": 3.0636,
      "step": 6783
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.933410371226347e-06,
      "loss": 3.0794,
      "step": 6784
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.904521161346238e-06,
      "loss": 3.0011,
      "step": 6785
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.8756319514661276e-06,
      "loss": 3.1414,
      "step": 6786
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.846742741586018e-06,
      "loss": 3.1243,
      "step": 6787
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.817853531705908e-06,
      "loss": 2.9568,
      "step": 6788
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.788964321825798e-06,
      "loss": 3.0535,
      "step": 6789
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.760075111945689e-06,
      "loss": 3.0086,
      "step": 6790
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.731185902065579e-06,
      "loss": 3.0518,
      "step": 6791
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.702296692185469e-06,
      "loss": 2.9835,
      "step": 6792
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.673407482305359e-06,
      "loss": 2.9584,
      "step": 6793
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.64451827242525e-06,
      "loss": 3.036,
      "step": 6794
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.61562906254514e-06,
      "loss": 3.0023,
      "step": 6795
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.58673985266503e-06,
      "loss": 2.9843,
      "step": 6796
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.55785064278492e-06,
      "loss": 3.0756,
      "step": 6797
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.528961432904811e-06,
      "loss": 3.0288,
      "step": 6798
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.5000722230247e-06,
      "loss": 2.9967,
      "step": 6799
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.471183013144591e-06,
      "loss": 3.0806,
      "step": 6800
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.442293803264481e-06,
      "loss": 3.0906,
      "step": 6801
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.413404593384371e-06,
      "loss": 2.9722,
      "step": 6802
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.384515383504261e-06,
      "loss": 3.0042,
      "step": 6803
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.355626173624152e-06,
      "loss": 3.1041,
      "step": 6804
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.326736963744042e-06,
      "loss": 3.0505,
      "step": 6805
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.297847753863932e-06,
      "loss": 3.1167,
      "step": 6806
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.268958543983822e-06,
      "loss": 3.0737,
      "step": 6807
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.240069334103712e-06,
      "loss": 2.9312,
      "step": 6808
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.2111801242236025e-06,
      "loss": 3.0523,
      "step": 6809
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.182290914343493e-06,
      "loss": 3.1139,
      "step": 6810
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.153401704463383e-06,
      "loss": 3.1251,
      "step": 6811
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.124512494583273e-06,
      "loss": 3.1028,
      "step": 6812
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.0956232847031634e-06,
      "loss": 3.0303,
      "step": 6813
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.066734074823054e-06,
      "loss": 2.9433,
      "step": 6814
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.0378448649429435e-06,
      "loss": 3.0504,
      "step": 6815
    },
    {
      "epoch": 0.97,
      "learning_rate": 6.008955655062834e-06,
      "loss": 3.0348,
      "step": 6816
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.980066445182725e-06,
      "loss": 3.0943,
      "step": 6817
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.951177235302615e-06,
      "loss": 3.1233,
      "step": 6818
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.922288025422505e-06,
      "loss": 3.0396,
      "step": 6819
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.893398815542396e-06,
      "loss": 3.0299,
      "step": 6820
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.864509605662285e-06,
      "loss": 3.071,
      "step": 6821
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.835620395782176e-06,
      "loss": 3.1082,
      "step": 6822
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.806731185902066e-06,
      "loss": 3.0669,
      "step": 6823
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.777841976021956e-06,
      "loss": 2.9796,
      "step": 6824
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.748952766141846e-06,
      "loss": 2.9577,
      "step": 6825
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.720063556261737e-06,
      "loss": 3.0809,
      "step": 6826
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.691174346381627e-06,
      "loss": 3.0723,
      "step": 6827
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.662285136501517e-06,
      "loss": 2.9341,
      "step": 6828
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.633395926621407e-06,
      "loss": 2.9875,
      "step": 6829
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.6045067167412975e-06,
      "loss": 2.9204,
      "step": 6830
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.575617506861187e-06,
      "loss": 3.0588,
      "step": 6831
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.5467282969810775e-06,
      "loss": 2.9367,
      "step": 6832
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.517839087100968e-06,
      "loss": 3.0618,
      "step": 6833
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.4889498772208576e-06,
      "loss": 3.0858,
      "step": 6834
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.460060667340749e-06,
      "loss": 3.086,
      "step": 6835
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.431171457460639e-06,
      "loss": 2.8882,
      "step": 6836
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.402282247580529e-06,
      "loss": 3.1037,
      "step": 6837
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.373393037700419e-06,
      "loss": 2.9876,
      "step": 6838
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.34450382782031e-06,
      "loss": 2.9321,
      "step": 6839
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.315614617940199e-06,
      "loss": 3.0526,
      "step": 6840
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.28672540806009e-06,
      "loss": 2.7436,
      "step": 6841
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.25783619817998e-06,
      "loss": 2.9718,
      "step": 6842
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.228946988299871e-06,
      "loss": 3.1092,
      "step": 6843
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.20005777841976e-06,
      "loss": 2.9495,
      "step": 6844
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.171168568539651e-06,
      "loss": 2.9479,
      "step": 6845
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.142279358659541e-06,
      "loss": 3.1482,
      "step": 6846
    },
    {
      "epoch": 0.97,
      "learning_rate": 5.113390148779431e-06,
      "loss": 2.9749,
      "step": 6847
    },
    {
      "epoch": 0.98,
      "learning_rate": 5.084500938899321e-06,
      "loss": 3.0205,
      "step": 6848
    },
    {
      "epoch": 0.98,
      "learning_rate": 5.0556117290192116e-06,
      "loss": 2.9451,
      "step": 6849
    },
    {
      "epoch": 0.98,
      "learning_rate": 5.026722519139101e-06,
      "loss": 3.0533,
      "step": 6850
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.997833309258992e-06,
      "loss": 3.0174,
      "step": 6851
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.968944099378882e-06,
      "loss": 2.9644,
      "step": 6852
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.9400548894987725e-06,
      "loss": 3.1064,
      "step": 6853
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.911165679618663e-06,
      "loss": 3.1061,
      "step": 6854
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.882276469738553e-06,
      "loss": 3.0388,
      "step": 6855
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.853387259858443e-06,
      "loss": 3.088,
      "step": 6856
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.824498049978333e-06,
      "loss": 2.6724,
      "step": 6857
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.795608840098224e-06,
      "loss": 3.093,
      "step": 6858
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.766719630218114e-06,
      "loss": 3.15,
      "step": 6859
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.737830420338004e-06,
      "loss": 3.016,
      "step": 6860
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.708941210457894e-06,
      "loss": 2.842,
      "step": 6861
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.680052000577785e-06,
      "loss": 3.1013,
      "step": 6862
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.651162790697674e-06,
      "loss": 3.0444,
      "step": 6863
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.622273580817565e-06,
      "loss": 3.098,
      "step": 6864
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.593384370937455e-06,
      "loss": 3.0278,
      "step": 6865
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.564495161057345e-06,
      "loss": 3.0869,
      "step": 6866
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.535605951177235e-06,
      "loss": 2.7475,
      "step": 6867
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.506716741297126e-06,
      "loss": 3.1009,
      "step": 6868
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.477827531417016e-06,
      "loss": 3.0051,
      "step": 6869
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.448938321536906e-06,
      "loss": 2.9816,
      "step": 6870
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.420049111656797e-06,
      "loss": 3.052,
      "step": 6871
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.3911599017766865e-06,
      "loss": 3.0582,
      "step": 6872
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.362270691896577e-06,
      "loss": 2.9568,
      "step": 6873
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.3333814820164674e-06,
      "loss": 2.86,
      "step": 6874
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.304492272136358e-06,
      "loss": 3.1061,
      "step": 6875
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.2756030622562475e-06,
      "loss": 3.0858,
      "step": 6876
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.246713852376138e-06,
      "loss": 2.994,
      "step": 6877
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.217824642496028e-06,
      "loss": 3.067,
      "step": 6878
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.188935432615918e-06,
      "loss": 2.9959,
      "step": 6879
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.160046222735808e-06,
      "loss": 2.8929,
      "step": 6880
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.131157012855699e-06,
      "loss": 3.0604,
      "step": 6881
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.102267802975588e-06,
      "loss": 3.0014,
      "step": 6882
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.073378593095479e-06,
      "loss": 2.997,
      "step": 6883
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.044489383215369e-06,
      "loss": 2.9582,
      "step": 6884
    },
    {
      "epoch": 0.98,
      "learning_rate": 4.015600173335259e-06,
      "loss": 3.0997,
      "step": 6885
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.986710963455149e-06,
      "loss": 3.0248,
      "step": 6886
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.95782175357504e-06,
      "loss": 3.1458,
      "step": 6887
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.92893254369493e-06,
      "loss": 2.9781,
      "step": 6888
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.900043333814821e-06,
      "loss": 3.1087,
      "step": 6889
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.871154123934711e-06,
      "loss": 2.7062,
      "step": 6890
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.842264914054601e-06,
      "loss": 3.1132,
      "step": 6891
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.813375704174491e-06,
      "loss": 3.1139,
      "step": 6892
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.7844864942943815e-06,
      "loss": 2.9154,
      "step": 6893
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.7555972844142715e-06,
      "loss": 2.9927,
      "step": 6894
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.726708074534162e-06,
      "loss": 3.1764,
      "step": 6895
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.697818864654052e-06,
      "loss": 3.0163,
      "step": 6896
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.668929654773942e-06,
      "loss": 2.9691,
      "step": 6897
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.6400404448938324e-06,
      "loss": 2.9428,
      "step": 6898
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.6111512350137224e-06,
      "loss": 3.0541,
      "step": 6899
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.582262025133613e-06,
      "loss": 3.0231,
      "step": 6900
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.553372815253503e-06,
      "loss": 3.0052,
      "step": 6901
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.524483605373393e-06,
      "loss": 2.9979,
      "step": 6902
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.4955943954932833e-06,
      "loss": 2.8792,
      "step": 6903
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.4667051856131733e-06,
      "loss": 3.1465,
      "step": 6904
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.4378159757330638e-06,
      "loss": 3.148,
      "step": 6905
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.408926765852954e-06,
      "loss": 3.0153,
      "step": 6906
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.3800375559728447e-06,
      "loss": 3.0236,
      "step": 6907
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.3511483460927347e-06,
      "loss": 3.0493,
      "step": 6908
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.322259136212625e-06,
      "loss": 3.0399,
      "step": 6909
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.293369926332515e-06,
      "loss": 3.0536,
      "step": 6910
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.2644807164524056e-06,
      "loss": 3.0736,
      "step": 6911
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.2355915065722956e-06,
      "loss": 3.1185,
      "step": 6912
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.2067022966921856e-06,
      "loss": 3.0759,
      "step": 6913
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.177813086812076e-06,
      "loss": 2.9949,
      "step": 6914
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.148923876931966e-06,
      "loss": 2.9747,
      "step": 6915
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.120034667051856e-06,
      "loss": 3.0707,
      "step": 6916
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.0911454571717465e-06,
      "loss": 3.0587,
      "step": 6917
    },
    {
      "epoch": 0.98,
      "learning_rate": 3.0622562472916365e-06,
      "loss": 3.1504,
      "step": 6918
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.033367037411527e-06,
      "loss": 3.1535,
      "step": 6919
    },
    {
      "epoch": 0.99,
      "learning_rate": 3.004477827531417e-06,
      "loss": 2.9331,
      "step": 6920
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.9755886176513074e-06,
      "loss": 2.8455,
      "step": 6921
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.946699407771198e-06,
      "loss": 2.8775,
      "step": 6922
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.917810197891088e-06,
      "loss": 2.9542,
      "step": 6923
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.888920988010978e-06,
      "loss": 3.0825,
      "step": 6924
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.8600317781308683e-06,
      "loss": 2.9671,
      "step": 6925
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.8311425682507583e-06,
      "loss": 3.0341,
      "step": 6926
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.8022533583706487e-06,
      "loss": 3.0661,
      "step": 6927
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.7733641484905388e-06,
      "loss": 3.0443,
      "step": 6928
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.7444749386104288e-06,
      "loss": 2.8629,
      "step": 6929
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.7155857287303196e-06,
      "loss": 3.1635,
      "step": 6930
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.6866965188502097e-06,
      "loss": 3.1478,
      "step": 6931
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.6578073089700997e-06,
      "loss": 2.8978,
      "step": 6932
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.62891809908999e-06,
      "loss": 2.9421,
      "step": 6933
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.60002888920988e-06,
      "loss": 3.0774,
      "step": 6934
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5711396793297706e-06,
      "loss": 3.1546,
      "step": 6935
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5422504694496606e-06,
      "loss": 3.0412,
      "step": 6936
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.5133612595695506e-06,
      "loss": 3.2098,
      "step": 6937
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.484472049689441e-06,
      "loss": 3.1855,
      "step": 6938
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.4555828398093315e-06,
      "loss": 2.9524,
      "step": 6939
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.4266936299292215e-06,
      "loss": 3.0207,
      "step": 6940
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.397804420049112e-06,
      "loss": 3.0075,
      "step": 6941
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.368915210169002e-06,
      "loss": 2.9822,
      "step": 6942
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.3400260002888924e-06,
      "loss": 3.1671,
      "step": 6943
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.3111367904087824e-06,
      "loss": 3.0587,
      "step": 6944
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.2822475805286724e-06,
      "loss": 2.9804,
      "step": 6945
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.253358370648563e-06,
      "loss": 3.0,
      "step": 6946
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.224469160768453e-06,
      "loss": 3.1084,
      "step": 6947
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.1955799508883433e-06,
      "loss": 2.9242,
      "step": 6948
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.1666907410082337e-06,
      "loss": 3.1705,
      "step": 6949
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.1378015311281237e-06,
      "loss": 2.9616,
      "step": 6950
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.108912321248014e-06,
      "loss": 3.0162,
      "step": 6951
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.080023111367904e-06,
      "loss": 3.0889,
      "step": 6952
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.051133901487794e-06,
      "loss": 3.18,
      "step": 6953
    },
    {
      "epoch": 0.99,
      "learning_rate": 2.0222446916076846e-06,
      "loss": 2.9249,
      "step": 6954
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9933554817275746e-06,
      "loss": 3.0066,
      "step": 6955
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.964466271847465e-06,
      "loss": 3.0889,
      "step": 6956
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9355770619673555e-06,
      "loss": 3.0842,
      "step": 6957
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.9066878520872455e-06,
      "loss": 2.9891,
      "step": 6958
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8777986422071358e-06,
      "loss": 2.9088,
      "step": 6959
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.848909432327026e-06,
      "loss": 3.0368,
      "step": 6960
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.8200202224469162e-06,
      "loss": 3.0724,
      "step": 6961
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7911310125668064e-06,
      "loss": 3.0317,
      "step": 6962
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7622418026866964e-06,
      "loss": 3.0485,
      "step": 6963
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.7333525928065867e-06,
      "loss": 3.0453,
      "step": 6964
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.704463382926477e-06,
      "loss": 3.0199,
      "step": 6965
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.6755741730463673e-06,
      "loss": 3.0125,
      "step": 6966
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.6466849631662576e-06,
      "loss": 2.9638,
      "step": 6967
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.6177957532861478e-06,
      "loss": 3.0054,
      "step": 6968
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.588906543406038e-06,
      "loss": 2.8664,
      "step": 6969
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.560017333525928e-06,
      "loss": 2.8927,
      "step": 6970
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5311281236458183e-06,
      "loss": 2.8634,
      "step": 6971
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.5022389137657085e-06,
      "loss": 2.9638,
      "step": 6972
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.473349703885599e-06,
      "loss": 3.0843,
      "step": 6973
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.444460494005489e-06,
      "loss": 3.1211,
      "step": 6974
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.4155712841253792e-06,
      "loss": 3.0891,
      "step": 6975
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.3866820742452694e-06,
      "loss": 3.0453,
      "step": 6976
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.3577928643651598e-06,
      "loss": 2.896,
      "step": 6977
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.3289036544850498e-06,
      "loss": 3.1392,
      "step": 6978
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.30001444460494e-06,
      "loss": 2.8895,
      "step": 6979
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.2711252347248303e-06,
      "loss": 3.0352,
      "step": 6980
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.2422360248447205e-06,
      "loss": 2.9906,
      "step": 6981
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.2133468149646107e-06,
      "loss": 2.8133,
      "step": 6982
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.184457605084501e-06,
      "loss": 3.0085,
      "step": 6983
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1555683952043912e-06,
      "loss": 2.9782,
      "step": 6984
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.1266791853242814e-06,
      "loss": 3.1336,
      "step": 6985
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0977899754441716e-06,
      "loss": 2.9151,
      "step": 6986
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.0689007655640619e-06,
      "loss": 3.0004,
      "step": 6987
    },
    {
      "epoch": 0.99,
      "learning_rate": 1.040011555683952e-06,
      "loss": 2.9352,
      "step": 6988
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.0111223458038423e-06,
      "loss": 3.1152,
      "step": 6989
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.822331359237325e-07,
      "loss": 2.9697,
      "step": 6990
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.533439260436228e-07,
      "loss": 3.1496,
      "step": 6991
    },
    {
      "epoch": 1.0,
      "learning_rate": 9.24454716163513e-07,
      "loss": 2.913,
      "step": 6992
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.955655062834032e-07,
      "loss": 3.0368,
      "step": 6993
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.666762964032933e-07,
      "loss": 2.9764,
      "step": 6994
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.377870865231837e-07,
      "loss": 3.0714,
      "step": 6995
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.088978766430739e-07,
      "loss": 3.0611,
      "step": 6996
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.80008666762964e-07,
      "loss": 3.0141,
      "step": 6997
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.511194568828542e-07,
      "loss": 2.9486,
      "step": 6998
    },
    {
      "epoch": 1.0,
      "learning_rate": 7.222302470027445e-07,
      "loss": 3.0568,
      "step": 6999
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.933410371226347e-07,
      "loss": 2.9752,
      "step": 7000
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.644518272425249e-07,
      "loss": 3.0737,
      "step": 7001
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.355626173624151e-07,
      "loss": 3.0703,
      "step": 7002
    },
    {
      "epoch": 1.0,
      "learning_rate": 6.066734074823054e-07,
      "loss": 2.9669,
      "step": 7003
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.777841976021956e-07,
      "loss": 3.0173,
      "step": 7004
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.488949877220858e-07,
      "loss": 3.0607,
      "step": 7005
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.20005777841976e-07,
      "loss": 3.0821,
      "step": 7006
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.911165679618663e-07,
      "loss": 3.0936,
      "step": 7007
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.622273580817565e-07,
      "loss": 3.0225,
      "step": 7008
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.3333814820164667e-07,
      "loss": 3.1118,
      "step": 7009
    },
    {
      "epoch": 1.0,
      "learning_rate": 4.0444893832153695e-07,
      "loss": 2.9692,
      "step": 7010
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.755597284414271e-07,
      "loss": 2.9797,
      "step": 7011
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.4667051856131735e-07,
      "loss": 3.0496,
      "step": 7012
    },
    {
      "epoch": 1.0,
      "learning_rate": 3.1778130868120757e-07,
      "loss": 3.0703,
      "step": 7013
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.888920988010978e-07,
      "loss": 3.0217,
      "step": 7014
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.60002888920988e-07,
      "loss": 3.0717,
      "step": 7015
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.3111367904087825e-07,
      "loss": 2.8553,
      "step": 7016
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.0222446916076847e-07,
      "loss": 3.0506,
      "step": 7017
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.7333525928065867e-07,
      "loss": 3.0767,
      "step": 7018
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.444460494005489e-07,
      "loss": 2.9865,
      "step": 7019
    },
    {
      "epoch": 1.0,
      "learning_rate": 1.1555683952043912e-07,
      "loss": 3.0635,
      "step": 7020
    },
    {
      "epoch": 1.0,
      "learning_rate": 8.666762964032934e-08,
      "loss": 3.1211,
      "step": 7021
    },
    {
      "epoch": 1.0,
      "learning_rate": 5.777841976021956e-08,
      "loss": 3.0291,
      "step": 7022
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.888920988010978e-08,
      "loss": 2.9262,
      "step": 7023
    }
  ],
  "logging_steps": 1,
  "max_steps": 7023,
  "num_train_epochs": 1,
  "save_steps": 500,
  "total_flos": 6.982535499177001e+18,
  "trial_name": null,
  "trial_params": null
}